CN105184072A - Data interpolation method and apparatus - Google Patents

Abstract

The invention discloses a data interpolation method and apparatus. The method comprises: obtaining multiple first data and multiple second data within a target time period, wherein the first data is data measured by a wind measurement tower, and the second data is meteorological analytical data; establishing a target network model according to the first data and the second data; and obtaining target data by calculation according to the target network model, wherein the target data is data that needs to be interpolated in wind measurement data within the target time period. According to the method and the apparatus disclosed by the invention, the technical problem of inability to obtain missing wind measurement data by using relevance of contiguous available data in time and space in the prior art caused by long-time data missing is solved.

Description

The method and apparatus of data interpolation

Technical field

The present invention relates to data processing field, in particular to a kind of method and apparatus of data interpolation.

Background technology

In wind-powered electricity generation construction in early stage, need to set up anemometer tower to be no less than the observation of 1 year to weather conditions in target construction area, then the wind-resources assessment of target construction area is carried out according to observation data, determine whether this region has Development volue, if have Development volue, how carrying out engineering design, the layout especially how designing Wind turbines could realize maximization of economic benefit.Observation process (hereinafter referred to as survey wind) has vital effect for the construction of wind energy turbine set and the realization of economic benefit.But continuing to be not less than in the survey wind of 1 year, often due to weather extremes, environment reason and the technical failure reason such as freezing, making survey wind data imperfect.If indivedual measuring point missing data at short notice, valuation can be carried out with its correlativity to it by time and (or) space adjacent other normal data, and then interpolation is carried out to missing data.If but whole anemometer tower long term deletion data, then above-mentioned technological means cannot be adopted to carry out interpolation, cause permanent data to lack.In this case, if missing data is more, continuous 1 year complete survey wind data is obtained, the survey wind of 1 year by a definite date can only be re-started, larger time cost and financial cost can be produced thus, even cause the target of the wind-powered electricity generation construction of expection not realize.

Current wind energy turbine set early development adopts anemometer tower to survey the mode of wind substantially, and this mode comprises following content:

Measuring position requires to install with anemometer tower; Surveying instrument technical requirement and surveying instrument are installed; Survey wind data require and survey wind data process.National standard requirement, in-site measurement should carry out continuously, should not be less than 1 year; The measurement data percentage of head rice of collection in worksite should more than 98%.Current anemometer tower apart from 10 meters, ground, 30 meters, 50 meters and hub height install anemometry, apart from ground 10 meter ampere dress temperature, air pressure, relative humidity measurement equipment, installing wind direction measuring equipment apart from 10 meters, ground and top layer.If indivedual measuring point missing data at short notice, valuation can be carried out with its correlativity to it by time and (or) space adjacent other normal data, and then interpolation is carried out to missing data.

Adopt current anemometer tower to survey the technical scheme of wind, often due to weather extremes, environment reason and the technical failure reason such as freezing, make survey wind data imperfect.If indivedual measuring point is at short time missing data, valuation can be carried out with its correlativity to it by time and (or) space adjacent other normal data, and then interpolation is carried out to missing data.If but whole anemometer tower long term deletion data, then above-mentioned technological means cannot be adopted to carry out interpolation, cause permanent data to lack.If the integrality surveying wind data is less than 98%, then do not meet national standard.

For above-mentioned problem, at present effective solution is not yet proposed.

Summary of the invention

Embodiments provide a kind of method and apparatus of data interpolation, at least to solve because long missing data causes the correlativity that can not utilize adjacent valid data in the time and space in prior art to obtain lacking the technical matters surveying wind data.

According to an aspect of the embodiment of the present invention, provide a kind of method of data interpolation, comprise: obtain the first data in object time section and the second data, wherein, the quantity of described first data and described second data is multiple, described first data are the data that anemometer tower is measured, and described second data are meteorologic analysis data; Objective network model is built according to described first data and described second data; And calculate target data according to described objective network model, wherein, described target data is the data needing interpolation in described object time section.

Further, after the first data obtained in object time section and the second data, described method also comprises: described first data are pre-conditionedly converted to the first sample according to first, and described second data are pre-conditionedly converted to the second sample according to second, build objective network model according to described first data and described second data and comprise: build objective network model according to described first sample and described second sample.

Further, multiple described first data are included in the data that the multiple first object moment in described object time section collect, and described first data are comprised according to the first pre-conditioned first sample that is converted to: read first object moment t in multiple described first object moment respectively _hthe first parameter and the second parameter, wherein, described first Parametric Representation is the wind speed of described anemometer tower at multiple first object height, described second Parametric Representation is the wind direction of described anemometer tower at described multiple first object height, h gets the positive integer of 1 to m1 successively, and m1 is the quantity in described first object moment; According to formula described first parameter and described second parameter are converted into primary vector u _fwith secondary vector v _f, wherein, S _ffor described first parameter, for described second parameter, f gets the positive integer of 1 to m2 successively, and m2 is the quantity of described first object height; By described first object moment t _hbe converted into the moment of object format; And will the described first object moment t after described object format be converted into _hand described first object moment t _hdescribed first parameter and the 3rd parameter form First ray, wherein, described 3rd parameter comprises following at least one: described anemometer tower is in the temperature of the second object height, air pressure and relative humidity, and the quantity of described First ray is m2, and multiple described First ray forms described first sample.

Further, multiple described second data are included in the data that multiple second object time in described object time section collect, the quantity of described second object time is greater than the quantity in described first object moment, described second data is comprised according to the second pre-conditioned second sample that is converted to: read described second data at the second object time t _kthe 4th parameter, wherein, k gets the positive integer of 1 to m3 successively, and m3 is the quantity in described second order moment, and the quantity of described 4th parameter is multiple; The 5th parameter Aij of described second sample is calculated according to described 4th parameter, wherein, described 5th parameter Aij at least comprises following at least one: wind speed, megadyne temperature, air pressure and vapor-to-liquid ratio, i gets the positive integer of 1 to a successively, a is the quantity of horizontal level, j gets the positive integer of 1 to b successively, and b is the quantity of differing heights on each described horizontal level, the parameter of the jth of described 5th Parametric Representation on an i-th horizontal level height; And by described second object time t _kdescribed 5th parameter Aij and described second object time t _kform the second sequence, wherein, the second sample described in multiple described second Sequence composition.

Further, described first data are pre-conditionedly being converted to the first sample according to first, and by described second data according to second pre-conditioned be converted to the second sample after, described method comprises: described second sample is split into first object sample and the second target sample, builds objective network model comprise according to described first data and described second data: build described objective network model according to described first object sample and described first sample.

Further, described second sample is split into first object sample and the second target sample comprises: extract very first time component and the second time component, wherein, described very first time component is the first row of described first sample, and described second time component is the secondary series of described second sample; Judge the described second object time t in described second time component _kwhether comprise the described first object moment t in described very first time component _h, obtaining judged result, wherein, is described second object time t in described judged result _kcomprise described first object moment t _hwhen, described second object time t _kbeing set to 1, is described second object time t in described judged result _kdo not comprise described first object moment t _hwhen, described second object time t _kbe set to 0; Upgrade described second time component according to described judged result, obtain the 3rd time component; Described second time component and described 3rd time component are carried out point multiplication operation, obtains the 4th time component; And described 4th time component is added in the 3rd sample, obtain the 4th sample, wherein, identical with described first sample in described 4th sample is described first object sample, not identical with described first sample in described 4th sample is described second target sample, and described 3rd sample is the sample that described second sample deletes after described second time component.

Further, build objective network model according to described first object sample and described second sample to comprise: delete the described 4th time component in described first object sample and the described very first time component in described first sample, wherein, the 5th time component is the first row of described first object sample; Respectively the described first object sample deleting described 5th time component is normalized calculating, the described first object sample after obtaining normalized and described first sample with described first sample deleting described very first time component; And the described first object sample after being normalized and described first sample are loaded in target program, to make described first object sample to the training of described objective network model, wherein, after training completes, described objective network model has carried.

Further, calculate target data according to described objective network model to comprise: described second target sample is updated in target simulator function, to make described target simulator function carry out analog computation in described objective network model, obtain first object result; Described first object result is carried out return normalization to calculate, obtain the second objective result; Described second objective result is added very first time component, forms the 3rd objective result; And described first sample and described 3rd objective result are carried out merging obtain described target data.

According to the another aspect of the embodiment of the present invention, additionally provide a kind of device of data interpolation, it is characterized in that, comprise: acquiring unit, for obtaining the first data in object time section and the second data, wherein, the quantity of described first data and described second data is multiple, described first data are the data that anemometer tower is measured, and described second data are meteorologic analysis data; Build unit, for building objective network model according to described first data and described second data; And computing unit, for calculating target data according to described objective network model, wherein, described target data is the data needing interpolation in described object time section.

Further, described device also comprises: converting unit, for after the first data obtained in object time section and the second data, described first data are pre-conditionedly converted to the first sample according to first, and described second data are pre-conditionedly converted to the second sample according to second, described unit of building comprises: first builds module, for building objective network model according to described first sample and described second sample.

Further, multiple described first data are included in the data that the multiple first object moment in described object time section collect, and described converting unit comprises: the first read module, for reading first object moment t in multiple described first object moment respectively _hthe first parameter and the second parameter, wherein, described first Parametric Representation is the wind speed of described anemometer tower at multiple first object height, described second Parametric Representation is the wind direction of described anemometer tower at described multiple first object height, h gets the positive integer of 1 to m1 successively, and m1 is the quantity in described first object moment; First conversion module, for according to formula described first parameter and described second parameter are converted into primary vector u _fwith secondary vector v _f, wherein, S _ffor described first parameter, for described second parameter, f gets the positive integer of 1 to m2 successively, and m2 is the quantity of described first object height; Second conversion module, for by described first object moment t _hbe converted into the moment of object format; And first composition module, for will the described first object moment t after described object format be converted into _hand described first object moment t _hdescribed first parameter and the 3rd parameter form First ray, wherein, described 3rd parameter comprises following at least one: described anemometer tower is in the temperature of the second object height, air pressure and relative humidity, and the quantity of described First ray is m2, and multiple described First ray forms described first sample.

Further, multiple described second data are included in the data that multiple second object time in described object time section collect, the quantity of described second object time is greater than the quantity in described first object moment, described converting unit comprises: the second read module, for reading described second data at the second object time t _kthe 4th parameter, wherein, k gets the positive integer of 1 to m3 successively, and m3 is the quantity in described second order moment, and the quantity of described 4th parameter is multiple; First computing module, for calculating the 5th parameter Aij of described second sample according to described 4th parameter, wherein, described 5th parameter Aij at least comprises following at least one: wind speed, megadyne temperature, air pressure and vapor-to-liquid ratio, i gets the positive integer of 1 to a successively, and a is the quantity of horizontal level, and j gets the positive integer of 1 to b successively, b is the quantity of differing heights on each described horizontal level, the parameter of the jth of described 5th Parametric Representation on an i-th horizontal level height; And second composition module, for by described second object time t _kdescribed 5th parameter Aij and described second object time t _kform the second sequence, wherein, the second sample described in multiple described second Sequence composition.

Further, described device comprises: split cells, for described first data being pre-conditionedly converted to the first sample according to first in described converting unit, and by described second data according to second pre-conditioned be converted to the second sample after, described second sample is split into first object sample and the second target sample, described unit of building comprises: second builds module, for building described objective network model according to described first object sample and described first sample.

Further, described split cells comprises: extraction module, and for extracting very first time component and the second time component, wherein, described very first time component is the first row of described first sample, and described second time component is the secondary series of described second sample; Judge module, for judging the described second object time t in described second time component _kwhether comprise the described first object moment t in described very first time component _h, obtaining judged result, wherein, is described second object time t in described judged result _kcomprise described first object moment t _hwhen, described second object time t _kbeing set to 1, is described second object time t in described judged result _kdo not comprise described first object moment t _hwhen, described second object time t _kbe set to 0; Update module, for upgrading described second time component according to described judged result, obtains the 3rd time component; Second computing module, for described second time component and described 3rd time component are carried out dot product calculating, obtains the 4th time component; And first adds module, for described 4th time component is added in the 3rd sample, obtain the 4th sample, wherein, identical with described first sample in described 4th sample is described first object sample, not identical with described first sample in described 4th sample is described second target sample, and described 3rd sample is the sample that described second sample deletes after described second time component.

Further, described second builds module comprises: delete submodule, for deleting the 5th time component in described first object sample and the described very first time component in described first sample, wherein, described 5th time component is the first row of described first object sample; First calculating sub module, for respectively the described first object sample deleting described 5th time component being normalized calculating, the described first object sample after obtaining normalized and described first sample with described first sample deleting described very first time component; And loading submodule, for the described first object sample after being normalized and described first sample are loaded in target program, to make described first object sample to the training of described objective network model, wherein, after training completes, described objective network model has carried.

Further, described computing unit comprises: substituting into module, for being updated in target simulator function by described second target sample, to make described target simulator function carry out analog computation in described objective network model, obtaining first object result; 3rd computing module, calculating for described first object result being carried out return normalization, obtaining the second objective result; Second adds module, for described second objective result is added very first time component, forms the 3rd objective result; And merging module, obtain described target data for described first sample and described 3rd objective result are carried out merging.

In embodiments of the present invention, adopt the first data and the second data that obtain in object time section, wherein, the quantity of described first data and described second data is multiple, described first data are the survey wind data that described anemometer tower is measured, and described second data are the meteorologic survey data of anemometer tower; Objective network model is built according to described first data and described second data; And calculate target data according to described objective network model, wherein, described target data is the mode needing the data of interpolation in described object time section in described survey wind data.By obtaining survey wind data in the same time period and meteorological measurement data (namely, mesoscale data), and build objective network model according to survey wind data and mesoscale data, and by Meso-Scale Analysis data with effectively survey wind data to the training of objective network model, obtain the target data in the time period of surveying wind data disappearance, reach the object of the interpolation of surveying wind missing data, thus the valid data achieved in utilization survey wind data and objective network model are to simulate the technique effect of the survey wind data of disappearance, and then solve because long missing data causes the correlativity that can not utilize adjacent valid data in the time and space in prior art to obtain lacking the technical matters surveying wind data.

Accompanying drawing explanation

Accompanying drawing described herein is used to provide a further understanding of the present invention, and form a application's part, schematic description and description of the present invention, for explaining the present invention, does not form inappropriate limitation of the present invention.In the accompanying drawings:

Fig. 1 is the process flow diagram of the method for a kind of optional data interpolation according to the embodiment of the present invention;

Fig. 2 is the process flow diagram of the method according to the optional data interpolation of the another kind of the embodiment of the present invention; And

Fig. 3 is the schematic diagram of the device of a kind of data interpolation according to the embodiment of the present invention.

Embodiment

The present invention program is understood better in order to make those skilled in the art person, below in conjunction with the accompanying drawing in the embodiment of the present invention, technical scheme in the embodiment of the present invention is clearly and completely described, obviously, described embodiment is only the embodiment of a part of the present invention, instead of whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art, not making the every other embodiment obtained under creative work prerequisite, should belong to the scope of protection of the invention.

It should be noted that, term " first ", " second " etc. in instructions of the present invention and claims and above-mentioned accompanying drawing are for distinguishing similar object, and need not be used for describing specific order or precedence.Should be appreciated that the data used like this can be exchanged in the appropriate case, so as embodiments of the invention described herein can with except here diagram or describe those except order implement.In addition, term " comprises " and " having " and their any distortion, intention is to cover not exclusive comprising, such as, contain those steps or unit that the process of series of steps or unit, method, system, product or equipment is not necessarily limited to clearly list, but can comprise clearly do not list or for intrinsic other step of these processes, method, product or equipment or unit.

According to the embodiment of the present invention, provide a kind of embodiment of the method for method of data interpolation, it should be noted that, can perform in the computer system of such as one group of computer executable instructions in the step shown in the process flow diagram of accompanying drawing, and, although show logical order in flow charts, in some cases, can be different from the step shown or described by order execution herein.

Fig. 1 is the process flow diagram of the method for a kind of optional data interpolation according to the embodiment of the present invention, and as shown in Figure 1, the method comprises the steps that S102 is to step S106:

Step S102, obtains the first data in object time section and the second data, and wherein, the quantity of the first data and the second data is multiple, and the first data are the data that anemometer tower is measured, and the second data are meteorologic analysis data.

It should be noted that, in embodiments of the present invention, the first data are the survey wind data of anemometer tower, and the second data are the mesoscale data of anemometer tower.Be chosen for according to national regulation object time section and be no less than year section.

Step S104, builds objective network model according to the first data and the second data.

It should be noted that, the objective network model in the embodiment of the present invention comprises the models such as neural network model, PSO Neural Network model, simulated annealing and genetic algorithm.Wherein, objective network model is built by the first data and the second data in simulation software, concrete simulation software is the simulation softwares such as Matlab, Matlab is as the software of general emulation, there is powerful data processing function, therefore in embodiments of the present invention, all for Matlab simulation software, building of objective network model is introduced.

Step S106, calculates target data according to objective network model, and wherein, target data is the data needing interpolation in object time section.

It should be noted that, the survey wind data in object time section causes data imperfect due to a variety of causes such as weather, causes permanent loss, and the data of this disappearance are the data surveyed and need interpolation in wind data.

In embodiments of the present invention, by obtaining survey wind data in the same time period and mesoscale data, and build objective network model according to survey wind data and mesoscale data, and by Meso-Scale Analysis data with effectively survey wind data to the training of objective network model, obtain the target data in the time period of surveying wind data disappearance, reach the object of the interpolation of surveying wind missing data, thus achieve the technique effect of the survey wind data of simulating disappearance according to the correlativity of adjacent valid data, and then solve because long missing data causes the correlativity that can not utilize adjacent valid data in the time and space in prior art to obtain lacking the technical matters surveying wind data.

Alternatively, after step S102 obtains the first data in object time section and the second data, the method for data interpolation provided by the invention also comprises the steps S1:

Step S1, first data are pre-conditionedly converted to the first sample according to first, and the second data are pre-conditionedly converted to the second sample according to second, build objective network model according to the first data and the second data and comprise: build objective network model according to the first sample and the second sample.

After the data acquisition unit in anemometer tower gets the first data and the second data, the first data and the second data are all stored in storer with the form of fixed text.When the first data and the second data build objective network model in Matlab, need be the text formatting of Matlab simulation software acquiescence by the first data and the second data transformations, the such as text of " .dat " form.Therefore when selecting Matlab simulation software, need by the first data according to first pre-conditioned and the second data according to the second pre-conditioned text being converted into " .dat " form.

And then in Matlab simulation software, build objective network model according to the first sample after conversion and the second sample.It should be noted that, if the simulation software selected is the software beyond Matlab simulation software, then needing the first data and the second data transformations is the text formatting corresponding with this simulation software.

Further, multiple first data are included in the data that the multiple first object moment in object time section collect, according to the first pre-conditioned first sample that is converted to, the first data are comprised the steps that S11 is to step S14 by step S1:

Step S11, reads first object moment t in multiple first object moment respectively _hthe first parameter and the second parameter, wherein, the first Parametric Representation is the wind speed of anemometer tower at multiple first object height, and the second Parametric Representation is the wind direction of anemometer tower at multiple first object height, h gets the positive integer of 1 to m1 successively, and m1 is the quantity in first object moment.

First data, the survey wind data of also i.e. anemometer tower measurement, divided into groups to the first data according to multiple first object moment.Wherein, each group data comprises the wind speed and direction apart from ground differing heights, wind speed S _f, be, the first parameter; Wind direction be, the second parameter.Specifically choose multiple first object height according to national regulation, multiple first object height comprises distance 10 meters, ground, 30 meters, 50 meters and these 4 height of anemometer tower top layer.

Step S12, according to formula first parameter and the second parameter are converted into primary vector u _fwith secondary vector v _f, wherein, S _fbe the first parameter, be the second parameter, f gets the positive integer of 1 to m2 successively, and m2 is the quantity of first object height.

According to above-mentioned formula wind speed S _fbe converted into the vector of both direction, i.e. primary vector u _fwith secondary vector v _f, wherein, primary vector u _ffor in first object height t _hthe wind velocity vector of upper east-west direction, secondary vector v _ffor in first object height t _hthe wind velocity vector of upper North and South direction.

Step S13, by first object moment t _hbe converted into the moment of object format.

First object moment t _hobject format be " YYYYMMDDHHMI ", wherein, " YYYY " 4 bit digital represent year, " MM " 2 bit digital represent the moon, " DD " 2 bit digital represent day, " HH " 2 bit digital to represent hour, " MI " 2 is numeral minute.Suppose the first object moment t chosen ₁for " 201501010000 ", namely from the morning on January 1st, 2015, gather the first data, gather a secondary data, then first object moment t every 10 minutes ₂be expressed as " 201501010010 ", then t ₃, t ₄..., t _m1first object moment t can be drawn by that analogy below _hobject format.

Step S14, will be converted into the first object moment t after object format _hand first object moment t _hthe first parameter and the 3rd parameter form First ray, wherein, the 3rd parameter comprises following at least one: anemometer tower is in the temperature of the second object height, air pressure and relative humidity, and the quantity of First ray is m2, and multiple First ray forms the first sample.

It should be noted that, in the first data except the first parameter and the second parameter, also comprise the 3rd parameter, the 3rd parameter can be the temperature of anemometer tower at the second object height, represents with symbol T; Air pressure, represents for symbol P; Relative humidity, represents with symbol R, and wherein, the quantity of the second object height is one.Particularly, at first object moment t _hthe first parameter of first object height layer and the 3rd parameter of the second object height layer and first object moment t _hform First ray, wherein, First ray is expressed as: A=(t _h, u _f, v _f, T, P, R), multiple First ray forms the first sample A'.When the value of f is 4, sequence A is for comprising 12 components, and the first sample A' is the matrix of m2 × 12, and the file according to the file of the first sample generation is called: windsample.dat.It should be noted that, in this first sample, do not comprise the data needing interpolation, i.e. target data, the data in the first sample are valid data, namely can the data that arrive of Obtaining Accurate by anemometer tower.

Alternatively, multiple second data are included in the data that multiple second object time in object time section collect, the quantity of the second object time is greater than the quantity in first object moment, according to the second pre-conditioned second sample that is converted to, step S1 just the second data comprises the steps that S15 is to step S17:

Step S15, reads the second data at the second object time t _kthe 4th parameter, wherein, k gets the positive integer of 1 to m3 successively, and m3 is the quantity of the second object time, and the quantity of the 4th parameter is multiple.

It should be noted that, second data, also be mesoscale data, the horizontal resolution of mesoscale data can reach 1.5 kilometers (namely lattice point distance is not more than 1.5 kilometers), the vertical direction number of plies is not less than 10, wherein be no less than 3 layers apart from floor level less than 150 meters, maximum height is not less than 1000 meters, and time granularity is 10 minutes.The 4th parameter in second data at least comprise following any one: level sequence number, represents with bottom_top; The north and south sequence number of lattice point, represents with south_north; The thing sequence number of lattice point, represents with west_east; The wind speed component of east-west direction, represents with U; The wind speed component of North and South direction, represents with V; The wind speed component of vertical direction, represents with W; Latitude, represents with XLAT; Longitude, represents with XLONG; Disturbance potential, represents with PH; Ground state potential, represents with PHB; Floor level, represents with HGT; Disturbance megadyne temperature, represents with T; Disturbance, represents with P; Ground state air pressure, represents with PB; Vapor-to-liquid ratio, represents with QVAPOR; 2 meters, ground height megadyne temperature, represents with TH2.

Step S16, the 5th parameter Aij of the second sample is calculated according to the 4th parameter, wherein, 5th parameter Aij at least comprises following at least one: wind speed, megadyne temperature, air pressure and vapor-to-liquid ratio, i gets the positive integer of 1 to a successively, and a is the quantity of horizontal level, and j gets the positive integer of 1 to b successively, b is the quantity of differing heights on each horizontal level, the parameter of the jth of the 5th Parametric Representation on an i-th horizontal level height.

Calculate the 5th parameter Aij in the second sample according to the 4th parameter, wherein, the 5th parameter Aij comprise according to object format transform after the second object time t _k, the u of jth the height of 3 components on i-th horizontal level of wind speed _ij, v _ij, w _ij, megadyne temperature θ _ij, vapor-to-liquid ratio q _ijand air pressure p _ij.

Calculate the 5th parameter according to formula (1), wherein, the equation left side is the 4th parameter, is the 5th parameter on the right of equation.

$\left\{\begin{array}{c}{u}_{ij}=U_{ij},v_{ij}=V_{ij},w_{ij}=W_{ij}\\ {\mathrm{\θ}}_{ij}=T_{ij}+TH{2}_{ij}\\ p_{ij}=P_{ij}+{\mathrm{PB}}_{ij}\\ q_{ij}={\mathrm{QVAPOR}}_{ij}\end{array}\right.---\left(1\right)$

Wherein, U _ij, V _ijand W _ijbe respectively the wind speed component of the jth height on i-th horizontal level.T _ijbe expressed as the disturbance megadyne temperature of the jth height on i-th horizontal level, TH2 _ijbe expressed as 2 meters, the ground height megadyne temperature of the jth height on i-th horizontal level, P _ijbe expressed as the disturbance of the jth height on i-th horizontal level, PB _ijbe expressed as the ground state air pressure of the jth height on i-th horizontal level, QVAPOR _ijbe expressed as the vapor-to-liquid ratio of the jth height on i-th horizontal level.

Step S17, by the second object time t _kthe 5th parameter Aij and the second object time t _kform the second sequence, wherein, multiple second Sequence composition second sample.

Second object time t _kthe 5th parameter Aij and the second object time t _kthe second sequence formed is expressed as the second sequence: B=(t _k, u _ij, v _ij, w _ij, θ _ij, p _ij, q _ij), wherein, a is the sequence number of horizontal level, b is the sequence number of differing heights on each horizontal level, when the span of a and b is 1<a<3,1<b<3, the second sequence B comprises 55 components, individual second Sequence composition second sample B of m3 ', the file of the second sample file B' is called mesosample.dat.

Alternatively, in step S1, the first data are pre-conditionedly converted to the first sample according to first, and by the second data according to second pre-conditioned be converted to the second sample after, the method of data interpolation provided by the invention also comprises the steps S2: the second sample is split into first object sample and the second target sample, builds objective network model comprise according to the first data and the second data: build objective network model according to first object sample and the first sample.

It should be noted that, first object sample is used for the training of objective network model, and to build the objective network model of energy Accurate Prediction target data, the second target sample is used for analog computation and goes out target data.

Alternatively, step S2 comprises the steps that S21 is to step S25:

Step S21, extract very first time component and the second time component, wherein, very first time component is the first row of the first sample, and the second time component is the secondary series of the second sample.

By following program by the first sample A' and the second sample B ' be carried in Matlab program: step one: load document is called the first sample A':load (' mesosample.dat', ' A ") of windsample.dat; Step 2: load document is called the second sample B ': load (' the windsample.dat' of windsample.dat, ' B ").

Very first time component by following Program extraction first sample: T1=A'(:, 1); And by the second time component: T2=B'(in following Program extraction second sample:, 1).

Step S22, judges the second object time t in the second time component _kwhether comprise the first object moment t in very first time component _h, obtaining judged result, wherein, is the second object time t in judged result _kcomprise first object moment t _hwhen, the second object time t _kbeing set to 1, is the second object time t in judged result _kdo not comprise first object moment t _hwhen, the second object time t _kbe set to 0.

Step S23, upgrades the second time component according to judged result, obtains the 3rd time component.

Be 0 by following program by needing the time mark of interpolation data in the second time component:

T3＝ismember(T2，T1，'rows')；

Wherein, the quantity of the second time component is greater than the quantity of very first time component, and the second time component and very first time component have identical time component.The 3rd time component finally calculated be only comprise 1 and/or 0 time component.

Step S24, carries out point multiplication operation by the second time component and the 3rd time component, obtains the 4th time component.

Need the time of interpolation all to become 0, T2=T2.*T3 by following program by the second time component, wherein, the T2 on the equation left side is the 4th time component.

Step S25,4th time component is added in the 3rd sample, obtain the 4th sample, wherein, identical with the first sample in 4th sample is first object sample, not identical with the first sample in 4th sample is the second target sample, and the 3rd sample is the sample that the second sample deletes after the second time component.

By following program, is replaced with the time of mark interpolation period the time in the second sample,

C=B'; Wherein, the Matrix C on the equation left side is the 3rd sample.

C(：，1)＝[]；

C=[T2, C]; Wherein, the Matrix C on the equation left side is the 4th sample.

Present Matrix C is, on the basis of the second sample, the interpolation time period is labeled as 0, matrix B ' the part identical with C be exactly the second target sample, matrix B ' the part different from C be exactly first object sample, specifically, realized the fractionation of first object sample and the second target sample by following program:

First object sample: P=setdiff (B', C, ' rows');

Second target sample: I=setdiff (B', P, ' rows');

Respectively first object sample and the second target sample are saved as file according to following program.

save('trainsample.dat'，P)；

save('inputsample.dat'，I)；

Like this, first object sample is identical with the line number of the second target sample file, and the quantity of time component is consistent with time point.

Alternatively, build objective network model according to first object sample and the second sample and comprise the steps S3 to S5:

Step S3, delete the 5th time component in first object sample and the very first time component in the first sample, wherein, the 5th time component is the first row of first object sample.

Step S4, is normalized calculating, the first object sample after obtaining normalized and the first sample by the first object sample deleting the 5th time component with the first sample deleting very first time component respectively.

Step S5, be loaded in target program by the first object sample after being normalized and the first sample, to make first object sample to the training of objective network model, wherein, after training completes, objective network model has carried.

In embodiments of the present invention, objective network model is described for neural network model.

The description of following step is each option setting of neural network based on Matlab.

Step one: build neural network structure

Neural network structure is divided into input layer, hidden layer and output layer.Wherein, hidden layer is set to 3 layers, and the neuron of all adjacent two layers is carried out entirely interconnected at interlayer.3 layers of hidden layer respectively arrange 4,12,4 neurons.

Step 2: transport function is set

Particularly, hidden layer adopts hyperbolic tangent function tansig () to be transport function, and output layer adopts pure linear function purelin () to be transport function.

Step 3: training function is set

Particularly, learning algorithm selects momentum and adaptive learning rate method, and corresponding training function selects trainlm to train function.

Step 4: the implementation process that use MATLAB programming language builds neural network model is as follows:

A, set up first object sample matrix

Perform instruction and carry out Data import: load (' trainsample.dat', ' P'), wherein P is expressed as first object sample, and the file name of first object sample is trainsample.dat.

Delete the 1st row of the matrix of first object sample, i.e. the time component of first object sample: P (:, 1)=[].

B, set up the second target sample matrix

Perform instruction and carry out Data import: load (' inputsample.dat', ' I'), wherein I is expressed as the second target sample, and the file name of the second target sample is inputsample.dat.

Delete the 1st row of the matrix of the second target sample, i.e. the time component of the second target sample: I (:, 1)=[].

C, set up expected response data

Perform instruction and carry out Data import: load (' windsample.dat', ' A "), wherein, matrix A ' be the matrix of the expectation corresponding data that the valid data in the first sample are formed, the name of this matrix is called windsample.dat.

Puncture table the 1st row, matrix A ' time component: A'(:, 1)=[];

D, data normalization

By as follows for the normalization process of first object sample matrix:

First object sample is splitted into two parts: Part I comprises wind speed, P1=P (:, 1:27); Part II comprises air pressure, the temperature and humidity of the second object height, P2=P (:, 28:54), wherein, the quantity i of the horizontal level chosen is 3, the quantity of the height j on each horizontal level chosen is 3, therefore the first object sample P not with time component is the matrix samples of N1 × 54, wherein, front 27 samples being classified as wind speed, 28th row are respectively the air pressure of the second object height, the sample of temperature and humidity to 54 row, and wherein, N1 is the quantity of the time component of first object sample.

Respectively Part I P1 and Part II P2 is normalized calculating, the detailed process of calculating is the maximal value of each component divided by all wind speed components of wind speed part, and specific formula for calculation (2) is expressed as:

$\left\{\begin{array}{c}a=max\left(max\right(P1\left)\right)\\ P{1}^{\&prime;}=\frac{1}{a}*\left(P1\right)\end{array}\right.---\left(2\right)$

Wherein, a is the maximal value of Part I P1, by each component of Part I divided by maximal value a, is Part I first object sample after normalization calculates, represents with P1'.

The air pressure of the second object height, temperature, humidifying part adopt standard normalization algorithm, and specific formula for calculation (3) is:

[P2',ps]＝mapminmax(P2)(3)

Wherein, the Part II normalization in first object sample represents with P2' after calculating, Part I P1' and Part II P2' is merged the first object sample P after obtaining normalized, wherein, the first object sample after normalized is expressed as: P=[P1', P2'].

Adopting uses the same method is normalized the second target sample, and concrete processing procedure is as follows:

Second target sample is splitted into two parts: Part I comprises wind speed, I1=I (:, 1:27); Part II comprises air pressure, the temperature and humidity of the second object height, I2=I (:, 28:54);

Respectively Part I I1 and Part II I2 is normalized calculating, the detailed process of calculating is the maximal value b of each component divided by all wind speed components of wind speed part, and specific formula for calculation (4) is expressed as:

$\left\{\begin{array}{c}b=max\left(max\right(I1\left)\right)\\ I{1}^{\&prime;}=\frac{1}{b}*\left(I1\right)\end{array}\right.---\left(4\right)$

Wherein, b is the maximal value of Part I I1, by each component of Part I divided by maximal value b, is the Part I of the second target sample after normalization calculates, represents with I1'.

The air pressure of the second object height, temperature, humidifying part adopt standard normalization algorithm, and specific formula for calculation (5) is:

[I2',ps]＝mapminmax(I2)(5)

Part I I1 and Part II I2 is merged the second target sample I after obtaining normalized, wherein, the second target sample after normalized is expressed as: I=[I1', I2'].

The following formula of concrete employing (6) is normalized each component in the first sample.

[A”,ps]＝mapminmax(A')(6)

E, set up neural network model

Following expression formula is adopted to set up neural network model:

net＝newff(P，A”，[4，12，4]，{'transig','transig','transig','purelin'}，'trainlm')；

First object sample after being normalized and the first sample are carried out model training to the neural network of tentatively putting up, shown in the following program of concrete training parameter:

Net.train.Param.epochs=1000; % initialization frequency of training.

Net.train.Param.goal=0.001; % sets training error.

Net.train.Param.lr=0.01; % sets learning rate.

Net.train.Param.mc=0.9; % sets factor of momentum.

By following program, training is started to the neural network of tentatively putting up:

net＝train(net，P，A”)；

It should be noted that, whole in first sample are effectively and the accurately data surveying wind data, so just by the Changing Pattern between neural network memory first object sample and the first sample, and then simulate target data more accurately, i.e. interpolation data.

Alternatively, step S108 calculates target data according to objective network model and comprises the steps that S1081 is to step S1087:

Step S1081, is updated to the second target sample in target simulator function, to make target simulator function carry out analog computation in objective network model, obtains first object result.

Step S1083, is undertaken returning normalization and calculates, obtain the second objective result by analog result.

Step S1085, adds very first time component by the second objective result, forms the 3rd objective result.

Step S1087, carries out merging by the first sample and the 3rd objective result and obtains target data.

Particularly, when neural network model has been trained, the second target sample is input in following program, has gone out to make neuron network simulation.

Use the second target sample to simulate according to following program, wherein, I is the second target sample, and O is that model exports, i.e. the normalization data of target data, and Pf, Af, E and perf are respectively the parameter in neural network:

[O,Pf,Af,E,perf]＝sim(net,I,[],[],[])；

The model obtained is exported O and carries out renormalization process according to following program:

O＝mapminmax('reverse',O,ts)；

Wherein, in said procedure, the model be after renormalization process on the equal sign left side exports, and then model is exported O and adds very first time component T1, obtain the O=[T1, O] with free component.

Interpolation data the most at last represented by this O=[T1, O] matrix is merged into be surveyed in wind data, obtains completing survey wind data in object time section: Temp=[A'; O]; And the file of the survey wind data after interpolation is preserved by following program: save (' windsample.dat', ' A "); So far, survey wind data interpolation to complete.

The method of data interpolation provided by the invention is also obtained and the mesoscale data surveying the wind data same time period by the survey wind data in a complete year of acquisition, and set up objective network model, by mesoscale data and effective wind data of surveying, it is trained, wherein, effective survey wind data is not limited only to adjacent time and survey wind data spatially, as long as the effective survey wind data in the survey wind data in a complete year; Surveying the time period of wind data disappearance, using Meso-Scale Analysis data as mode input, exporting as interpolation data using model, realizing the interpolation of surveying wind missing data.

According to above-mentioned training method of building routine objective network model and model, ensure to survey in the wind cycle in standard to complete survey wind, avoid repeating to survey wind; Avoid extending the wind energy turbine set construction period in early stage, save construction cost; Solve the indeterminable problem of method of carrying out data interpolation according to adjacent valid data correlativity.

Fig. 2 is the process flow diagram of the method according to the optional data interpolation of the another kind of the embodiment of the present invention, as shown in Figure 2, respectively by the second time component extraction in the very first time component in the first sample and the second sample out, obtain the second sample without time component and the first sample, wherein, without three sample of the second sample also namely in above-described embodiment of time component.Very first time component and the second time component are updated in the ismember function in Matlab and calculate, obtain the 3rd time component, 3rd time component and the second time divide and measure point multiplication operation, obtain the 4th time component, combine by the 4th time component with without the second sample of time component, obtaining the second sample with the 4th time component, is also the 4th sample in the above embodiment of the present invention.4th sample is split into first object sample and the second target sample, and the time component of first object sample and the second target sample is deleted, do not provide concrete delete procedure in fig. 2, the first object sample of erasing time component, the second target sample and the second sample without time component are normalized.Then the first object sample after normalized and the first sample are brought in objective network model and train, after training terminates, second target sample is brought in objective network model and carries out analog computation, obtaining Output rusults, is also the first object result in the above embodiment of the present invention.Carry out first object result returning the target data that normalized obtains without time component, also be the 3rd objective result in the above embodiment of the present invention, finally the 4th time component and the target data without time component are merged, obtain the target data of free component, i.e. target data as shown in Figure 2, the computation process of objectives data does not provide detailed process in fig. 2.The target data of free component and the first sample are carried out merging and is the first complete sample.

The embodiment of the present invention additionally provides a kind of device of data interpolation, the device of this data interpolation is mainly used in the method performing the data interpolation that embodiment of the present invention foregoing provides, and does concrete introduction below to the device of the data interpolation that the embodiment of the present invention provides.

Fig. 3 is the schematic diagram of the device of a kind of data interpolation according to the embodiment of the present invention, and as shown in Figure 3, the device of this data interpolation mainly comprises acquiring unit 10, builds unit 20 and computing unit 30, wherein:

Acquiring unit 10, for obtaining the first data in object time section and the second data, wherein, the quantity of the first data and the second data is multiple, and the first data are the data that anemometer tower is measured, and the second data are meteorologic analysis data.

It should be noted that, in embodiments of the present invention, the first data are the survey wind data of anemometer tower, and the second data are the mesoscale data of anemometer tower.Be chosen for according to national regulation object time section and be no less than year section.

Build unit 20, for building objective network model according to the first data and the second data.

It should be noted that, the objective network model in the embodiment of the present invention comprises the models such as neural network model, PSO Neural Network model, simulated annealing and genetic algorithm.Wherein, objective network model is built by the first data and the second data in simulation software, concrete simulation software is the simulation softwares such as Matlab, Matlab is as the software of general emulation, there is powerful data processing function, therefore in embodiments of the present invention, all for Matlab simulation software, building of objective network model is introduced.

Computing unit 30, for calculating target data according to objective network model, wherein, target data is the data needing interpolation in object time section.

It should be noted that, the survey wind data in object time section causes data imperfect due to a variety of causes such as weather, causes permanent loss, and the data of this disappearance are the data surveyed and need interpolation in wind data.

In embodiments of the present invention, by obtaining survey wind data in the same time period and mesoscale data, and build objective network model according to survey wind data and mesoscale data, and by Meso-Scale Analysis data with effectively survey wind data to the training of objective network model, obtain the target data in the time period of surveying wind data disappearance, reach the object of the interpolation of surveying wind missing data, thus achieve the technique effect of the survey wind data of simulating disappearance according to the correlativity of adjacent valid data, and then solve because long missing data causes the correlativity that can not utilize adjacent valid data in the time and space in prior art to obtain lacking the technical matters surveying wind data.

Alternatively, data difference provided by the invention must not also comprise by device: converting unit, for after the first data obtained in object time section and the second data, first data are pre-conditionedly converted to the first sample according to first, and the second data are pre-conditionedly converted to the second sample according to second, build unit to comprise: first builds module, for building objective network model according to the first sample and the second sample.

After the data acquisition unit in anemometer tower gets the first data and the second data, the first data and the second data are all stored in storer with the form of fixed text.When the first data and the second data build objective network model in Matlab, need be the form of Matlab simulation software acquiescence by the first data and the second data transformations, the such as text of " .dat " form.Therefore when selecting Matlab simulation software, need by converting unit by the first data according to first pre-conditioned and the second data according to the second pre-conditioned text being converted into " .dat " form.

And then in Matlab simulation software, build objective network model according to the first sample after conversion and the second sample.It should be noted that, if the simulation software selected is the software beyond Matlab simulation software, then needing the first data and the second data transformations is the text formatting corresponding with this simulation software.

Alternatively, multiple first data are included in the data that the multiple first object moment in object time section collect, and converting unit comprises: the first read module, for reading first object moment t in multiple first object moment respectively _hthe first parameter and the second parameter, wherein, the first Parametric Representation is the wind speed of anemometer tower at multiple first object height, and the second Parametric Representation is the wind direction of anemometer tower at multiple first object height, h gets the positive integer of 1 to m1 successively, and m1 is the quantity in first object moment; First conversion module, according to formula first parameter and the second parameter are converted into primary vector u _fwith secondary vector v _f, wherein, S _fbe the first parameter, be the second parameter, f gets the positive integer of 1 to m2 successively, and m2 is the quantity of first object height; Second conversion module, for by first object moment t _hbe converted into the moment of object format; First composition module, for being converted into the first object moment t after object format _hand first object moment t _hthe first parameter and the 3rd parameter form First ray, wherein, the 3rd parameter comprises following at least one: anemometer tower is in the temperature of the second object height, air pressure and relative humidity, and the quantity of First ray is m2, and multiple First ray forms the first sample.

Read the first data by the first read module, the survey wind data of also i.e. anemometer tower measurement, and according to multiple first object moment, the first data are divided into groups.Wherein, each group data comprises the wind speed and direction apart from ground differing heights, wind speed S _f, be, the first parameter; Wind direction be, the second parameter.By the first conversion module by wind speed S _fbe converted into the vector of both direction, i.e. primary vector u _fwith secondary vector v _f, wherein, primary vector u _ffor in first object height t _hthe wind velocity vector of upper east-west direction, secondary vector v _ffor in first object height t _hthe wind velocity vector of upper North and South direction.And by the second conversion unit by first object moment t _hbe converted into object format, objectives form is expressed as " YYYYMMDDHHMI ", wherein, " YYYY " 4 bit digital represent year, " MM " 2 bit digital represent the moon, " DD " 2 bit digital represent day, " HH " 2 bit digital to represent hour, " MI " 2 is numeral minute.

Multiple second data are included in the data that multiple second object time in object time section collect, and the quantity of the second object time is greater than the quantity in first object moment, and converting unit comprises:

Second read module, for reading the second data at the second object time t _kthe 4th parameter, wherein, k gets the positive integer of 1 to m3 successively, and m3 is the quantity in the second order moment, and the quantity of the 4th parameter is multiple.

It should be noted that, the second data are read by the second read module, second data are also mesoscale data, the horizontal resolution of mesoscale data can reach 1.5 kilometers (namely lattice point distance is not more than 1.5 kilometers), the vertical direction number of plies is not less than 10, wherein be no less than 3 layers apart from floor level less than 150 meters, maximum height is not less than 1000 meters, and time granularity is 10 minutes.

First computing module, for calculating the 5th parameter Aij of the second sample according to the 4th parameter, wherein, 5th parameter Aij at least comprises following at least one: wind speed, megadyne temperature, air pressure and vapor-to-liquid ratio, i gets the positive integer of 1 to a successively, and a is the quantity of horizontal level, and j gets the positive integer of 1 to b successively, b is the quantity of differing heights on each horizontal level, the parameter of the jth of the 5th Parametric Representation on an i-th horizontal level height.

First computing module calculates the 5th parameter Aij in the second sample according to formula (1), wherein, the 5th parameter Aij comprise according to object format transform after the second object time t _k, the u of jth the height of 3 components on i-th horizontal level of wind speed _ij, v _ij, w _ij, megadyne temperature θ _ij, vapor-to-liquid ratio q _ijand air pressure p _ij.

Calculate the 5th parameter according to following formula (1), wherein, the equation left side is the 4th parameter, is the 5th parameter on the right of equation.

$\left\{\begin{array}{c}{u}_{ij}=U_{ij},v_{ij}=V_{ij},w_{ij}=W_{ij}\\ {\mathrm{\&theta;}}_{ij}=T_{ij}+TH{2}_{ij}\\ p_{ij}=P_{ij}+{\mathrm{PB}}_{ij}\\ q_{ij}={\mathrm{QVAPOR}}_{ij}\end{array}\right.---\left(1\right)$

Wherein, U _ij, V _ijand W _ijbe respectively the wind speed component of the jth height on i-th horizontal level.T _ijbe expressed as the disturbance megadyne temperature of the jth height on i-th horizontal level, TH2 _ijbe expressed as 2 meters, the ground height megadyne temperature of the jth height on i-th horizontal level, P _ijbe expressed as the disturbance of the jth height on i-th horizontal level, PB _ijbe expressed as the ground state air pressure of the jth height on i-th horizontal level, QVAPOR _ijbe expressed as the vapor-to-liquid ratio of the jth height on i-th horizontal level.

Second composition module, for by the second object time t _kthe 5th parameter Aij and the second object time t _kform the second sequence, wherein, multiple second Sequence composition second sample.

The device of data interpolation provided by the invention also comprises: split cells, for the first data being pre-conditionedly converted to the first sample according to first in converting unit, and by the second data according to second pre-conditioned be converted to the second sample after, second sample is split into first object sample and the second target sample, build unit to comprise: second builds module, for building objective network model according to first object sample and the first sample.

Split cells comprises: extraction module, and for extracting very first time component and the second time component, wherein, very first time component is the first row of the first sample, and the second time component is the secondary series of the second sample; Judge module, for judging the second object time t in the second time component _kwhether comprise the first object moment t in very first time component _h, obtaining judged result, wherein, is the second object time t in judged result _kcomprise first object moment t _hwhen, the second object time t _kbeing set to 1, is the second object time t in judged result _kdo not comprise first object moment t _hwhen, the second object time t _kbe set to 0; Update module, for upgrading the second time component according to judged result, obtains the 3rd time component; Second computing module, for the second time component and the 3rd time component are carried out dot product calculating, obtains the 4th time component; First adds module, for the 4th time component is added in the 3rd sample, obtain the 4th sample, wherein, identical with the first sample in 4th sample is first object sample, not identical with the first sample in 4th sample is the second target sample, and the 3rd sample is the sample that the second sample deletes after the second time component.

By following program by the first sample A' and the second sample B ' be carried in Matlab program: step one: load document is called the first sample A':load (' mesosample.dat', ' A ") of windsample.dat; Step 2: load document is called the second sample B ': load (' the windsample.dat' of windsample.dat, ' B ").

The very first time component in the first sample is extracted: T1=A'(by extraction module:, 1); And the second time component: the T2=B'(extracted by extraction module in the second sample:, 1).

Be 0 by following program by needing the time mark of interpolation data in the second time component:

T3＝ismember(T2，T1，'rows')；

Wherein, the quantity of the second time component is greater than the quantity of very first time component, and the second time component and very first time component have identical time component.The 3rd time component finally calculated be only comprise 1 and/or 0 time component.

Need the time of interpolation all to become 0, T2=T2.*T3 by following program by the second time component, wherein, the T2 on the equation left side is the 4th time component.

By following program, is replaced with the time of mark interpolation period the time in the second sample,

C=B'; Wherein, the Matrix C on the equation left side is the 3rd sample.

C(：，1)＝[]；

C=[T2, C]; Wherein, the Matrix C on the equation left side is the 4th sample.

Present Matrix C is, on the basis of the second sample, the interpolation time period is labeled as 0, and matrix B ' the part identical with C be exactly the second target sample, matrix B ' the part different from C is exactly first object sample.

Wherein, second builds module comprises: delete submodule, for deleting the 5th time component in first object sample and the very first time component in the first sample, wherein, the 5th time component is the first row of first object sample; First calculating sub module, for being normalized calculating, the first object sample after obtaining normalized and the first sample by the first object sample deleting the 5th time component with the first sample deleting very first time component respectively; Load submodule, for being loaded in target program by the first object sample after being normalized and the first sample, to make first object sample to the training of objective network model, wherein, after training completes, objective network model has carried.

The description of following step is each option setting of neural network based on Matlab.

Step one: build neural network structure; Step 2: transport function is set; Step 3: training function is set; Step 4: use Matlab programming language to build neural network model.

Computing unit comprises: substituting into module, for being updated in target simulator function by the second target sample, to make target simulator function carry out analog computation in objective network model, obtaining first object result; 3rd computing module, calculating for analog result being carried out return normalization, obtaining the second objective result; Second adds module, for the second objective result is added very first time component, forms the 3rd objective result; Merge module, obtain target data for the first sample and the 3rd objective result are carried out merging.

Particularly, when neural network model has been trained, the second target sample is input in following program, has gone out to make neuron network simulation.

Use the second target sample to simulate according to following program, wherein, I is the second target sample, and O is that model exports, i.e. the normalization data of target data, and Pf, Af, E and perf are respectively the parameter in neural network:

[O,Pf,Af,E,perf]＝sim(net,I,[],[],[])；

The model obtained is exported O and carries out renormalization process according to following program:

O＝mapminmax('reverse',O,ts)；

Wherein, in said procedure, the model be after renormalization process on the equal sign left side exports, and then model is exported O and adds very first time component T1, obtain the O=[T1, O] with free component.

Interpolation data the most at last represented by this O=[T1, O] matrix is merged into be surveyed in wind data, obtains completing survey wind data in object time section: Temp=[A'; O]; And the file of the survey wind data after interpolation is preserved by following program: save (' windsample.dat', ' A "); So far, survey wind data interpolation to complete.

The invention described above embodiment sequence number, just to describing, does not represent the quality of embodiment.

In the above embodiment of the present invention, the description of each embodiment is all emphasized particularly on different fields, in certain embodiment, there is no the part described in detail, can see the associated description of other embodiments.

In several embodiments that the application provides, should be understood that, disclosed technology contents, the mode by other realizes.Wherein, device embodiment described above is only schematic, the such as division of described unit, can be that a kind of logic function divides, actual can have other dividing mode when realizing, such as multiple unit or assembly can in conjunction with or another system can be integrated into, or some features can be ignored, or do not perform.Another point, shown or discussed coupling each other or direct-coupling or communication connection can be by some interfaces, and the indirect coupling of unit or module or communication connection can be electrical or other form.

The described unit illustrated as separating component or can may not be and physically separates, and the parts as unit display can be or may not be physical location, namely can be positioned at a place, or also can be distributed on multiple unit.Some or all of unit wherein can be selected according to the actual needs to realize the object of the present embodiment scheme.

In addition, each functional unit in each embodiment of the present invention can be integrated in a processing unit, also can be that the independent physics of unit exists, also can two or more unit in a unit integrated.Above-mentioned integrated unit both can adopt the form of hardware to realize, and the form of SFU software functional unit also can be adopted to realize.

If described integrated unit using the form of SFU software functional unit realize and as independently production marketing or use time, can be stored in a computer read/write memory medium.Based on such understanding, the part that technical scheme of the present invention contributes to prior art in essence in other words or all or part of of this technical scheme can embody with the form of software product, this computer software product is stored in a storage medium, comprises all or part of step of some instructions in order to make a computer equipment (can be personal computer, server or the network equipment etc.) perform method described in each embodiment of the present invention.And aforesaid storage medium comprises: USB flash disk, ROM (read-only memory) (ROM, Read-OnlyMemory), random access memory (RAM, RandomAccessMemory), portable hard drive, magnetic disc or CD etc. various can be program code stored medium.

The above is only the preferred embodiment of the present invention; it should be pointed out that for those skilled in the art, under the premise without departing from the principles of the invention; can also make some improvements and modifications, these improvements and modifications also should be considered as protection scope of the present invention.

Claims (16)

1. a method for data interpolation, is characterized in that, comprising:

Obtain the first data in object time section and the second data, wherein, the quantity of described first data and described second data is multiple, and described first data are the data that anemometer tower is measured, and described second data are meteorologic analysis data;

Objective network model is built according to described first data and described second data; And

Calculate target data according to described objective network model, wherein, described target data is the data needing interpolation in described object time section.

2. method according to claim 1, is characterized in that, after the first data obtained in object time section and the second data, described method also comprises:

Described first data are pre-conditionedly converted to the first sample according to first, and described second data are pre-conditionedly converted to the second sample according to second,

Build objective network model according to described first data and described second data to comprise: build objective network model according to described first sample and described second sample.

3. method according to claim 2, is characterized in that, multiple described first data are included in the data that the multiple first object moment in described object time section collect, and described first data are comprised according to the first pre-conditioned first sample that is converted to:

Read first object moment t in multiple described first object moment respectively _hthe first parameter and the second parameter, wherein, described first Parametric Representation is the wind speed of described anemometer tower at multiple first object height, described second Parametric Representation is the wind direction of described anemometer tower at described multiple first object height, h gets the positive integer of 1 to m1 successively, and m1 is the quantity in described first object moment;

According to formula described first parameter and described second parameter are converted into primary vector u _fwith secondary vector v _f, wherein, S _ffor described first parameter, for described second parameter, f gets the positive integer of 1 to m2 successively, and m2 is the quantity of described first object height;

By described first object moment t _hbe converted into the moment of object format; And

Described first object moment t after described object format will be converted into _hand described first object moment t _hdescribed first parameter and the 3rd parameter form First ray, wherein, described 3rd parameter comprises following at least one: described anemometer tower is in the temperature of the second object height, air pressure and relative humidity, and the quantity of described First ray is m2, and multiple described First ray forms described first sample.

4. method according to claim 3, it is characterized in that, multiple described second data are included in the data that multiple second object time in described object time section collect, the quantity of described second object time is greater than the quantity in described first object moment, described second data is comprised according to the second pre-conditioned second sample that is converted to:

Read described second data at the second object time t _kthe 4th parameter, wherein, k gets the positive integer of 1 to m3 successively, and m3 is the quantity in described second order moment, and the quantity of described 4th parameter is multiple;

The 5th parameter Aij of described second sample is calculated according to described 4th parameter, wherein, described 5th parameter Aij at least comprises following at least one: wind speed, megadyne temperature, air pressure and vapor-to-liquid ratio, i gets the positive integer of 1 to a successively, a is the quantity of horizontal level, j gets the positive integer of 1 to b successively, and b is the quantity of differing heights on each described horizontal level, the parameter of the jth of described 5th Parametric Representation on an i-th horizontal level height; And

By described second object time t _kdescribed 5th parameter Aij and described second object time t _kform the second sequence, wherein, the second sample described in multiple described second Sequence composition.

5. method according to claim 4, is characterized in that, described first data are pre-conditionedly being converted to the first sample according to first, and by described second data according to second pre-conditioned be converted to the second sample after, described method comprises:

Described second sample is split into first object sample and the second target sample,

Build objective network model according to described first data and described second data to comprise: build described objective network model according to described first object sample and described first sample.

6. method according to claim 5, is characterized in that, described second sample is split into first object sample and the second target sample comprises:

Extract very first time component and the second time component, wherein, described very first time component is the first row of described first sample, and described second time component is the secondary series of described second sample;

Judge the described second object time t in described second time component _kwhether comprise the described first object moment t in described very first time component _h, obtaining judged result, wherein, is described second object time t in described judged result _kcomprise described first object moment t _hwhen, described second object time t _kbeing set to 1, is described second object time t in described judged result _kdo not comprise described first object moment t _hwhen, described second object time t _kbe set to 0;

Upgrade described second time component according to described judged result, obtain the 3rd time component;

Described second time component and described 3rd time component are carried out point multiplication operation, obtains the 4th time component; And

Described 4th time component is added in the 3rd sample, obtain the 4th sample, wherein, identical with described first sample in described 4th sample is described first object sample, not identical with described first sample in described 4th sample is described second target sample, and described 3rd sample is the sample that described second sample deletes after described second time component.

7. method according to claim 6, is characterized in that, builds objective network model comprise according to described first object sample and described second sample:

Delete the 5th time component in described first object sample and the described very first time component in described first sample, wherein, described 5th time component is the first row of described first object sample;

Respectively the described first object sample deleting described 5th time component is normalized calculating, the described first object sample after obtaining normalized and described first sample with described first sample deleting described very first time component; And

Described first object sample after being normalized and described first sample are loaded in target program, to make described first object sample to the training of described objective network model, wherein, after training completes, described objective network model has carried.

8. method according to claim 5, is characterized in that, calculates target data comprise according to described objective network model:

Described second target sample is updated in target simulator function, to make described target simulator function carry out analog computation in described objective network model, obtains first object result;

Described first object result is carried out return normalization to calculate, obtain the second objective result;

Described second objective result is added very first time component, forms the 3rd objective result; And

Described first sample and described 3rd objective result are carried out merging and obtains described target data.

9. a device for data interpolation, is characterized in that, comprising:

Acquiring unit, for obtaining the first data in object time section and the second data, wherein, the quantity of described first data and described second data is multiple, and described first data are the data that anemometer tower is measured, and described second data are meteorologic analysis data;

Build unit, for building objective network model according to described first data and described second data; And

Computing unit, for calculating target data according to described objective network model, wherein, described target data is the data needing interpolation in described object time section.

10. device according to claim 9, is characterized in that, described device also comprises:

Described first data, for after obtaining the first data in object time section and the second data, are pre-conditionedly converted to the first sample according to first by converting unit, and described second data are pre-conditionedly converted to the second sample according to second,

Described unit of building comprises: first builds module, for building objective network model according to described first sample and described second sample.

11. devices according to claim 10, is characterized in that, multiple described first data are included in the data that the multiple first object moment in described object time section collect, and described converting unit comprises:

First read module, for reading first object moment t in multiple described first object moment respectively _hthe first parameter and the second parameter, wherein, described first Parametric Representation is the wind speed of described anemometer tower at multiple first object height, described second Parametric Representation is the wind direction of described anemometer tower at described multiple first object height, h gets the positive integer of 1 to m1 successively, and m1 is the quantity in described first object moment;

First conversion module, for according to formula described first parameter and described second parameter are converted into primary vector u _fwith secondary vector v _f, wherein, S _ffor described first parameter, for described second parameter, f gets the positive integer of 1 to m2 successively, and m2 is the quantity of described first object height;

Second conversion module, for by described first object moment t _hbe converted into the moment of object format; And

First composition module, for being converted into the described first object moment t after described object format _hand described first object moment t _hdescribed first parameter and the 3rd parameter form First ray, wherein, described 3rd parameter comprises following at least one: described anemometer tower is in the temperature of the second object height, air pressure and relative humidity, and the quantity of described First ray is m2, and multiple described First ray forms described first sample.

12. devices according to claim 11, it is characterized in that, multiple described second data are included in the data that multiple second object time in described object time section collect, and the quantity of described second object time is greater than the quantity in described first object moment, and described converting unit comprises:

Second read module, for reading described second data at the second object time t _kthe 4th parameter, wherein, k gets the positive integer of 1 to m3 successively, and m3 is the quantity in described second order moment, and the quantity of described 4th parameter is multiple;

First computing module, for calculating the 5th parameter Aij of described second sample according to described 4th parameter, wherein, described 5th parameter Aij at least comprises following at least one: wind speed, megadyne temperature, air pressure and vapor-to-liquid ratio, i gets the positive integer of 1 to a successively, and a is the quantity of horizontal level, and j gets the positive integer of 1 to b successively, b is the quantity of differing heights on each described horizontal level, the parameter of the jth of described 5th Parametric Representation on an i-th horizontal level height; And

Second composition module, for by described second object time t _kdescribed 5th parameter Aij and described second object time t _kform the second sequence, wherein, the second sample described in multiple described second Sequence composition.

13. devices according to claim 12, is characterized in that, described device comprises:

Split cells, for described first data being pre-conditionedly converted to the first sample according to first in described converting unit, and by described second data according to second pre-conditioned be converted to the second sample after, described second sample is split into first object sample and the second target sample

Described unit of building comprises: second builds module, for building described objective network model according to described first object sample and described first sample.

14. devices according to claim 13, is characterized in that, described split cells comprises:

Extraction module, for extracting very first time component and the second time component, wherein, described very first time component is the first row of described first sample, and described second time component is the secondary series of described second sample;

Judge module, for judging the described second object time t in described second time component _kwhether comprise the described first object moment t in described very first time component _h, obtaining judged result, wherein, is described second object time t in described judged result _kcomprise described first object moment t _hwhen, described second object time t _kbeing set to 1, is described second object time t in described judged result _kdo not comprise described first object moment t _hwhen, described second object time t _kbe set to 0;

Update module, for upgrading described second time component according to described judged result, obtains the 3rd time component;

Second computing module, for described second time component and described 3rd time component are carried out dot product calculating, obtains the 4th time component; And

First adds module, for described 4th time component is added in the 3rd sample, obtain the 4th sample, wherein, identical with described first sample in described 4th sample is described first object sample, not identical with described first sample in described 4th sample is described second target sample, and described 3rd sample is the sample that described second sample deletes after described second time component.

15. devices according to claim 14, is characterized in that, described second builds module comprises:

Delete submodule, for deleting the 5th time component in described first object sample and the described very first time component in described first sample, wherein, described 5th time component is the first row of described first object sample;

First calculating sub module, for respectively the described first object sample deleting described 5th time component being normalized calculating, the described first object sample after obtaining normalized and described first sample with described first sample deleting described very first time component; And

Load submodule, for the described first object sample after being normalized and described first sample are loaded in target program, to make described first object sample to the training of described objective network model, wherein, after training completes, described objective network model has carried.

16. devices according to claim 13, is characterized in that, described computing unit comprises:

Substituting into module, for being updated in target simulator function by described second target sample, to make described target simulator function carry out analog computation in described objective network model, obtaining first object result;

3rd computing module, calculating for described first object result being carried out return normalization, obtaining the second objective result;

Second adds module, for described second objective result is added very first time component, forms the 3rd objective result; And

Merge module, obtain described target data for described first sample and described 3rd objective result are carried out merging.