CN116231763B

CN116231763B - Household energy management system optimization scheduling method and device with self-learning capability

Info

Publication number: CN116231763B
Application number: CN202310504391.XA
Authority: CN
Inventors: 罗剑洪; 卢雪明; 欧阳家淦; 张家杰; 陈建松
Original assignee: Guangzhou Sanjing Electric Co Ltd
Current assignee: Guangzhou Sanjing Electric Co Ltd
Priority date: 2023-05-08
Filing date: 2023-05-08
Publication date: 2023-08-04
Anticipated expiration: 2043-05-08
Also published as: CN116231763A

Abstract

The application relates to a self-learning family energy management system optimization scheduling method and device, which are used for determining constraint conditions of an energy storage system model, determining an objective function and constraint conditions of the family energy management system optimization scheduling model according to first photovoltaic output prediction data, first load prediction data and constraint conditions of the energy storage system model, and determining a preliminary scheduling strategy. And establishing an error probability distribution model according to the errors of the second photovoltaic output prediction data and the first photovoltaic output data and the errors of the second load prediction data and the first load data, updating a household energy management system optimization scheduling model according to the error probability distribution model, obtaining an updated scheduling strategy, and configuring the updated scheduling strategy into a scheduling household energy management system. And the dispatching strategy is updated through the actual error, so that the dispatching accuracy of the household energy management system is improved, and the dispatching effect is optimized.

Description

Household energy management system optimization scheduling method and device with self-learning capability

Technical Field

The application relates to the technical field of household energy management, in particular to a household energy management system optimization scheduling method and device with self-learning capability.

Background

At present, energy and environmental problems have become important factors restricting social development, which puts higher demands on the development of electric power systems. In order to guide users to better use electricity, a household energy management system based on distributed power generation and energy storage is generated.

The household energy management system (Home Energy Management System, HEMS) is an extension of a smart grid, and depends on a smart meter to interact with the power grid to obtain information such as real-time electricity price, photovoltaic output and the like. Meanwhile, the HEMS integrates the distributed power generation equipment in the home, and the intelligent home is efficiently managed and controlled, so that the HEMS can intelligently replace a user to respond to electricity price, and make an equipment coordination optimization decision, thereby being the embodiment of intelligent electricity utilization and distributed power generation at a user side. The household energy management system can save energy on the electricity utilization side and has important practical application value.

At present, the optimal scheduling of the home energy management system is usually solved by adopting an intelligent algorithm, and mainly comprises a genetic algorithm, a particle swarm algorithm, an ant colony algorithm and the like. However, because the home energy management system often has a large difference between the pre-executed schedule and the real situation in the actual running process, the accuracy of the schedule is affected.

Disclosure of Invention

Based on the above, it is necessary to provide a method and a device for optimizing and dispatching a home energy management system with self-learning capability, aiming at the defect that the accuracy of dispatching is affected due to the fact that the pre-executed dispatching and the real situation often have larger difference.

At least one embodiment of the present disclosure provides a home energy management system optimization scheduling method with self-learning capability, including the steps of:

acquiring first photovoltaic output prediction data and first load prediction data;

determining constraint conditions of an energy storage system model;

determining an objective function and constraint conditions of a family energy management system optimization scheduling model according to the first photovoltaic output prediction data, the first load prediction data and the constraint conditions of the energy storage system model, so as to determine a preliminary scheduling strategy;

the household energy management system is instructed to operate according to the preliminary scheduling strategy, and second photovoltaic output prediction data, second load prediction data, first photovoltaic output data and first load data in an operation period are obtained;

establishing an error probability distribution model according to errors of the second photovoltaic output prediction data and the first photovoltaic output data and errors of the second load prediction data and the first load data;

Updating the optimal scheduling model of the household energy management system according to the error probability distribution model to obtain an updated scheduling strategy; wherein the updated scheduling policy is configured to schedule the home energy management system.

According to the household energy management system optimization scheduling method with the self-learning capability, the constraint condition of the energy storage system model is determined, and the objective function and the constraint condition of the household energy management system optimization scheduling model are determined according to the first photovoltaic output prediction data, the first load prediction data and the constraint condition of the energy storage system model, so that the preliminary scheduling strategy is determined. Further, an error probability distribution model is established according to errors of the second photovoltaic output prediction data and the first photovoltaic output data and errors of the second load prediction data and the first load data, and a household energy management system optimization scheduling model is updated according to the error probability distribution model to obtain an updated scheduling strategy which is configured to schedule the household energy management system. Based on the method, the scheduling strategy is updated through the actual error, the scheduling accuracy of the home energy management system is improved, and the scheduling effect is optimized.

The present disclosure also provides a home energy management system optimization scheduling apparatus with self-learning capability, including:

the data acquisition module is used for acquiring first photovoltaic output prediction data and first load prediction data;

the condition determining module is used for determining constraint conditions of the energy storage system model according to the first photovoltaic output prediction data and the first load prediction data;

the strategy determining module is used for determining an objective function and constraint conditions of the optimal scheduling model of the household energy management system according to the first photovoltaic output prediction data, the first load prediction data and the constraint conditions of the energy storage system model, so as to determine a preliminary scheduling strategy;

the data operation module is used for indicating the operation of the home energy management system according to the preliminary scheduling strategy and acquiring second photovoltaic output prediction data, second load prediction data, first photovoltaic output data and first load data in an operation period;

the error analysis module is used for establishing an error probability distribution model according to the errors of the second photovoltaic output prediction data and the first photovoltaic output data and the errors of the second load prediction data and the first load data;

the scheduling updating module is used for updating the optimal scheduling model of the household energy management system according to the error probability distribution model to obtain an updated scheduling strategy; wherein the updated scheduling policy is configured to schedule the home energy management system.

The household energy management system optimization scheduling device with the self-learning capability determines the constraint condition of the energy storage system model, and determines the objective function and the constraint condition of the household energy management system optimization scheduling model according to the first photovoltaic output prediction data, the first load prediction data and the constraint condition of the energy storage system model, so as to determine the preliminary scheduling strategy. Further, an error probability distribution model is established according to errors of the second photovoltaic output prediction data and the first photovoltaic output data and errors of the second load prediction data and the first load data, and a household energy management system optimization scheduling model is updated according to the error probability distribution model to obtain an updated scheduling strategy which is configured to schedule the household energy management system. Based on the method, the scheduling strategy is updated through the actual error, the scheduling accuracy of the home energy management system is improved, and the scheduling effect is optimized.

At least one embodiment of the present disclosure further provides a data scheduling apparatus, including:

one or more memories non-transitory storing computer-executable instructions;

the system comprises one or more processors configured to execute computer-executable instructions, wherein the computer-executable instructions, when executed by the one or more processors, implement a self-learning-enabled home energy management system optimization scheduling method according to any embodiment of the present disclosure.

The load prediction device determines the constraint condition of the energy storage system model, and determines the objective function and the constraint condition of the optimal scheduling model of the household energy management system according to the first photovoltaic output prediction data, the first load prediction data and the constraint condition of the energy storage system model, so as to determine the preliminary scheduling strategy. Further, an error probability distribution model is established according to errors of the second photovoltaic output prediction data and the first photovoltaic output data and errors of the second load prediction data and the first load data, and a household energy management system optimization scheduling model is updated according to the error probability distribution model to obtain an updated scheduling strategy which is configured to schedule the household energy management system. Based on the method, the scheduling strategy is updated through the actual error, the scheduling accuracy of the home energy management system is improved, and the scheduling effect is optimized.

At least one embodiment of the present disclosure also provides another non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer-executable instructions that, when executed by a processor, implement a self-learning-capable home energy management system optimization scheduling method according to any embodiment of the present disclosure.

The non-transitory computer readable storage medium determines constraint conditions of the energy storage system model, and determines objective functions and constraint conditions of the optimal scheduling model of the home energy management system according to the first photovoltaic output prediction data, the first load prediction data and the constraint conditions of the energy storage system model, so as to determine a preliminary scheduling strategy. Further, an error probability distribution model is established according to errors of the second photovoltaic output prediction data and the first photovoltaic output data and errors of the second load prediction data and the first load data, and a household energy management system optimization scheduling model is updated according to the error probability distribution model to obtain an updated scheduling strategy which is configured to schedule the household energy management system. Based on the method, the scheduling strategy is updated through the actual error, the scheduling accuracy of the home energy management system is improved, and the scheduling effect is optimized.

Drawings

Fig. 1 is a schematic view of a home energy system scenario according to an embodiment of the present disclosure;

FIG. 2 is a flowchart of an optimized dispatching method of a home energy management system with self-learning capability according to an embodiment;

FIG. 3 is a flowchart of another embodiment of a method for optimizing and scheduling a home energy management system with self-learning capability;

FIG. 4 is a 12-point period photovoltaic power prediction error probability model of a specific application example;

FIG. 5 is a 12-point period load power prediction error probability model of a specific application example;

FIG. 6 is a graph of real-time electricity prices for a specific application example;

FIG. 7 is a diagram showing the power consumption cost change process after self-learning scheduling according to a specific application example;

FIG. 8 is an energy balance diagram after self-learning scheduling according to a specific application example;

FIG. 9 is a diagram illustrating an operating state of the energy storage system after self-learning scheduling according to an embodiment of the present invention;

FIG. 10 is a graph showing the power consumption cost after conventional scheduling according to a specific application example;

FIG. 11 is a graph of energy balance after conventional scheduling for a specific application example;

FIG. 12 is a block diagram of a home energy management system optimization scheduler module with self-learning capabilities according to an embodiment;

FIG. 13 is a schematic block diagram of a data scheduling apparatus provided in accordance with at least one embodiment of the present disclosure;

fig. 14 is a schematic diagram of a non-transitory computer-readable storage medium provided by at least one embodiment of the present disclosure.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present disclosure more apparent, the technical solutions of the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present disclosure. It will be apparent that the described embodiments are some, but not all, of the embodiments of the present disclosure. All other embodiments, which can be made by one of ordinary skill in the art without the need for inventive faculty, are within the scope of the present disclosure, based on the described embodiments of the present disclosure.

Unless defined otherwise, technical or scientific terms used in this disclosure should be given the ordinary meaning as understood by one of ordinary skill in the art to which this disclosure belongs. The terms "first," "second," and the like, as used in this disclosure, do not denote any order, quantity, or importance, but rather are used to distinguish one element from another. The word "comprising" or "comprises", and the like, means that elements or items preceding the word are included in the element or item listed after the word and equivalents thereof, but does not exclude other elements or items. The terms "connected" or "connected," and the like, are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "upper", "lower", "left", "right", etc. are used merely to indicate relative positional relationships, which may also be changed when the absolute position of the object to be described is changed.

In order to keep the following description of the embodiments of the present disclosure clear and concise, the present disclosure omits a detailed description of some known functions and known components.

At least one embodiment of the present disclosure provides a home energy management system optimization scheduling method with self-learning capability.

In order to better understand the application scenario of the embodiments of the present disclosure, the application scenario of the embodiments of the present disclosure is first explained below. Fig. 1 is a schematic view of a home energy system according to an embodiment of the present disclosure, and as shown in fig. 1, the home energy system includes a distributed photovoltaic power generation system 1, an energy storage system 2, a power grid 3, and a system load 4. The distributed photovoltaic power generation system 1 performs distributed photovoltaic power generation to supply power to the energy storage system 2 and to supply power to the system load 4, or directly feeds into the power grid 3. The energy storage system 2 can be switched between charging and discharging, charged by the grid 3 or distributed photovoltaic power generation, or discharged to power the system load 4 or feed into the grid 3. According to the purchase of electricity to the power grid 3, the power supply to the energy storage system 2 or the power supply to the system load 4 is realized. Therefore, the whole system determines the working modes of the energy storage system 2 and the power grid 3 according to the photovoltaic power generation and the load balance of the system load, and records historical data.

As shown in fig. 1, the home energy management system 5 is configured as a management system, and includes controlling charge-discharge switching, power, electricity purchasing, electricity selling, and the like of the energy storage system 2 to balance loss and benefit of the home energy system. For example, the lower the number of charge-discharge switches of the energy storage system 2, the lower the system loss. The electricity purchasing and selling strategies are determined according to the real-time electricity price of the power grid 3 so as to improve the income.

It should be noted that, as the management system of the distributed photovoltaic power generation system 1, the energy storage system 2, the power grid 3 and the system load 4, the home energy management system 5 may obtain various operation indexes, such as power, electric quantity, charge and discharge states, of the distributed photovoltaic power generation system 1, the energy storage system 2, the power grid 3 and the system load 4 in the actual operation process, which are used as the execution data base of the home energy management system with self-learning capability in any embodiment, and will not be described in detail later.

Based on the above application scenario, fig. 2 is a flowchart of an embodiment of a self-learning home energy management system optimization scheduling method, as shown in fig. 2, and an embodiment of a self-learning home energy management system optimization scheduling method includes steps S100 to S105:

s100, acquiring first photovoltaic output prediction data and first load prediction data;

s101, determining constraint conditions of an energy storage system model;

s102, determining an objective function and constraint conditions of an optimal scheduling model of a household energy management system according to first photovoltaic output prediction data, first load prediction data and constraint conditions of an energy storage system model, so as to determine a preliminary scheduling strategy;

S103, indicating the operation of the home energy management system according to the preliminary scheduling strategy, and acquiring second photovoltaic output prediction data, second load prediction data, first photovoltaic output data and first load data in an operation period;

s104, establishing an error probability distribution model according to errors of the second photovoltaic output prediction data and the first photovoltaic output data and errors of the second load prediction data and the first load data;

s105, updating a family energy management system optimization scheduling model according to the error probability distribution model to obtain an updated scheduling strategy; wherein the updated scheduling policy is configured to schedule the home energy management system.

In the embodiment of the disclosure, the photovoltaic output prediction data includes first photovoltaic output prediction data, second photovoltaic output prediction data and third photovoltaic output prediction data, which are all the results of the photovoltaic output prediction model for pre-prediction, represent the working prediction state of the distributed photovoltaic power generation, and are mainly characterized as the output power predicted by the photovoltaic power generation. The corresponding first photovoltaic output data and second photovoltaic output data represent the working actual state of the distributed photovoltaic power generation, and represent the actual output power of the photovoltaic power generation.

In the embodiment of the disclosure, the load prediction data includes first load prediction data, second load prediction data and third load prediction data, which are all the results of the load prediction model for performing the prediction in advance, represent the load state of the system load, and are mainly represented as the load power predicted by the system load or the energy storage system. The corresponding first load data and second load data represent the real load of the system, and represent the actual load power of the system load or the energy storage system.

In the embodiments of the present disclosure, in order to better explain the execution of the features, the explanation of the embodiments is performed with a fixed scheduling period and a unit time length, while unifying the data output of the photovoltaic output prediction model and the load prediction model.

In one embodiment, the photovoltaic output prediction model predicts photovoltaic output values for 24 hours a day in the future, for a total of 24 time points, in units of 1 hour. The method comprises the following steps: the short-term photovoltaic output model inputs weather forecast information for 24 hours in the future, and outputs forecast results for the photovoltaic output for 24 hours in the future. Because the integrated model accuracy is higher than that of a single model, the embodiment establishes an XGBoost-RF model to realize short-term prediction of the photovoltaic. The XGBoost continuously generates new trees, each tree is learned based on the difference value between the last photovoltaic prediction result and the photovoltaic target value, and the results of all the trees are accumulated to be used as the final photovoltaic output prediction, so that the deviation of the model is reduced. RF establishes a forest in a random manner, and the forest is internally provided with a plurality of decision trees, and the decision trees are not related. After the forest is obtained, when a new meteorological data sample is input, each decision tree in the forest is respectively subjected to photovoltaic output prediction, and the photovoltaic output prediction results are obtained in a parallel mode.

In one embodiment, the load prediction model predicts the load of 24 hours a day into the future in units of 1 hour, and the total value of 24 time points. The method comprises the following steps: the short-term load model is input into historical meteorological data and historical load data with the time length of 4 days and the resolution of 1 hour, the LSTNet model is adopted for prediction, and the load prediction result of 24 hours in the future is output. Wherein the LSTNet model consists of a nonlinear part and a linear part. The nonlinear part is composed of a CNN convolution layer, an RNN loop layer, a jump layer and a linear part is composed of an autoregressive linear layer. LSTNet exploits the advantages of CNN, which can extract the volatility characteristics of the load, and RNN, which can capture the long-term trends of the load sequence. Finally, a traditional autoregressive linear model is added on the basis of a nonlinear neural network part, so that the nonlinear deep learning model has stronger robustness to a load sequence with scale violations. And superposing the output results of the nonlinear part and the linear part to obtain a final load prediction result.

As a preferred implementation mode, the dispatching of the household energy management system optimizing dispatching model is specifically to solve the optimal charging power and discharging power of the energy storage system in the future day, and the unit time length is 1 hour. And taking the output of the photovoltaic output prediction model and the output of the load prediction model as the photovoltaic output and the load of one day in the future. The scheduling targets include: 1) The reliable and stable power supply of the system is ensured; 2) The lowest electricity cost of the system is ensured; 3) And limiting the switching times of the energy storage system. The dispatching construction of the household energy optimization system is as follows:

Step 1: and constructing an energy storage system model. The energy storage system is mainly used for peak clipping and valley filling, and reducing the influence of the micro-grid on the power quality of the power distribution network. The energy storage system comprises a charging mode and a discharging mode, and a dynamic model of the energy storage system is specifically constructed as follows.

And (3) a constraint equation of the residual electric quantity of the energy storage system:

；

dynamic and constraint equations for energy storage system charging:

；

dynamic and constraint equations for energy storage system discharge:

；

wherein ,is the self-discharge rate of the energy storage system;The unit duration is T, and the scheduling period is T; andRespectively charging efficiency and discharging efficiency of the energy storage system;Is->State of charge of the time period energy storage system +.>State of charge indicating the start period of the scheduling period, < >>State of charge indicating the end period of the scheduling period, +.>Is the minimum value of the state of charge of the energy storage system;Is the maximum value of the state of charge of the energy storage system; andRespectively is an energy storage system->Discharge power and charge power during a period; andRespectively put for energy storage systemsMaximum value of electric power and charging power; andRespectively is an energy storage system->Status of time period>1 and->A discharge state and a charge state are respectively indicated by 0.

Step 2: and determining an objective function of the optimal scheduling model of the home energy management system. Under the condition of grid-connected operation, the electricity cost of the user is mainly electricity purchasing and selling cost. The objective function of the optimal scheduling model of the home energy management system is to minimize the electricity cost of one day and limit the number of charge-discharge switching times of one day, and can be expressed as:

；

Wherein F represents the electricity cost; t is a scheduling period, which is 24 hours in this embodiment; andThe unit electricity prices of electricity purchase and electricity selling to the power grid in different time periods are respectively shown; andRespectively representing the electricity purchasing power and the electricity selling power of the power grid in different time periods;The upper limit of the charge-discharge switching times of the energy storage system in the dispatching cycle is set; andRespectively is an energy storage system->The state of the period, the discharge state and the charge state are respectively represented by 1 and 0;A punishment function for switching times of charge and discharge of the energy storage system in a dispatching cycle is larger than limit times; because the switching times of the energy storage system have larger influence on the service life, the cost problem is considered, and the optimal solution is sought to ensure that the switching times of the energy storage system are +.>Within a time.

Step 3: and determining constraint conditions of the optimal scheduling model of the home energy management system. Specific constraints are as follows:

energy balance constraint:

；

wherein ,representing photovoltaic +.>Output power during a period>Representation->Load power in period> andRespectively is an energy storage system->Discharge power and charge power in a period of time +.> andRespectively the power grid is->Purchase power and sell power during a time period.

Mathematical model of energy storage system:

；

and (3) restraining the charge and discharge amount of the energy storage system: ；

Constraint of residual electric quantity of an energy storage system:；

mutual exclusion constraint of energy storage system:；

energy storage system cycle start-end energy storage balance constraint:；

the household energy management system is constrained with the interaction electric quantity of the power grid:

；

wherein ,representing maximum power of the household energy management system to the grid, < >>Representing the maximum power sold by the home energy management system to the grid.

In one embodiment, a quantum particle swarm algorithm is used to solve the optimal scheduling model of the home energy management system. The standard PSO algorithm is improved by utilizing quantum behaviors and probability expression characteristics in the quantum theory, the particle throwing speed and the particle trajectory are used for representing the condition of particles in the quantum range, and the problem of solving the optimal scheduling model of the household energy management system can be well solved. The algorithm sets a random energy storage system charging and discharging strategy in an initial stage, because photovoltaic output and load predicted values are known, the power purchase and power selling of a household energy system to a power grid in different periods in a dispatching period can be calculated, then the value of an objective function F is obtained, the quantum particle swarm algorithm minimizes the value of the objective function through continuous iteration, and finally the optimal energy storage system charging and discharging strategy is found out and used as a preliminary dispatching strategy. The particle swarm optimization is easy to fall into a local optimal problem, and the embodiment adopts the quantum particle swarm optimization to solve the optimal problem in the dispatching of the household energy management system. Compared with the particle swarm algorithm, the quantum particle swarm algorithm only needs to consider the position variable, so that the algorithm is easier to encode, the algorithm can also enable particles to search in the whole quantum space, and the capability of jumping out of a local optimal solution and global convergence is enhanced on the basis of increasing the stability of the algorithm.

After the preliminary scheduling strategy is obtained, the household energy management system is instructed to operate, trial operation is carried out to carry out self-learning, and the preliminary scheduling strategy is updated and adjusted. The prediction result of the photovoltaic output prediction model in the running period of the test run is second photovoltaic output prediction data, the prediction result of the load prediction module is second load prediction data, the photovoltaic output result actually recorded by the system is first photovoltaic output data, and the load result actually recorded by the system is first load data.

Fig. 3 is a flowchart of another embodiment of a self-learning home energy management system optimization scheduling method, as shown in fig. 3, in step S104, a process of establishing an error probability distribution model according to an error between the second photovoltaic output prediction data and the first photovoltaic output data and an error between the second load prediction data and the first load data includes steps S200 and S201:

s200, an error sample data set is manufactured according to errors of the second photovoltaic output prediction data and the first photovoltaic output data and errors of the second load prediction data and the first load data;

s201, a discrete probability distribution model of the photovoltaic output error and a discrete probability distribution model of the load error are established according to the error sample data set.

As a preferred embodiment, the self-learning adjustment period is set to. Operating a cycle +.>And then, performing self-learning and self-adjustment of the online scheduling strategy. Since the schedule is in days ∈ ->Also in days. It is assumed that after one period of operation +.>Photovoltaic output prediction data and load prediction data. In addition, actual photovoltaic output data and load data over the period may be obtained as well. Based on these data, error probability distribution models are established for photovoltaic output prediction and load prediction, respectively.

Step 1: an error sample dataset of photovoltaic and load power is made. The embodiment respectively counts the latest adjustment periodsThe true value and the predicted value of the photovoltaic output force and the load in the system are respectively manufactured into error sample data sets of the photovoltaic output force and the load, and the error sample data sets of the photovoltaic output force and the load can be obtained>Each error sample is an error of 24 photovoltaic or load predicted values and a true value (predicted value minus true value) of a day, namely the error sample data set shares +.>Each sample has 24 features.

Step 2: and constructing an error probability distribution model of the photovoltaic output and the load. Continuous data discretization refers to segmenting continuous data into sections that are discretized in one segment. The discretization method can comprise two embodiments, wherein one is an unsupervised learning mode and the other is a supervised learning mode. The principle of segmentation is based on equidistant, equifrequency, clustering or optimization methods. The equal interval method can better keep the complete distribution of the data, so that the embodiment adopts the equidistant discretization method to discretize the continuous photovoltaic output error data and the continuous load error data. Features for photovoltaic output error samples And characteristics of the load error samples->, whereinIn this embodiment, the characteristic ++is obtained by the equidistant discrete method one by one> andDiscretizing, and fixing into +.>Discrete intervals> andCharacteristics of the photovoltaic output error samples are respectively represented +.>And characteristics of the load error samples->In->Error values of discrete intervals, wherein ∈>. Further counting the number of elements occupied by each discrete interval of each characteristic in the photovoltaic output and load error samples> and. Will be andThe ratio of the number of the characteristic total elements to the number of the characteristic total elements is used as the probability of the discrete interval, namely +.>，Photovoltaic output error and load error are respectively in +.>Probability of discrete interval. To sum upIn this embodiment, the discrete probability distribution model of the photovoltaic output error is successfully built>And a discrete probability distribution model of load error +.>As shown in tables 1 and 2, respectively.

TABLE 1 discrete probability distribution model for photovoltaic output errors

TABLE 2 discrete probability distribution model of load error

And solving a preliminary scheduling strategy by adopting the established error probability distribution model. First, a probability distribution of future photovoltaic output and load is determined. The photovoltaic output predicted value and the load predicted value of one day in the future can be obtained based on the photovoltaic output predicted model and the load predicted model, and the photovoltaic output error and the load error discrete probability distribution model obtained in the step 2 is adopted to carry out error correction on the photovoltaic output and the load predicted value of one day in the future so as to obtain the photovoltaic output probability distribution of one day in the future And load probability distribution->As shown in tables 3 and 4.

TABLE 3 photovoltaic output discrete probability distribution model for future day

TABLE 4 load discrete probability distribution model for future day

Because the preliminary scheduling models are all based on the photovoltaic output predicted value of the future dayAnd load predictive value->Scheduling is performed if the probability distribution of photovoltaic output based on the future day +.>And load probability distribution->Scheduling is performed, and the scheduling policy needs to be readjusted. The method comprises the steps that a random energy storage system charging and discharging strategy is set in an initial stage by a quantum particle swarm algorithm, a commercial power probability distribution model in a dispatching period is needed to be calculated by combining the calculated photovoltaic output probability distribution and the calculated load probability distribution, the expectation of the commercial power at the corresponding moment is calculated according to the probability models of the commercial power at different moments, the value of an objective function F is obtained, the algorithm minimizes the value of the objective function through continuous iteration, and finally the optimal energy storage system charging and discharging strategy is found. Thus, the online self-learning phase requires the establishment of a mains probability model and the updating of the objective function F.

Because the photovoltaic output probability distribution and the load probability distribution are mutually independent, when the grid-connected operation is performed, if the photovoltaic output cannot meet the load demand The load power supply demand prioritizes the internal energy storage supply, otherwise insufficient power is supplied by the grid. The probability distribution of the corresponding purchase power is as followsThe following is shown:

；

wherein ,。

if the photovoltaic output is sufficient and meets the load power supply requirement, the redundant electric quantity is charged for the energy storage system preferentially, and when the photovoltaic output meets the load and the energy storage system is charged, the residual electric quantity is fed into the power grid. Corresponding probability distribution of electricity selling powerThe calculation is as follows:

；

and finally, calculating the expectations corresponding to the mains supply at different moments according to the mains supply probability model, and updating the value of the objective function F as shown below.

；

In one example, as shown in fig. 3, the home energy management system optimization scheduling method with self-learning capability according to another embodiment further includes step S300 and step S301:

s300, acquiring third photovoltaic output prediction data, third load prediction data, second photovoltaic output data and second load data in an adjustment period based on the operation of a home energy management system;

s301, updating an error probability distribution model according to errors of the third photovoltaic output prediction data and the second photovoltaic output data and errors of the third load prediction data and the second load data.

Through step S300 and step S301, self-updating of the error probability distribution model is achieved.

As a preferred embodiment, at each subsequent adjustment periodError data of corresponding photovoltaic output and load can be obtained, and the embodiment realizes self-updating of the error probability distribution model. The method comprises the following steps:

step 1: for each newly obtained photovoltaic output and load error sample, i.e. during the last adjustment periodThe photovoltaic output error sample and the load error sample are divided intoDiscrete intervals> andCharacteristics of the photovoltaic output error samples are respectively represented +.>And characteristics of the load error samples->In->Error values of discrete intervals, the embodiment counts the number of elements occupied by each discrete interval of each characteristic in the latest photovoltaic output and load error samples> andPhotovoltaic output error and load error are at +.>Probability of discrete interval andThe updating method can be updated as follows:

；

wherein ,is a smaller value and the magnitude of the value depends on the importance of the error sample in the current adjustment period.

Step 2: the probability of each discrete interval for each feature in the photovoltaic output and load error samples is normalized as:

；

After updating the error probability distribution model of the photovoltaic output and the load, the solution of the preliminary scheduling strategy adopts the self-learning mode.

In order to better explain the effects of the embodiments of the present disclosure, the following shows technical effects in a specific application example.

The construction site of the home energy management system of this embodiment is located in south australia, and this embodiment will be described in further detail with reference to the actual operation diagram. In the system of the specific application example, the rated capacity of the photovoltaic power generation system installed in the home is 10.2kWp, and the home energy management system can predict the power generation amount of the photovoltaic power generation system in each period of the following day. In order to have a good effect on the dispatching of the energy storage system, 2 lead-acid batteries of 5.12kWh are installed in the household, and the battery energy storage systemThe maximum capacity of the system is about 10.24kWh, and the charge and discharge efficiency is highAll take 1, maximum charge and discharge power +.> andThe initial state of charge value and the final state of charge value of the storage battery are respectively 0.3, the minimum state of charge value and the maximum state of charge value of the storage battery are respectively 0.1 and 1, and the self discharge rate of the energy storage system is>. Because of the limitation of household equipment and networking conditions, the specific application example does not consider the fact that a household management system sells electricity to a large power grid, and the maximum electricity purchasing power is realized 3kW was taken.

Taking a certain intelligent household in summer as an example, the scheduling period is one day, and the electric energy cost of running in one day is counted as a target value of the specific application example. Scheduling period of systemFor 24h, it is divided into 24 time periods per unit time, i.e. +.>. The method and the system take a daily zero point as a time boundary, and divide a daily scheduling period from the yesterday zero point to the today zero point, namely, the system schedules a household energy system in the next day at the daily zero point according to user parameter input.

Using historical data (from 2021, 8, 21 to 2022, 8, 21), the conventional scheduling method (without self-learning capability) was constructed using the content initialization portion of the embodiments of the present disclosure.

The specific application setting system is online to learn the adjustment cycle certainlyFor 30 days, from day 22 of year 2022 8 to day 22 of year 2022 9, the obtained traditional scheduling method is adopted to run for one month on line. The practical application example calculates the actual value and the predicted value of the photovoltaic output and the load in the self-learning adjustment period, and respectively prepares error sample data sets of the photovoltaic output and the load. Then, for 24 time periods of a day, the error data of the photovoltaic output and the load in each time period are discretized by adopting an equidistant discretization method. The specific application example sets that error samples of photovoltaic output and load in each period are fixedly divided into 20 different discrete intervals, and then the number of elements in each discrete interval in each period is counted one by one. And finally, calculating the ratio of the number of elements in each interval to the total number of elements one by one, and taking the ratio as the probability of the discrete interval. In order to better demonstrate the modeling process, this particular application example focuses on the analysis process of the 12-point period in the day schedule. The specific application example shows an error probability model of the photovoltaic output and the load of the 12-point period in the form of a histogram, as shown in fig. 4 and 5, and further shows the photovoltaic output and the probability of the division of the load error interval and the correspondence between different intervals in the period by using tables 5 and 6 respectively. Thus, an error probability distribution model of the photovoltaic output and the load can be obtained.

TABLE 5 interval division of photovoltaic output errors at 12 Point period and probabilities of correspondence between different intervals

TABLE 6 interval division of 12 Point period load errors and probabilities corresponding to different intervals

The specific application example integrates the error information model obtained based on learning into a scheduling algorithm, and performs real-time scheduling on the day of 9 months and 23 years of 2022, and compares the error information model with the traditional scheduling algorithm (without self-learning capability).

This particular applicationAn example is the real-time electricity price (RTP) profile of 2022, 9, 23 days provided by AEMO in dollars per watt-hour ($/Wh), as shown in FIG. 6. The household energy management system respectively adopts an Xgboost-RF fusion model and an LSTNet neural network model to predict and obtain photovoltaic output and load of 9 months and 23 days in advance, and adjusts the period based on the self-learningError correction is carried out on predicted values of the photovoltaic output and the load of the future day by using the photovoltaic output and load error probability distribution model in the future to obtain photovoltaic output probability distribution of the future day>And load probability distribution->. Probability distribution of photovoltaic output based on future day +.>And load probability distribution->Fig. 7 shows the process of finding the optimal solution using a quantum group optimization algorithm, where the optimal solution determined last may result in a minimum power cost of 0.1882 dollars/Wh for day 9 and 23. Fig. 8 shows the dispatching result of the on-line self-learning corrected household energy management system to the storage battery and the power grid, and the energy conservation law is satisfied. In connection with fig. 6 and 8, it can be found that the purpose of using the storage battery by the user is to store the electric energy in the period of low electricity prices, and to release the electric energy in the period of relatively high electricity prices, so as to reduce the electricity consumption cost. It can also be found that grid output is more used when electricity prices are low. In conclusion, the scheduling strategy can reduce electricity consumption to the greatest extent.

In addition, since the number of times of cyclic charge and discharge of the storage battery has a great influence on the service life thereof, under the condition of real-time electricity price, if the storage battery is frequently charged and discharged only for helping a user save electricity charge, the service life of the storage battery can be greatly shortened. Therefore, in this specific application, the number of times of battery switching is desirably controlled to be within 5 times, taking into consideration the parameter of the number of times of battery charging and discharging in the scheduling target of the home energy consumption management system. Fig. 9 is an operating state of the battery after the on-line self-learning correction, wherein 1 indicates that the battery is in a charged state and 0 indicates that the battery is in a discharged state. As can be seen from fig. 9, the number of times of switching the battery is only 4 in 24 scheduling periods, which satisfies the requirements.

In order to verify the effectiveness of the specific application, the specific application adopts a traditional scheduling method to solve the optimal scheduling strategy for 9 months and 23 days. As shown in FIG. 10, the resulting optimal solution may result in a minimum power cost of 0.6116 dollars/Wh for 9 months and 23 days. Fig. 11 shows the dispatching result of the household energy management system to the storage battery and the power grid based on the traditional dispatching method, and the law of conservation of energy is also satisfied. By contrast, through one month of self-learning, the method provided by the specific application example can save 0.4234 dollars/Wh of electricity consumption in the experimental day, and the effectiveness of the online self-learning method is verified.

According to the home energy management system optimization scheduling method with the self-learning capability, constraint conditions of the energy storage system model are determined, and objective functions and constraint conditions of the home energy management system optimization scheduling model are determined according to the first photovoltaic output prediction data, the first load prediction data and the constraint conditions of the energy storage system model, so that a preliminary scheduling strategy is determined. Further, an error probability distribution model is established according to errors of the second photovoltaic output prediction data and the first photovoltaic output data and errors of the second load prediction data and the first load data, and a household energy management system optimization scheduling model is updated according to the error probability distribution model to obtain an updated scheduling strategy which is configured to schedule the household energy management system. Based on the method, the scheduling strategy is updated through the actual error, the scheduling accuracy of the home energy management system is improved, and the scheduling effect is optimized.

The embodiment of the disclosure also provides an optimized dispatching device of the home energy management system with self-learning capability.

Fig. 12 is a block diagram of a self-learning-capable home energy management system optimal scheduling apparatus, and as shown in fig. 12, the self-learning-capable home energy management system optimal scheduling apparatus according to an embodiment includes:

The data acquisition module 100 is used for acquiring first photovoltaic output prediction data and first load prediction data;

a condition determining module 101, configured to determine constraint conditions of the energy storage system model;

the strategy determining module 102 is configured to determine an objective function and constraint conditions of the optimal scheduling model of the home energy management system according to the first photovoltaic output prediction data, the first load prediction data and constraint conditions of the energy storage system model, so as to determine a preliminary scheduling strategy;

the data operation module 103 is configured to instruct the home energy management system to operate according to the preliminary scheduling policy, and obtain second photovoltaic output prediction data, second load prediction data, first photovoltaic output data and first load data in an operation period;

the error analysis module 104 is configured to establish an error probability distribution model according to the errors of the second photovoltaic output prediction data and the first photovoltaic output data and the errors of the second load prediction data and the first load data;

the scheduling updating module 105 is used for updating the optimal scheduling model of the home energy management system according to the error probability distribution model to obtain an updated scheduling strategy; wherein the updated scheduling policy is configured to schedule the home energy management system.

At least one embodiment of the present disclosure further provides a data scheduling apparatus. Fig. 13 is a schematic block diagram of a data scheduling apparatus provided in at least one embodiment of the present disclosure. For example, as shown in fig. 13, the data scheduling apparatus 20 may include one or more memories 200 and one or more processors 201. Memory 200 is used to non-transitory store computer-executable instructions; the processor 201 is configured to execute computer-executable instructions that, when executed by the processor 201, may cause the processor 201 to perform one or more steps in a self-learning-capable home energy management system optimization scheduling method in accordance with any embodiment of the present disclosure.

The specific implementation and relevant explanation content of each step of the self-learning home energy management system optimization scheduling method can be referred to the relevant content in the embodiment of the power load prediction model training method, and will not be described herein. It should be noted that the components of the data scheduler 20 shown in fig. 13 are only exemplary and not limiting, and that the data scheduler 20 may also have other components according to practical application needs.

In one embodiment, the processor 201 and the memory 200 may communicate with each other directly or indirectly. For example, the processor 201 and the memory 200 may communicate over a network connection. The network may include a wireless network, a wired network, and/or any combination of wireless and wired networks, the disclosure is not limited in type and function of the network herein. For another example, processor 201 and memory 200 may also communicate via a bus connection. The bus may be a peripheral component interconnect standard (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. For example, the processor 201 and the memory 200 may be disposed at a remote data server (cloud) or a distributed energy system (local), or may be disposed at a client (e.g., a mobile device such as a mobile phone). For example, the processor 201 may be a Central Processing Unit (CPU), tensor Processor (TPU), or graphics processor GPU, among other devices having data processing and/or instruction execution capabilities, and may control other components in the data prediction apparatus 20 to perform desired functions. The Central Processing Unit (CPU) can be an X86 or ARM architecture, etc.

In one embodiment, memory 200 may comprise any combination of one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and/or non-volatile memory. Volatile memory can include, for example, random Access Memory (RAM) and/or cache memory (cache) and the like. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, erasable programmable read-only memory (EPROM), portable compact disc read-only memory (CD-ROM), USB memory, flash memory, and the like. One or more computer-executable instructions may be stored on the computer-readable storage medium and the processor 201 may execute the computer-executable instructions to implement the various functions of the data prediction device 20. Various applications and various data, as well as various data used and/or generated by the applications, etc., may also be stored in the memory 200.

It should be noted that, the data scheduling device 20 may achieve similar technical effects as the foregoing optimized scheduling method of the home energy management system with self-learning capability, and the repetition is omitted.

At least one embodiment of the present disclosure also provides a non-transitory computer-readable storage medium. Fig. 14 is a schematic diagram of a non-transitory computer-readable storage medium provided by at least one embodiment of the present disclosure. For example, as shown in FIG. 14, one or more computer-executable instructions 301 may be non-transitory stored on a non-transitory computer-readable storage medium 30. For example, the computer-executable instructions 301, when executed by a computer, may cause the computer to perform one or more steps in a self-learning-capable home energy management system optimization scheduling method according to any embodiment of the present disclosure.

In one embodiment, the non-transitory computer readable storage medium 30 may be applied in the data scheduler 20 described above, which may be, for example, the memory 200 in the data scheduler 20.

In one embodiment, the description of the non-transitory computer readable storage medium 30 may refer to the description of the memory 200 in the embodiment of the data scheduling apparatus 20, and the repetition is omitted.

It should be noted that the memory 200 stores different non-transitory computer executable instructions that, when executed by the processor 201, may cause the processor 201 to perform one or more steps in a self-learning-capable home energy management system optimization scheduling method according to any of the embodiments of the present disclosure, the data scheduling apparatus 20 corresponds to a load prediction apparatus.

For the purposes of this disclosure, the following points are also noted:

(1) The drawings of the embodiments of the present disclosure relate only to the structures to which the embodiments of the present disclosure relate, and reference may be made to the general design for other structures.

(2) In the drawings for describing embodiments of the present invention, thicknesses and dimensions of layers or structures are exaggerated for clarity. It will be understood that when an element such as a layer, film, region or substrate is referred to as being "on" or "under" another element, it can be "directly on" or "under" the other element or intervening elements may be present.

(3) The embodiments of the present disclosure and features in the embodiments may be combined with each other to arrive at a new embodiment without conflict. The above is only a specific embodiment of the present disclosure, but the protection scope of the present disclosure is not limited thereto, and the protection scope of the present disclosure should be subject to the protection scope of the claims

The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.

The foregoing examples represent only a few embodiments of the present application, which are described in more detail and are not thereby to be construed as limiting the scope of the claims. It should be noted that it would be apparent to those skilled in the art that various modifications and improvements could be made without departing from the spirit of the present application, which would be within the scope of the present application. Accordingly, the scope of protection of the present application is to be determined by the claims appended hereto.

Claims

1. The household energy management system optimization scheduling method with the self-learning capability is characterized by comprising the following steps of:

determining constraint conditions of an energy storage system model;

the energy storage system model comprises a mathematical model and constraint conditions; the constraint conditions of the energy storage system model comprise a charge and discharge amount constraint condition, a residual electric quantity constraint condition, a mutual exclusion constraint condition and a period start and end energy storage balance constraint condition of the energy storage system;

the mathematical model of the energy storage system model has the following formula:

；

the charge and discharge amount constraint conditions of the energy storage system model are as follows:

；

the remaining capacity constraint condition of the energy storage system model is as follows:

；

the mutual exclusion constraint condition of the energy storage system model is as follows:

；

the cycle start and end energy storage balance constraint conditions of the energy storage system model are as follows:

；

wherein ,is the self-discharge rate of the energy storage system;The unit duration is T, and the scheduling period is T; andRespectively charging efficiency and discharging efficiency of the energy storage system;Is->State of charge of the time period energy storage system +.>State of charge indicating the start period of the scheduling period, < >>State of charge indicating the end period of the scheduling period, +.>Is the minimum value of the state of charge of the energy storage system;Is the maximum value of the state of charge of the energy storage system; / > andRespectively is an energy storage system->Discharge power and charge power during a period; andThe maximum values of the discharging power and the charging power of the energy storage system are respectively; andRespectively is an energy storage system->Status of time period>1 and->A discharge state and a charge state are respectively represented by 0;

the objective function of the optimal scheduling model of the household energy management system is to minimize the electricity cost and the charge-discharge switching times of the energy storage system in the scheduling period, and the objective function is as follows:

；

wherein F represents electricity cost; t is a scheduling period; andRespectively indicate->The unit electricity price of electricity purchase and electricity selling to the power grid in the period; andRespectively indicate->Purchasing and selling electric power to the power grid in a period;The upper limit of the charge-discharge switching times of the energy storage system in the dispatching cycle is set; andRespectively is an energy storage system->Status of time period of1 and->A discharge state and a charge state are respectively represented by 0;A punishment function for switching times of charge and discharge of the energy storage system in a dispatching cycle is larger than limit times;

Wherein the optimal solution sought by the objective function is that the charge-discharge switching times of the energy storage system is less than or equal toSecondary times;

the electricity purchasing powerThe formula is as follows:

；

the electricity selling powerThe formula is as follows:

；

wherein ,representing photovoltaic +.>Output power during a period>Representation->Load power in period> andRespectively is an energy storage system->Discharge power and charge power during a period;

the constraint conditions of the optimal scheduling model of the household energy management system comprise energy balance constraint conditions and constraint conditions of interaction electric quantity of the household energy management system and the electric network;

the energy balance constraint bar has the following formula:

；

the constraint condition of the interaction electric quantity between the household energy management system and the electric network is as follows:

；

wherein ,representing maximum power of the household energy management system to the grid, < >>Representing the maximum electricity selling power of the household energy management system to the power grid;

establishing an error probability distribution model according to the errors of the second photovoltaic output prediction data and the first photovoltaic output data and the errors of the second load prediction data and the first load data;

Updating the optimal scheduling model of the home energy management system according to the error probability distribution model to obtain an updated scheduling strategy; wherein the updated scheduling policy is configured to schedule the home energy management system.

2. The home energy management system optimizing scheduling method with self-learning ability according to claim 1, further comprising the steps of:

acquiring third photovoltaic output prediction data, third load prediction data, second photovoltaic output data and second load data in an adjustment period based on the operation of the home energy management system;

and updating the error probability distribution model according to the errors of the third photovoltaic output prediction data and the second photovoltaic output data and the errors of the third load prediction data and the second load data.

3. The method for optimizing and scheduling a home energy management system with self-learning ability according to claim 1, wherein the process of establishing an error probability distribution model according to the errors of the second photovoltaic output prediction data and the first photovoltaic output data and the errors of the second load prediction data and the first load data comprises the steps of:

An error sample data set is manufactured according to errors of the second photovoltaic output prediction data and the first photovoltaic output data and errors of the second load prediction data and the first load data;

and establishing a discrete probability distribution model of the photovoltaic output error and a discrete probability distribution model of the load error according to the error sample data set.

4. The method for optimizing and scheduling a home energy management system with self-learning ability according to any one of claims 1 to 3, wherein the process of updating the home energy management system optimizing and scheduling model according to the error probability distribution model and obtaining the updated scheduling policy comprises the steps of:

and updating the first photovoltaic output prediction data and the first load prediction data according to an error probability distribution model.

5. An optimized dispatching device of a home energy management system with self-learning capability is characterized by comprising:

the condition determining module is used for determining constraint conditions of the energy storage system model;

；

wherein ,is the self-discharge rate of the energy storage system;The unit duration is T, and the scheduling period is T; andRespectively charging efficiency and discharging efficiency of the energy storage system;Is->State of charge of the time period energy storage system +.>State of charge indicating the start period of the scheduling period, < >>State of charge indicating the end period of the scheduling period, +.>Is the minimum value of the state of charge of the energy storage system;For storingMaximum value of the state of charge of the system; andRespectively is an energy storage system->Discharge power and charge power during a period; andThe maximum values of the discharging power and the charging power of the energy storage system are respectively; andRespectively is an energy storage system->Status of time period>1 and->A discharge state and a charge state are respectively represented by 0;

；

wherein F represents electricity cost; t is a scheduling period; andRespectively indicate->The unit electricity price of electricity purchase and electricity selling to the power grid in the period; andRespectively indicate->Purchasing and selling electric power to the power grid in a period;The upper limit of the charge-discharge switching times of the energy storage system in the dispatching cycle is set; andRespectively is an energy storage system->Status of time period ofIs 1 anda discharge state and a charge state are respectively represented by 0;A punishment function for switching times of charge and discharge of the energy storage system in a dispatching cycle is larger than limit times;

the electricity purchasing powerThe formula is as follows:

；

the electricity selling powerThe formula is as follows:

；

The energy balance constraint bar has the following formula:

；

the scheduling updating module is used for updating the optimal scheduling model of the home energy management system according to the error probability distribution model to obtain an updated scheduling strategy; wherein the updated scheduling policy is configured to schedule the home energy management system.

6. A data scheduling apparatus comprising:

one or more memories non-transitory storing computer-executable instructions;

One or more processors configured to execute the computer-executable instructions, wherein the computer-executable instructions, when executed by the one or more processors, implement the self-learning-enabled home energy management system optimization scheduling method of any one of claims 1 to 4.

7. A non-transitory computer readable storage medium storing computer executable instructions which when executed by a processor implement the self-learning-enabled home energy management system optimization scheduling method of any one of claims 1 to 4.