CN105738974B - The forecasting procedure and system of air heavily contaminated weather - Google Patents

Abstract

A kind of forecasting procedure of air heavily contaminated weather, including：Collect, obtain the history meteorological data for the long-term sequence for waiting for forecast area；Several meteorologic parameters are chosen as impact factor；By way of fitting, air heavily contaminated prognostic equation is established using several meteorologic parameters of selection as the factor, the prediction numerical value of the Meteorological Models WRF meteorologic parameters forecast is substituted into the prognostic equation, following heavily contaminated weather is predicted by solving equation.And a kind of forecast system of air heavily contaminated weather.The heavy air pollution process that the present invention combines weather typing and meteorological element discriminant equation that future may occur for the first time differentiates；The representative meteorological factor quantity that the discriminant equation of foundation is chosen is few, and heavily contaminated case differentiates that effect is good；It is to heavily contaminated statistical fluctuation technology, effective supplement of Numerical Forecast Technology, method is simple and practicable, and input is less, and the pollution prewarning that can effectively attach most importance to provides technical support；Establish more simple and practical operation interface.

Description

Method and system for forecasting heavy air pollution weather

technical field

The invention relates to the field of weather forecasting, and more particularly to a forecasting method and system for heavy air pollution weather.

Background technique

In recent years, with the development of my country's economy and the expansion of production capacity, large-scale smog and other heavily polluted weather have begun to appear in the north, which has seriously affected normal production and life and damaged the physical and mental health of the majority of people. To this end, some areas have begun to implement a red warning mechanism for heavily polluted weather. By issuing red warning information, there are expected suspensions of classes and work, as well as traffic restrictions, to control further emissions from pollution sources and prevent primary school students and young children from being exposed to heavily polluted outdoors.

However, the current heavy pollution forecasting mechanism mainly relies on manual forecasting, and there are large errors. Once the red warning is activated, but the weather conditions do not actually deteriorate, it will bring great losses to the economy of our country and the lives of the people. . In addition, my country only began to officially monitor and release PM2.5 values in 2013. Due to the lack of previous data, many existing prediction methods cannot be carried out due to lack of data. How to accurately predict the occurrence and development of heavy air pollution after a certain period of time is a difficult problem that needs to be solved urgently.

Contents of the invention

In view of this, the object of the present invention is to provide a method and system for forecasting heavy air pollution weather.

In order to achieve the above object, as an aspect of the present invention, the present invention provides a method for forecasting heavy air pollution weather, comprising the following steps:

Step S1: Collect and obtain historical meteorological data of the area to be forecasted;

Step S2: Select several meteorological parameters as influencing factors;

Step S3: By means of fitting, establish a heavy air pollution forecasting equation with selected meteorological parameters as factors, substitute the predicted and/or measured values of the meteorological parameters into the forecasting equation, and calculate the Whether to predict heavy pollution weather in the future.

Preferably, the influencing factors selected in step S2 are average wind speed, 24-hour variable pressure, meteorological variable of temperature difference between 850 and 1000hP, and yesterday's national control fine particle average concentration.

As preferably, the prediction equation fitted in step S3 is:

Among them, c is the predicted daily average concentration of PM2.5 on the day to be predicted, a ₀ is a constant, a1, a2, a3, and a4 are regression coefficients; x1, x2, and x3 are the 24 hours of the day to be predicted The predicted value of average wind speed, 24-hour variable pressure, and the temperature difference between 850 and 1000hPa at 08:00, x4 is the measured or predicted value of the average concentration of fine particulate matter under national control yesterday.

Preferably, the day to be predicted includes today, tomorrow and the day after tomorrow, corresponding to the next 24 hours, 48 hours and 72 hours respectively.

As a preference, the values of x1, x2, and x3 are selected from the simulation forecast results of the WRF model. The initial and boundary data of the WRF model are the reanalysis daily data GFS of NCAR and NCEP, with a resolution of 1°×1° and a time resolution of 6h (00:00, 06:00, 12:00, 18:00); the topographic and underlying surface input data are from USGS 30s global topographic and MODIS underlying surface classification data, respectively. For x4, when predicting today's heavily polluted weather conditions, use the measured value of the nationally controlled fine particulate matter concentration yesterday; when predicting tomorrow's or the day after tomorrow's heavily polluted weather conditions, use yesterday's nationally controlled fine particulate matter average value relative to tomorrow or the day after tomorrow predicted concentration.

As a preference, the forecast equation for 2013 fitted to the urban area of Beijing is: c=103.23-24.974x1-3.8127x2+1.5025x3+0.53945x4.

As another aspect of the present invention, the present invention also provides a forecasting system for heavy air pollution weather, said forecasting system implements the forecasting method for heavy air pollution weather as described above based on matlab software, to predict the weather at a certain time in the future Forecast of heavily polluted weather.

Based on the above technical solutions, it can be seen that the forecasting method and system of the present invention have the following beneficial effects:

(1) For the first time, combined with weather classification and meteorological element discriminant equation to discriminate the process of heavy pollution that may occur in the future; this method is the first to classify the weather system that affects heavy pollution in Beijing, and combined with artificial forecast to establish a statistical discrimination of heavy pollution Equation is a heavy pollution forecasting technique;

(2) The number of representative meteorological factors selected by the established discriminant equation is small, and the discrimination effect of heavy pollution cases is good;

(3) The method of the present invention is an effective supplement to heavy pollution statistical forecasting technology and numerical forecasting technology. The method is simple and easy to implement, and has less investment, can effectively provide technical support for heavy pollution early warning, and is convenient and feasible for application demonstration and promotion;

(4) Based on the combination of pollutants, meteorological measurement database and WRF forecast database for the first time based on matlab, a relatively simple and practical method operation interface has been established; this method will provide a better idea for the future dynamic statistical forecast of heavy air pollution in Beijing .

Description of drawings

Figure 1 is a comparison chart of statistical discriminant equation prediction and actual measurement in Beijing in 2013;

Figure 2 is a comparison chart of the 58-day heavy pollution statistics and forecast in Beijing in 2013 and the actual measurement;

Fig. 3 is the flowchart of forecasting method of the present invention;

Fig. 4 (a), 4 (b) are respectively the result curve diagram that the forecasting method of the present invention carries out the verification assessment to PM2.5 concentration during 2 red warnings in Beijing in December, 2015;

Fig. 5 is the software operation interface of the forecasting system of the present invention.

Detailed ways

In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with specific embodiments and with reference to the accompanying drawings.

As shown in Figure 3, the present invention discloses a method for forecasting heavy air pollution weather, comprising the following steps:

Step S1: Collect and obtain the historical meteorological data of the area to be forecasted; among them, in order to improve the fitting accuracy, it is possible to manually identify the weather background field where heavy pollution may occur, and focus on some meteorological events that have a greater impact on the heavily polluted weather parameter. In addition, the above-mentioned historical meteorological data are preferably long-term historical meteorological data, such as historical meteorological data of more than 6 months, or even 12 months, 2 years and other long time periods.

Step S2: Select a number of meteorological parameters as influencing factors, and the selection criteria can be based on automatic ranking, for example, from the perspective of the strength of correlation between historical meteorological parameters and PM2.5 concentration.

Step S3: By means of fitting, the air heavy pollution forecasting equation is established with selected meteorological parameters as factors, and the meteorological parameters forecasted by the meteorological model WRF are substituted into the forecasting equation. Make predictions.

In the above step S2, there are many meteorological parameters to choose from, such as temperature, humidity, wind speed, wind direction, air pressure, distribution of high pressure troughs, and so on. Preferably, the influencing factors selected in step S2 are, for example, the average wind speed, 24-hour variable pressure, the difference between 850 and 1000hP temperature meteorological variables, and the average concentration of fine particulate matters controlled by the state yesterday. After trial and error, the selection of these four variables can best approximate the change of pollutants and predict whether it will be heavily polluted at a certain time in the future.

As preferably, the prediction equation fitted in step S3 is, for example:

Among them, c is the predicted concentration of PM2.5 on the day to be predicted, a ₀ is a constant, a1, a2, a3, and a4 are regression coefficients; x1, x2, and x3 respectively represent the 24-hour average wind speed on the day to be predicted , 24-hour variable pressure, and the predicted value of the temperature difference between 850 and 1000hPa at 08:00, and x4 represents the measured or predicted value of the average concentration of fine particulate matter under national control yesterday.

For this forecasting equation, it can be used to forecast the air pollution situation of today, tomorrow and the day after tomorrow (that is, in the next 24h, 48h, 72h).

As a preference, the above-mentioned values of x1, x2, and x3 are selected from the simulation forecast results of the WRF model. The initial and boundary data of the WRF model are the reanalysis daily data GFS of NCAR and NCEP, with a resolution of 1°×1° and a time resolution of 6h (00:00, 06:00, 12:00, 18:00); input data of topography and underlying surface are from USGS 30s global topography and MODIS underlying surface classification data respectively. For x4, when predicting today's heavily polluted weather conditions (c value), directly use the measured value of yesterday's national control mean value; when predicting tomorrow's heavily polluted weather conditions (c value), due to the relative )’s national control mean value (x4) has not yet been calculated, you can first use the above forecasting equation to predict today’s c value, and use this c value as a forecast value relative to tomorrow’s “yesterday’s national control mean value” (x4). Similarly, for predicting the heavily polluted weather conditions (c value) of the day after tomorrow, the c values of today, tomorrow and the day after tomorrow are predicted in sequence, so as to obtain a forecast of whether the day after tomorrow is heavily polluted weather.

After careful research, for the urban area of Beijing, the fitting forecast equation in 2013 is: c=103.23-24.974x1-3.8127x2+1.5025x3+0.53945x4. The prediction equation can more accurately fit the pollutant distribution data in 2013. For 2014 and 2015, the historical fine particulate matter and meteorological observation data from 2014 to 2015 can be added to the database, and a new statistical discriminant equation can be updated and established to obtain more accurate results based on more historical data. The new forecasting equation is used to dynamically verify and evaluate the two red warnings in Beijing in 2015. The predicted value and the measured value have a good time trend. The forecasting equation can better predict the process of heavy pollution.

The invention also discloses a forecasting system for heavily air-polluted weather. The forecasting system executes the above-mentioned heavy-air-polluted weather forecasting method based on matlab to predict the heavy-polluted weather for a certain period of time in the future. Based on matlab, the forecasting system combines the pollutants with the weather measurement database and the WRF forecast database for the first time, and establishes a relatively simple and practical method operation interface.

The forecast system includes pollutant and meteorological observation database, WRF forecast database and user operation platform based on MATLAB. The system adopts layered and distributed mechanism design, and the whole system is divided into two layers: data storage layer and system operation layer. The data storage layer mainly stores the manually input data and the system output results. As shown in Figure 5, the manually input data includes the measured concentration and forecast date of fine particulate matter yesterday and the forecasted value of meteorological factors on that day, and the output result of the system is mainly the predicted concentration value. The system operation layer is a program that reads the initial input data, performs related operations and displays the results. The forecast system as a whole has dynamic characteristics, new pollution samples and meteorological samples will be added to the system data set in a timely manner, and the forecast system will be adjusted so that the model system can reflect the changing pollution conditions. The current operation interface and display method are relatively simple, and will be further improved in the near future. Compared with the numerical forecast system, the statistical forecast system is easy to operate, does not require more professional computing hardware equipment, and does not require high programming ability of operation and maintenance personnel. application tools.

In order to better illustrate the technical solution of the present invention, the Beijing area is taken as an example below to further illustrate the technical solution of the present invention. However, it should be clear that the solution of the present invention is not limited to the Beijing area, and is equally effective for other areas.

The traditional heavy air pollution in Beijing can be divided into four types: static accumulation type, dust type, composite type and special type. The present invention conducts classification research based on the weather situation of heavy pollution. Analyzing the surface and upper-altitude weather conditions on heavily polluted days, it is found that the surface pressure field and 850hPa temperature advection, 500hPa situation field have a good correspondence with the pollution day. Shallow troughs, weak northwest airflow or ridges, etc.), 850hPa is mostly controlled by warm ridges, and the ground is mostly in weak pressure gradient fields or low pressure convergence areas. The mid- and low-altitude wind field is dominated by southwest or southeast winds, the daily average relative humidity of the ground is high, and the average wind speed continues to be low. According to the surface pressure field situation, 500hPa situation field, and 850hPa cold and warm advection conditions, the analysis of the surface weather situation affecting Beijing is summarized into the following three weather types: high pressure (high pressure rear, high pressure bottom, weak high pressure); low pressure (closed low pressure , low pressure bottom rear, inverted trough); pressure equalization (equal pressure field, saddle field). These three types of weather pollution are all unfavorable diffusion types, the weather situation is relatively stable, and the occurrence time is relatively continuous. The following table counts the number of surface weather types during the heavy pollution days in Beijing in 2013. It can be seen that among the 58 heavy pollution days in 2013, the surface pressure situation on the heavy pollution days was divided into three types: high pressure, low pressure, and equal pressure. Each accounts for 38%, 41%, and 21%. Among the various surface weather conditions, the high pressure rear (14 days), equal pressure (12 days), low pressure bottom, and rear (12 days) are the main ones.

Table 1 Statistics of weather types on heavy pollution days in Beijing in 2013

Note: The high pressure rear weather type includes the low pressure front weather type

Forecasting is to deduce the future based on the present, so in the statistical forecasting method, the choice of predictors is to take the meteorological parameters at the starting time or the past time. However, some forecast quantities are most closely related to the meteorological conditions at the time of its occurrence. After the dynamic forecast is available, the forecast values of air pressure, temperature, humidity and three wind speed components u, v, w at various heights can be obtained from the dynamic forecast, and many other physical quantities (such as temperature advection) can be calculated from these basic physical quantities. , vorticity advection, water vapor transport flux, stability index, etc.).

The dynamic statistical model is used for prediction and forecasting. The dynamic statistical model assumes that the pollution level is mainly controlled by meteorological conditions, and the change of pollution sources is small; the meteorological factors with good stability, strong representativeness and good correlation with pollution are selected, and similar meteorological factors are analyzed at the same time. Combine to reduce the number of factors.

The following table shows the statistical characteristics of ground pollutants and meteorological monitoring data in Beijing in 2013. It can be seen that under different air quality levels, with the change of fine particle concentration, each meteorological element has obvious differences, especially the average wind speed, 24-hour variable pressure, temperature difference between 850hPa and 1000hPa at 08:00, 850hPa dew point temperature.

Table 2 Statistical characteristics of ground pollutants and meteorological monitoring data by level in Beijing in 2013

Table 3 Statistical characteristics of meteorological monitoring data by level in Beijing (08:00) Observatory

After screening and referring to the research results of other scholars, meteorological factors with good stability, strong representativeness, and good correlation with pollution were selected, including surface and high-altitude meteorological factors, and combined with each meteorological factor; finally, the average wind speed, 24-hour variable Pressure, 850 and 1000hP temperature difference meteorological variables are used as the discriminant indicators of heavy pollution days, which not only considers the ability of horizontal diffusion, but also shows the important role of vertical diffusion in the identification of heavy pollution, while reducing meteorological factors and increasing their correlation with pollutant concentrations relevance. Considering that only historical data can be used for the concentration of fine particles in the air quality forecast, the new variable yesterday’s national control fine particle concentration is added.

Use the screened physical quantity to establish the discrimination index of Beijing’s heavy pollution day. The data used after screening include 24-hour average wind speed, 24-hour variable pressure, temperature difference between 850 and 1000hPa at 08:00, and yesterday’s average concentration of nationally controlled fine particles, which are respectively set to x1 , x2, x3, x4, reflecting the influence of weather system evolution, cumulative concentration and other meteorological elements on heavy pollution. Based on matlab multi-parameter linear fitting, the prediction equation is obtained as follows:

Among them, c is the predicted concentration, a ₀ is a constant, and a1, a2, a3, a4 are regression coefficients.

Establish the statistical forecasting equation for 2013 as follows: Y (average concentration of predicted PM2.5) = 103.23-24.974x1-3.8127x2+1.5025x3+0.53945x4, wherein, Y is the average concentration of predicted PM2.5, and its definition is the same as c . Figure 1 shows the annual air quality forecast by the statistical forecast equation. It can be seen that the predicted value has a consistent trend of change compared with the measured value.

Validation of Discrimination Effects for Heavy Pollution Cases

Analyze the weather situation in the next few days. If it conforms to the characteristics of heavy pollution weather classification, use the established discriminant equation for forecasting and discrimination. parameters to verify the evaluation, using statistical parameters standardized mean deviation (NMB) and standardized mean error (NME), root mean square error (RMSE) to evaluate the degree of agreement between the simulation results and the measured values, as can be seen from Figure 2, most of the scattered points Concentrated between Y=2x and Y=0.5x, the predicted value has a larger correlation coefficient and a smaller NMB value than the measured value, and the average value of the predicted value and the measured value is relatively consistent. The statistical forecast established by statistics The discriminant rate of the equation for the 58 days of heavy pollution in a year is above 65%, and the result is good, showing the ability of the established discriminant equation to capture the process of heavy pollution.

Table 4 Statistical results of simulated and monitored values of heavy pollution for 58 days in 2013

Technology Roadmap

In actual operation, the forecast meteorological data is the simulation forecast result of WRF model, the initial and boundary data of WRF model are reanalysis daily data GFS of NCAR and NCEP, the resolution is 1°×1°, and the time resolution is 6h (00: 00, 06:00, 12:00, 18:00); the input data of terrain and underlying surface are from USGS 30s global terrain and MODIS underlying surface classification data respectively.

In order to verify the accuracy of the fitting results, the fine particulate matter and meteorological observation data from 2014 to 2015 were added to the database, and a new statistical relationship was updated and established. Using the new forecasting equation to verify and evaluate the PM2.5 concentration during the two red warnings in Beijing in December 2015, it can be seen from Figure 4(a) and 4(b) that the predicted value and the measured value have a good time The forecast equation can capture the heavy pollution process, and the results are relatively consistent, which can serve as an early warning.

The specific embodiments described above have further described the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above descriptions are only specific embodiments of the present invention, and are not intended to limit the present invention. Within the spirit and principles of the present invention, any modifications, equivalent replacements, improvements, etc., shall be included in the protection scope of the present invention.

Claims (6)

1. a method for forecasting heavy air pollution weather, is characterized in that, comprises the following steps: Step S1: Collect and obtain historical meteorological data of the area to be forecasted; Step S2: Select the following four meteorological parameters: 24-hour average wind speed, 24-hour variable pressure, the temperature difference between 850 and 1000hPa at 08:00, and the average concentration of fine particulate matter controlled by the state yesterday as the influencing factors; Step S3: By means of fitting, the air heavy pollution forecasting equation is established with the selected meteorological parameters as factors, and the predicted and/or measured values of the meteorological parameters are substituted into the forecasting equation. Whether a certain day to be predicted is predicted as heavily polluted weather; Wherein, the described prediction equation of fitting is: Among them, c is the predicted daily average concentration of PM _2.5 on a certain day to be predicted, a ₀ is a constant, a ₁ , a ₂ , a ₃ , and a ₄ are regression coefficients; x ₁ , x ₂ , and x ₃ are respectively The predicted value of the 24-hour average wind speed, 24-hour variable pressure, and the temperature difference between 850 and 1000hPa at 08:00 on a certain day, and x ₄ is the measured or predicted value of yesterday’s nationally controlled fine particulate matter concentration on a certain day to be predicted. 2. The forecasting method of heavy air pollution weather as claimed in claim 1, is characterized in that, described a certain day to be predicted can be selected as today, tomorrow or the day after tomorrow relative to the current day, respectively corresponding to the next 24 hours, Forecast of heavily polluted weather within 24-48 hours or within 48-72 hours. 3. The forecasting method of heavy air pollution weather as claimed in claim 1, is characterized in that, the numerical value of x ₁ , x ₂ , x ₃ is selected from the simulated forecast result of WRF model, and the initial and boundary data of WRF model are NCAR and NCEP The reanalysis of daily data GFS, the resolution is 1°×1°, the time resolution is 6h, and the corresponding time is 00:00, 06:00, 12:00, 18:00; the input data of topography and underlying surface come from USGS 30s global terrain and MODIS underlying surface classification data. 4. the method for forecasting heavy air pollution weather as claimed in claim 1 is characterized in that, for x ₄ , when predicting today's heavy pollution weather situation, select the measured value of yesterday's national control fine particle mean concentration; when predicting tomorrow In case of heavy pollution weather or the day after tomorrow, use the predicted value of the national control fine particle mean concentration yesterday or the day after tomorrow. 5. The method for forecasting heavy air pollution weather as claimed in claim 1, characterized in that the forecast equation for 2013 fitted to the urban area of Beijing is: c=103.23-24.974x ₁ -3.8127x ₂ +1.5025x ₃ + _0.53945x4 . 6. A forecasting system for heavy air pollution weather, characterized in that, the forecasting system executes the forecasting method for heavy air pollution weather as claimed in any one of claims 1 to 5 based on matlab software, to predict future Forecast the heavily polluted weather on a certain day.