Electric quantity prediction method based on meteorological factors

Abstract

The invention discloses an electric quantity prediction method based on meteorological factors, which belongs to the technical field of electric quantity prediction, and comprises the following steps of: separating original electric quantity data, and analyzing a coupling effect of meteorological electric quantity in the separated electric quantity data to obtain a meteorological comprehensive index by considering that the influence degree of the meteorological factors on electric quantity prediction is relatively large; a meteorological data correction algorithm is adopted to correct the obtained data to reduce prediction errors, a meteorological electric quantity prediction model is constructed according to a model averaging method to predict the electric quantity, the model averaging method is applied to meteorological electric quantity prediction, the influence of the combination of various influence factors on electric quantity prediction is fully considered, and the prediction accuracy is improved. Therefore, the reference data during electric quantity prediction is more perfect, and the accuracy of electric quantity prediction is remarkably improved on the premise that the prediction efficiency is not affected.

Description

Technical Field

[0001] The present invention relates to the technical field of electricity forecasting, and in particular to an electricity forecasting method based on meteorological factors. Background Art

[0002] With the deepening of the reform of the power economic system and the continuous advancement of the construction of energy Internet, the method of using a single model for electricity forecasting can no longer meet the forecasting needs. The electricity forecasting of a single model is easily affected by the training samples and model parameters, which affects the prediction accuracy and the reliability is not high. In order to improve the prediction accuracy and prediction reliability, a common method is to use multiple models for joint prediction. However, due to the different training objects, model structures and output formats among different models, it is necessary to unify the formats and process the data when conducting joint prediction, which increases the amount of calculation, makes the prediction steps more complicated, and affects the prediction efficiency. Therefore, how to solve the problem of low prediction accuracy caused by the single model's easy external influence on electricity forecasting while ensuring prediction efficiency needs to be solved urgently.

[0003] Chinese patent, publication number: CN116307161A, publication date: June 23, 2023, discloses a method for predicting electricity consumption, which includes: preprocessing the acquired meteorological data and electricity consumption data; constructing an electricity consumption prediction model based on the preprocessed meteorological data and electricity consumption data using the PSO-AWRF algorithm, and using Spark's parallel data processing capabilities to reduce the data training optimization time in the data preprocessing and modeling process, thereby improving the prediction speed; predicting electricity consumption through the trained electricity consumption prediction model, but using a single model to predict electricity, and the prediction accuracy is not high. Summary of the invention

[0004] The present invention aims at the problem that the prior art adopts a single model to predict electricity quantity, which is easily affected by external conditions and the influencing factors are not fully considered, resulting in low accuracy of electricity prediction. The present invention provides an electricity prediction method based on meteorological factors, by separating the original electricity data, and considering that meteorological factors have a greater impact on electricity prediction, the coupling effect of meteorological electricity in the separated electricity data is analyzed to obtain a meteorological comprehensive index, and a meteorological data correction algorithm is used to correct the obtained data to reduce the prediction error, and then a meteorological electricity prediction model is constructed according to the model averaging method to predict electricity. The model averaging method is applied to meteorological electricity prediction, and the influence of the combination of multiple influencing factors on electricity prediction is fully considered, so that the reference data for electricity prediction is more complete, and the accuracy of electricity prediction is significantly improved without affecting the prediction efficiency.

[0005] In a first aspect, a technical solution provided in an embodiment of the present invention is: a method for predicting electricity consumption based on meteorological factors, comprising the following steps: S1. Separating the original electricity data based on the fluctuation of meteorological data to obtain separated electricity; S2. Analyze the coupling effect and cumulative effect of meteorological factors on the meteorological electricity in the separated electricity to obtain a comprehensive meteorological index; S3, analyzing and correcting the meteorological data and the meteorological comprehensive index based on the set meteorological data correction algorithm to obtain meteorological correction data; S4. Construct a meteorological power forecast model based on the model averaging method, and use the meteorological correction data as input to output the predicted power.

[0006] In this scheme, by decomposing the original electricity data, the meteorological electricity is separated from the original electricity, so that the electricity affected by the weather can be analyzed intuitively. The use of meteorological data for electricity forecasting can improve the accuracy of the forecast and enhance the ability to analyze and forecast electricity in response to extreme weather and natural disasters. By analyzing the coupling effect and cumulative effect of meteorological factors on the meteorological electricity in the separated electricity, the meteorological electricity can be analyzed from multiple dimensions, and the meteorological factors affecting the change of electricity can be extracted, so that the electricity can be predicted according to the influence of meteorological factors on electricity in the future, which reduces the amount of prediction calculation and improves the prediction speed and accuracy. The meteorological data and the meteorological comprehensive index are analyzed and corrected by the meteorological data correction algorithm, which can not only effectively quantify the uncertainty contained in the meteorological data and its impact on the electricity forecast model, but also realize the fine adjustment of the electricity forecast model to improve the prediction accuracy, reduce the prediction error, and enhance the applicability of the model in different seasons and different weather conditions. Based on the model averaging method, a meteorological electricity combination prediction model is constructed to realize the accurate prediction of future meteorological electricity, realize the lean management of power companies, improve the power supply quality and service level, reduce operating costs, and increase the stability and reliability of the power system.

[0007] Preferably, in S1, separating the original power data based on the fluctuation of meteorological data to obtain separated power includes the following steps: Obtain the original electricity data, and divide the electricity sequence in the original electricity data into a short-term fluctuation sequence and a long-term trend sequence based on the fluctuation of meteorological data; Based on the data characteristics of short-term fluctuation series and long-term trend series, the original electricity data is separated to obtain meteorological electricity, trend electricity and random electricity, and the above electricity is recorded to obtain separated electricity.

[0008] In this scheme, since electricity consumption is affected by scientific and technological progress and social and economic development, as well as by various factors such as meteorology, as meteorological data fluctuates, part of the electricity consumption is affected, and the affected electricity is meteorological electricity. The original electricity data can be decomposed into trend electricity, meteorological electricity and random electricity by using the HP filtering method, and then the meteorological data can be used for electricity forecasting, which can improve the accuracy of the forecast and enhance the electricity analysis and forecasting capabilities in response to extreme weather and natural disasters.

[0009] Preferably, in S2, the coupling effect and cumulative effect of meteorological factors on the meteorological electricity in the separated electricity are analyzed to obtain a meteorological comprehensive index, including the following steps: Based on the influence of humidity and temperature on electricity, a temperature and humidity index is constructed. Based on the temperature and humidity index, the coupling effect of meteorological electricity in the separated electricity is analyzed to obtain the coupling factor of multiple meteorological factors. Based on the multi-day cumulative effect of meteorological factors on electricity consumption, the cumulative effect analysis of the meteorological electricity in the separated electricity is carried out to obtain the cumulative effect factor; the coupling factors of multiple meteorological factors and the cumulative effect factors are statistically analyzed to obtain the meteorological comprehensive index.

[0010] In this scheme, since the fluctuation of electricity consumption is affected by meteorological factors, and the impact of meteorological factors on electricity consumption fluctuations includes coupling effect and cumulative effect, among which the coupling effect of humidity and temperature will have a great impact on electricity consumption. Therefore, a temperature and humidity index is constructed to reflect the coupling effect of humidity and temperature on electricity consumption. In the cumulative effect, it is necessary to consider the multi-day cumulative effect of meteorological factors on electricity consumption. For example, for the electricity consumption on a certain day, the electricity consumption on that day may be different in the two cases of high temperature for three consecutive days and sudden high temperature on the same day. Therefore, in order to improve the prediction accuracy, it is necessary to analyze the coupling effect and cumulative effect of the meteorological electricity consumption affected by meteorological factors.

[0011] Preferably, the meteorological data correction algorithm analyzes meteorological data and meteorological comprehensive index based on Fisher information processing method, comprising the following steps: Separate the meteorological factors in the meteorological data to obtain single meteorological factors, and divide the meteorological comprehensive index corresponding to each single meteorological factor into a data window according to the sampling order; The meteorological comprehensive indexes corresponding to different single meteorological factors in the data window are analyzed, the types of single meteorological factors are boxed based on the random boxing method and the corresponding probability density functions are set, and the Fisher information value of each single meteorological factor is obtained based on the probability density function.

[0012] In this scheme, since the meteorological data used in predicting future meteorological electricity is the future meteorological forecast data, there are errors between the meteorological forecast data and the actual meteorological data. In order to reduce the impact of these errors on the predicted meteorological electricity, the Fisher information processing method is introduced as the core to construct a meteorological data correction algorithm. Through detailed analysis of single meteorological factors such as temperature, humidity, wind speed, etc., and processing of complex meteorological comprehensive indexes such as temperature and humidity index, lagged meteorological factors, multi-meteorological factor coupling factors and cumulative effect factors, etc., and constructing corresponding probability density functions for different meteorological factors, the Fisher information value is calculated based on the probability density function, which can effectively quantify the uncertainty contained in the meteorological data and its impact on the electricity prediction model, reduce the difficulty of prediction, and improve the prediction accuracy without reducing the prediction efficiency.

[0013] Preferably, in S3, the meteorological data and the meteorological comprehensive index are corrected based on a set meteorological data correction algorithm to obtain meteorological correction data, including the following steps: Extract the meteorological forecast data from the meteorological data to obtain the forecast time of the meteorological forecast data, and interpolate the actual meteorological data at several sampling times before the forecast time to obtain a discrete meteorological data set; Based on the meteorological data correction algorithm, the Fisher information value of the discrete meteorological data set is calculated and normalized to obtain meteorological quantitative data. Based on the meteorological quantitative data, the meteorological comprehensive index and meteorological forecast data, a meteorological forecast correction formula is constructed, and meteorological correction data is obtained based on the meteorological forecast correction formula.

[0014] In this scheme, in order to correct the meteorological forecast data at a certain moment, a discrete meteorological data set is obtained by interpolating the actual meteorological data at several sampling moments before the forecast moment. In order to measure the stability of the discrete meteorological data set, a meteorological data correction algorithm is used to calculate the Fisher information value of the discrete meteorological data set, and the value range of the meteorological quantitative data is obtained by normalizing it. Then, considering the factors that are most easily affected by the meteorological forecast data, the meteorological quantitative data is mapped to the interval with the greatest impact on the corresponding meteorological forecast data to obtain the meteorological forecast correction formula, and the meteorological forecast data is corrected. Through the optimization measures of the Fisher information processing method, the power forecast model is fine-tuned, the forecast accuracy is improved, the forecast error is reduced, and the applicability of the model to different seasons and different weather conditions is enhanced.

[0015] Preferably, in S4, a meteorological power forecasting model is constructed based on a model averaging method, and the meteorological correction data is used as an input value to output the predicted power, including the following steps: The meteorological factors in the meteorological correction data are screened for their characteristics to obtain meteorological factors, and the important meteorological factors are screened based on the degree of influence of meteorological factors on electricity. For each important meteorological factor, a corresponding sub-combination model is constructed based on different variable combinations, and the weight of each sub-combination model is calculated using the corresponding weight calculation criteria based on the variable combination of each sub-combination model; The variable combination corresponding to the sub-combination model corresponding to the maximum weight value of each sub-combination model is selected as the prediction influencing variable to predict the power consumption, and the error between the power prediction value and the actual power value is fitted to optimize the power prediction value to obtain the predicted power consumption.

[0016] In this scheme, by introducing machine learning models such as LGBM to screen meteorological factors, we can screen out the features with high importance to improve the prediction efficiency. For the n features after screening, we use the method of permutation and combination to construct 2 n -1 sub-combination model, so that each feature combination is included in the model set, not only comprehensively covers the combined impact of various meteorological factors on electricity, but also because the difference between the sub-combination models lies in the different combinations of variables, the model structure and output format of the sub-combination models are basically the same. There is no need to unify the format and process the data when making predictions, which will not increase the amount of calculation and make the prediction steps more complicated. Therefore, it will not affect the prediction efficiency. The prediction accuracy is significantly improved while ensuring that the prediction efficiency is not reduced; the weights of each sub-model are calculated through the AIC criterion, BIC criterion, Jackknife criterion, equal weight method, LASSO method, etc., and the optimal model is selected. Finally, machine learning neural network models, such as convolutional neural network (CNN) and long short-term memory network (LSTM), are introduced to correct the prediction error, optimize the prediction results, and improve the accuracy of electricity prediction.

[0017] Preferably, screening various meteorological factors based on the degree of influence of meteorological factors on electricity to obtain important meteorological factors includes the following steps: A meteorological factor evaluation index system is constructed based on the impact of meteorological factors on electricity volume. Each meteorological factor is evaluated based on the meteorological factor evaluation index system to obtain a meteorological impact score. The meteorological factors corresponding to the meteorological impact scores exceeding the set score threshold are screened to obtain important meteorological factors.

[0018] In this scheme, since there are many types of meteorological factors, different meteorological factors have different degrees of influence on electricity consumption. The impact of some meteorological factors in electricity consumption forecasting is almost negligible. Therefore, directly ignoring these meteorological factors with less influence in electricity consumption forecasting can greatly reduce the calculation steps in the forecasting process, thereby improving the forecasting efficiency without affecting the forecasting accuracy.

[0019] Preferably, the sub-combination model automatically selects the most suitable weight calculation criterion from the database based on the number of combinations to calculate the weight of each sub-combination model, and the weight calculation criterion at least includes AIC criterion, BIC criterion, Jackknife criterion and LASSO method.

[0020] In this scheme, since different sub-combination models have different combination variables, the number of different combination variables is calculated differently when calculating the weight of the sub-combination model, and the weight calculation steps of different weight calculation criteria are different. The weight calculation steps of the sub-combination model with simpler combination variables are relatively simple, and the weight calculation steps of the sub-combination model with more complex combination variables are relatively complex. Therefore, the calculation time is longer. Selecting the most suitable weight calculation criterion from the database to calculate the weight of each sub-combination model can effectively reduce the calculation time, thereby indirectly improving the prediction efficiency of electricity forecasting.

[0021] Preferably, the meteorological data at least includes meteorological forecast data, meteorological factors and actual meteorological data.

[0022] In this scheme, when predicting electricity, part of the electricity is affected by meteorological changes, while other electricity is not greatly affected. Therefore, the focus is on analyzing meteorological factors to obtain the relationship between electricity changes and meteorological changes. Therefore, it is necessary to analyze meteorological forecast data, types of meteorological factors and actual meteorological data, and fully consider the corresponding changes in various data when the meteorological changes occur, so as to conduct a comprehensive analysis of the meteorology to improve the accuracy of electricity prediction based on meteorological factors.

[0023] Preferably, the meteorological factors include at least temperature, humidity, wind speed and rainfall.

[0024] In this scheme, common meteorological factors include temperature, humidity, wind speed and rainfall. These types have the greatest impact on meteorological power, while other meteorological factors have a smaller impact on power. Therefore, the above-mentioned meteorological factors are mainly analyzed and discussed to predict meteorological power. The prediction steps are simplified as much as possible while ensuring the reliability of the prediction, thereby indirectly improving the prediction efficiency.

[0025] The beneficial effects of the present invention are as follows: (1) The present invention decomposes the original power data and separates the meteorological power from the original power, so that the power affected by the weather can be analyzed intuitively. The use of meteorological data for power forecasting can improve the accuracy of the forecast and enhance the power analysis and forecasting capabilities in response to extreme weather and natural disasters; (2) The present invention analyzes the coupling effect and cumulative effect of meteorological factors on the meteorological electricity in the separated electricity, and can analyze the meteorological electricity from multiple dimensions and extract the meteorological factors that affect the change of electricity, so as to facilitate the subsequent prediction of electricity according to the influence of meteorological factors on electricity, thereby reducing the amount of prediction calculation and improving the prediction calculation speed and accuracy; (3) The present invention analyzes and corrects meteorological data and meteorological comprehensive indexes through a meteorological data correction algorithm, which can not only effectively quantify the uncertainty contained in meteorological data and its impact on the power forecast model, but also realize the fine adjustment of the power forecast model to improve the prediction accuracy, reduce the prediction error, and enhance the applicability of the model in different seasons and different weather conditions; (4) The present invention constructs a meteorological power combination prediction model based on the model averaging method, which improves the prediction accuracy without reducing the prediction efficiency, realizes the accurate prediction of future meteorological power, realizes the lean management of power companies, improves the power supply quality and service level, reduces operating costs, and increases the stability and reliability of the power system.

[0026] The above invention content is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Other features, objects and advantages of the present invention will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings. The drawings are only for the purpose of illustrating preferred embodiments and are not to be considered as limiting the present invention. Also, the same reference symbols are used throughout the drawings to represent the same parts.

[0028] Figure 1 The present invention is a flow chart of a method for predicting electricity quantity based on meteorological factors. DETAILED DESCRIPTION

[0029] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific implementation method described herein is only an optimal embodiment of the present invention, which is only used to explain the present invention and does not limit the scope of protection of the present invention. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0030] Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flow charts. Although the flow charts describe the operations (or steps) as sequential processes, many of the operations (or steps) therein can be implemented in parallel, concurrently, or simultaneously. In addition, the order of the operations can be rearranged. The process can be terminated when its operation is completed, but can also have additional steps not included in the drawings; the process can correspond to a method, function, procedure, subroutine, subprogram, etc.

[0031] Example: Figure 1 As shown, in order to solve the problem that the existing technology uses a single model to predict electricity, which is easily affected by external conditions and the influencing factors are not fully considered, resulting in low accuracy of electricity prediction, this embodiment provides an electricity prediction method based on meteorological factors, including the following steps: S1: Separate the original electricity data based on the fluctuation of meteorological data to obtain separated electricity.

[0032] In this embodiment, the original power data is separated based on the fluctuation of meteorological data to obtain separated power, including the following steps: Obtain the original electricity data, and divide the electricity sequence in the original electricity data into a short-term fluctuation sequence and a long-term trend sequence based on the fluctuation of meteorological data; Based on the data characteristics of short-term fluctuation series and long-term trend series, the original electricity data is separated to obtain meteorological electricity, trend electricity and random electricity, and the above electricity is recorded to obtain separated electricity.

[0033] In this embodiment, since electricity consumption is affected by scientific and technological progress and social and economic development, as well as by multiple factors such as weather, as meteorological data fluctuates, part of the electricity consumption is affected accordingly, and the affected part of the electricity is meteorological electricity. The HP filtering method can be used to decompose the original electricity data into trend electricity, meteorological electricity and random electricity, and then the meteorological data can be used to predict electricity, which can improve the accuracy of the prediction and enhance the electricity analysis and prediction capabilities in response to extreme weather and natural disasters.

[0034] In this embodiment, the HP filter method can separate the long-term trend from the short-term fluctuation in the data sequence. Its essence is a nonlinear regression technology method. The principle is similar to that of a high-pass filter. By performing spectral analysis on the time series, it can more intuitively display the different measurements in the separated period. The HP filter method can have a better fit in small data samples. Assume that the original power sequence is Y, and the meteorological power has random fluctuations and can be regarded as a short-term fluctuation sequence, denoted by C = {c 1 ,c 2 ,...,c n}; The trend quantity is relatively stable and can be regarded as a long-term trend sequence, denoted by G = {g 1 ,g 2 ,...,g n}, when separating meteorological power and trend power, the function minimum principle should be followed: in, and are the standard deviations of the long-term trend series and the short-term fluctuation series, respectively.

[0035] S2: Analyze the coupling effect and cumulative effect of meteorological factors on the meteorological electricity in the separated electricity to obtain the meteorological comprehensive index.

[0036] In this embodiment, the coupling effect and cumulative effect of meteorological factors on the meteorological electricity in the separated electricity are analyzed to obtain a meteorological comprehensive index, including the following steps: Based on the influence of humidity and temperature on electricity, a temperature and humidity index is constructed. Based on the temperature and humidity index, the coupling effect of meteorological electricity in the separated electricity is analyzed to obtain the coupling factor of multiple meteorological factors. Based on the multi-day cumulative effect of meteorological factors on electricity consumption, the cumulative effect analysis of the meteorological electricity in the separated electricity is carried out to obtain the cumulative effect factor; the coupling factors of multiple meteorological factors and the cumulative effect factors are statistically analyzed to obtain the meteorological comprehensive index.

[0037] In this embodiment, since the fluctuation of electricity is affected by meteorological factors, and the influence of meteorological factors on the fluctuation of electricity includes coupling effect and cumulative effect, among which the coupling effect of humidity and temperature will have a great influence on the electricity, a temperature and humidity index is constructed to reflect the coupling effect of humidity and temperature on electricity, and in the cumulative effect, it is necessary to consider the multi-day cumulative effect of meteorological factors on electricity consumption. For example, for the electricity consumption on a certain day, the electricity consumption on that day may be different in the two cases of high temperature for three consecutive days and sudden high temperature on the same day. Therefore, in order to improve the prediction accuracy, it is necessary to analyze the coupling effect and cumulative effect of the meteorological electricity being affected by meteorological factors.

[0038] The specific steps of analyzing the coupling effect of meteorological electricity in this embodiment are as follows: considering that the coupling effect of humidity and temperature will have a great impact on electricity, a temperature and humidity index is constructed to reflect the coupling effect of humidity and temperature on electricity. The formula is defined as follows: Among them, Temp C is Celsius temperature, which is the daily average temperature; Hmd is the humidity percentage.

[0039] The specific steps of this embodiment in analyzing the cumulative effect of meteorological electricity are as follows: Consider the cumulative effect of meteorological factors on electricity consumption over multiple days. For example, for a certain day's electricity consumption, the daily electricity consumption may be different in the case of three consecutive days of high temperature and the case of sudden high temperature on the same day. In order to better quantify the cumulative effect of meteorological factors, extract the three-period lagged meteorological factor data, and the specific formula is as follows: X′=X t-1 X″=X t-2 X″′=X t-3 Where X = (x 1 ,x 2 ,...,x n ) represents various meteorological factors, X t-1 Indicates the meteorological factor data of the previous period, X t-2 represents the meteorological factor data of the last two periods, X t-3 Represents the meteorological factor data of the last three periods.

[0040] S3: Analyze and correct the meteorological data and meteorological comprehensive index based on the set meteorological data correction algorithm to obtain meteorological correction data.

[0041] In this embodiment, the meteorological data correction algorithm analyzes meteorological data and meteorological comprehensive index based on Fisher information processing method, including the following steps: Separate the meteorological factors in the meteorological data to obtain single meteorological factors, and divide the meteorological comprehensive index corresponding to each single meteorological factor into a data window according to the sampling order; The meteorological comprehensive indexes corresponding to different single meteorological factors in the data window are analyzed, the types of single meteorological factors are boxed based on the random boxing method and the corresponding probability density functions are set, and the Fisher information value of each single meteorological factor is obtained based on the probability density function.

[0042] In this embodiment, when predicting future meteorological electricity, the meteorological data used is the future meteorological forecast data. However, there are errors between the meteorological forecast data and the actual meteorological data. In order to reduce the impact of these errors on the predicted meteorological electricity, the Fisher information processing method is introduced as the core to construct a meteorological data correction algorithm. Through detailed analysis of single meteorological factors such as temperature, humidity, wind speed, etc., and processing of complex meteorological comprehensive indexes such as temperature and humidity index, lagged meteorological factors, multi-meteorological factor coupling factors and cumulative effect factors, corresponding probability density functions are constructed for different meteorological factors. The Fisher information value is calculated based on the probability density function, which can effectively quantify the uncertainty contained in the meteorological data and its impact on the electricity prediction model, reduce the difficulty of prediction, and improve the prediction accuracy without reducing the prediction efficiency.

[0043] In this embodiment, the meteorological data and the meteorological comprehensive index are corrected based on the set meteorological data correction algorithm to obtain meteorological correction data, including the following steps: Extract the meteorological forecast data from the meteorological data to obtain the forecast time of the meteorological forecast data, and interpolate the actual meteorological data at several sampling times before the forecast time to obtain a discrete meteorological data set; Based on the meteorological data correction algorithm, the Fisher information value of the discrete meteorological data set is calculated and normalized to obtain meteorological quantitative data. Based on the meteorological quantitative data, the meteorological comprehensive index and meteorological forecast data, a meteorological forecast correction formula is constructed, and meteorological correction data is obtained based on the meteorological forecast correction formula.

[0044] This embodiment takes the correction of the predicted temperature in the weather forecast data as an example to illustrate the correction of the weather forecast data in this embodiment, and the specific process is as follows: If the temperature value T in the weather forecast for one hour later is 1 For correction, the real temperature values ​​of the previous 12 hours are selected. Since the meteorological data obtained are at intervals of one hour, the selected temperature data are 12 discrete data points. In order to increase the number of discrete data, the 12 temperature data are interpolated to obtain 48 discrete temperature data points at intervals of 15 minutes. The temperature discrete data set is T 0 ={T 0 (i)|i=1,2,...,48}, the time series length is 48.

[0045] To measure Y 0 The stability of this data set is calculated by calculating the Fisher information value I of the discrete data set. FI (i) I FI The value of (i) is between 0 and 4, and I FI (i) Perform normalization processing, and the formula is as follows: in, The value range is between [0,1].

[0046] right The processing formula is as follows: in, The value of is also between [0,1].

[0047] Since the temperature value at a certain time is most susceptible to the temperature change in the past hour, The value range of is mapped to the interval [0,1.5], and the following weather forecast temperature correction formula is obtained: Among them, T 1 is the weather forecast temperature value at the time to be predicted, T 2 is the temperature correction value of weather forecast, T 0 (44) is the temperature value 2 hours before the temperature point to be calibrated, T 0 (48) is the temperature value 1 hour before the temperature point to be calibrated.

[0048] In order to correct the meteorological forecast data at a certain moment, this embodiment obtains a discrete meteorological data set by interpolating the actual meteorological data at several sampling moments before the forecast moment, and in order to measure the stability of the discrete meteorological data set, a meteorological data correction algorithm is used to calculate the Fisher information value of the discrete meteorological data set, and normalization is performed on it to obtain the value range of meteorological quantitative data, and then the factors that are most likely to be affected by the meteorological forecast data are considered, and the meteorological quantitative data are mapped to the interval with the greatest impact on the corresponding meteorological forecast data to obtain the meteorological forecast correction formula, and the meteorological forecast data are corrected. Through the optimization measures of the Fisher information processing method, the power forecast model is finely adjusted, the forecast accuracy is improved, the forecast error is reduced, and the applicability of the model to different seasons and different weather conditions is enhanced.

[0049] S4: Construct a meteorological power forecast model based on the model averaging method, and use the meteorological correction data as input to output the predicted power.

[0050] In this embodiment, a meteorological power forecasting model is constructed based on a model averaging method, and the meteorological correction data is used as an input value to output the predicted power, including the following steps: The meteorological factors in the meteorological correction data are screened for their characteristics to obtain meteorological factors, and the important meteorological factors are screened based on the degree of influence of meteorological factors on electricity. For each important meteorological factor, a corresponding sub-combination model is constructed based on different variable combinations, and the weight of each sub-combination model is calculated using the corresponding weight calculation criteria based on the variable combination of each sub-combination model; The variable combination corresponding to the sub-combination model corresponding to the maximum weight value of each sub-combination model is selected as the prediction influencing variable to predict the power consumption, and the error between the power prediction value and the actual power value is fitted to optimize the power prediction value to obtain the predicted power consumption.

[0051] The specific implementation process of constructing the meteorological power forecast model in this embodiment is as follows: use the LGBM algorithm to calculate the importance score of each meteorological factor on the meteorological power, and select the features with higher scores; secondly, based on the selected features, use the permutation and combination method to build a combination model; then, select the weight to calculate the predicted value; finally, optimize the prediction result through the LSTM model. The specific process is as follows: Step 1: Feature screening: By using the LGBM algorithm to calculate the importance scores of the impact of features such as temperature, humidity, wind speed, rainfall, temperature and humidity index, and lagged meteorological factors on meteorological electricity, the meteorological factors with higher scores are screened out.

[0052] Step 2: Combination model construction: First, for the n features after screening, use the permutation and combination method to construct 2 n -1 sub-model. If there are three features after screening: x 1 , x 2 and x 3 , then there are 2 in the model set 3 -1, that is, the independent variable combinations of the sub-models are (x 1 )、(x 2 )、(x 3 )、(x 1 ,x 2 )、(x 1 ,x 3 )、(x 2 ,x 3 ) and (x 1 ,x 2 ,x 3 ). For each combination of variables, the most appropriate model can be selected. Secondly, the weights of each sub-model are calculated by AIC criterion, BIC criterion, Jackknife criterion, equal weight method, LASSO method, etc., and higher weights are given to more preferred sub-models. This method takes into account all candidate models and can minimize the prediction error.

[0053] Step 3: Model optimization: Assume that y is the true value of meteorological electricity. is the meteorological power forecast value of the model averaging method, then the forecast error is: Introduce the LSTM model to fit ε to obtain the corrected value of ε The optimization result of meteorological power forecast is:

[0054] This embodiment introduces machine learning models such as LGBM to screen meteorological factors and select features with high importance to improve prediction efficiency. For the n features after screening, 2 n-1 sub-combination model, so that each feature combination is included in the model set, not only comprehensively covers the combined impact of various meteorological factors on electricity, but also because the difference between the sub-combination models lies in the different combinations of variables, the model structure and output format of the sub-combination models are basically the same. There is no need to unify the format and process the data when making predictions, which will not increase the amount of calculation and make the prediction steps more complicated. Therefore, it will not affect the prediction efficiency. The prediction accuracy is significantly improved while ensuring that the prediction efficiency is not reduced; the weights of each sub-model are calculated through the AIC criterion, BIC criterion, Jackknife criterion, equal weight method, LASSO method, etc., and the optimal model is selected. Finally, machine learning neural network models, such as convolutional neural network (CNN) and long short-term memory network (LSTM), are introduced to correct the prediction error, optimize the prediction results, and improve the accuracy of electricity prediction.

[0055] In this embodiment, the various meteorological factors are screened based on the degree of influence of the meteorological factors on the power to obtain important meteorological factors, including the following steps: A meteorological factor evaluation index system is constructed based on the impact of meteorological factors on electricity volume. Each meteorological factor is evaluated based on the meteorological factor evaluation index system to obtain a meteorological impact score. The meteorological factors corresponding to the meteorological impact scores exceeding the set score threshold are screened to obtain important meteorological factors.

[0056] In this embodiment, since there are many types of meteorological factors, different meteorological factors have different degrees of influence on electricity consumption. The influence of some meteorological factors on electricity consumption forecasting is almost negligible. Therefore, these meteorological factors with less influence can be directly ignored during electricity consumption forecasting, which can greatly reduce the calculation steps in the forecasting process, thereby improving the forecasting efficiency without affecting the forecasting accuracy.

[0057] In this embodiment, the sub-combination model automatically selects the most suitable weight calculation criterion from the database based on the number of combinations to calculate the weight of each sub-combination model, and the weight calculation criterion at least includes the AIC criterion, the BIC criterion, the Jackknife criterion and the LASSO method.

[0058] In this embodiment, since different sub-combination models have different combination variables, the number of different combination variables is calculated differently when calculating the weight of the sub-combination model, and the weight calculation steps of different weight calculation criteria are different, the weight calculation steps of the sub-combination model with simpler combination variables are relatively simple, and the weight calculation steps of the sub-combination model with more complex combination variables are relatively complex, so the calculation time is longer. Selecting the most suitable weight calculation criterion from the database to calculate the weight of each sub-combination model can effectively reduce the calculation time, thereby indirectly improving the prediction efficiency of electricity forecasting.

[0059] In this embodiment, the meteorological data at least includes meteorological forecast data, meteorological factors and actual meteorological data.

[0060] In this embodiment, when predicting electricity, part of the electricity is affected by meteorological changes, while other electricity is not greatly affected. Therefore, the focus is on analyzing meteorological factors to obtain the relationship between electricity changes and meteorological changes. Therefore, it is necessary to analyze meteorological forecast data, types of meteorological factors and actual meteorological data, and fully consider the corresponding changes in various data when the meteorological changes occur, so as to conduct a comprehensive analysis of the meteorology to improve the accuracy of predicting electricity based on meteorological factors.

[0061] In this embodiment, the meteorological factors include at least temperature, humidity, wind speed and rainfall.

[0062] Common meteorological factors in this embodiment include temperature, humidity, wind speed and rainfall. These factors have the greatest impact on meteorological power, while other meteorological factors have a smaller impact on power. Therefore, the above-mentioned meteorological factors are mainly analyzed and discussed to predict meteorological power. The prediction steps are simplified as much as possible while ensuring the reliability of the prediction, thereby indirectly improving the prediction efficiency.

[0063] This embodiment decomposes the original electricity data and separates the meteorological electricity from the original electricity, so that it can intuitively analyze the part of the electricity affected by the weather. Using meteorological data to predict electricity can improve the accuracy of the prediction and enhance the ability to analyze and predict electricity in response to extreme weather and natural disasters. By analyzing the coupling effect and cumulative effect of meteorological factors on the meteorological electricity in the separated electricity, the meteorological electricity can be analyzed from multiple dimensions, and the meteorological factors affecting the change of electricity can be extracted, so that the electricity can be predicted according to the influence of meteorological factors on the electricity in the future, which reduces the amount of prediction calculation and improves the prediction calculation speed and accuracy. The meteorological data and the meteorological comprehensive index are analyzed and corrected by the meteorological data correction algorithm, which can not only effectively quantify the uncertainty contained in the meteorological data and its influence on the electricity prediction model, but also realize the fine adjustment of the electricity prediction model to improve the prediction accuracy, reduce the prediction error, and enhance the applicability of the model in different seasons and different weather conditions. Based on the model averaging method, a meteorological electricity combination prediction model is constructed to realize the accurate prediction of future meteorological electricity, realize the lean management of power companies, improve the power supply quality and service level, reduce operating costs, and increase the stability and reliability of the power system.

[0064] It can be seen from the above embodiments that at least the following substantial effects are achieved: (1) The present invention decomposes the original power data and separates the meteorological power from the original power, so that the power affected by the weather can be analyzed intuitively. The use of meteorological data for power forecasting can improve the accuracy of the forecast and enhance the power analysis and forecasting capabilities in response to extreme weather and natural disasters. (2) The present invention analyzes the coupling effect and cumulative effect of meteorological factors on the meteorological electricity in the separated electricity, and can analyze the meteorological electricity from multiple dimensions and extract the meteorological factors that affect the change of electricity, so as to facilitate the subsequent prediction of electricity according to the influence of meteorological factors on electricity, thereby reducing the amount of prediction calculation and improving the prediction calculation speed and accuracy; (3) The present invention analyzes and corrects meteorological data and meteorological comprehensive indexes through a meteorological data correction algorithm, which can not only effectively quantify the uncertainty contained in meteorological data and its impact on the power forecast model, but also realize the fine adjustment of the power forecast model to improve the prediction accuracy, reduce the prediction error, and enhance the applicability of the model in different seasons and different weather conditions; (4) The present invention constructs a meteorological power combination prediction model based on the model averaging method, which improves the prediction accuracy without reducing the prediction efficiency, realizes the accurate prediction of future meteorological power, realizes the lean management of power companies, improves the power supply quality and service level, reduces operating costs, and increases the stability and reliability of the power system.

[0065] The specific implementation described above is a preferred implementation of a method for electricity forecasting based on meteorological factors of the present invention, and is not intended to limit the specific implementation scope of the present invention. The scope of the present invention includes but is not limited to this specific implementation. All equivalent changes made in accordance with the shape and structure of the present invention are within the protection scope of the present invention.

Claims

1. A method for predicting electricity consumption based on meteorological factors, characterized in that: The following steps are involved: S1. Separating the original electricity data based on the fluctuation of meteorological data to obtain separated electricity; S2. Analyze the coupling effect and cumulative effect of meteorological factors on the meteorological electricity in the separated electricity to obtain a comprehensive meteorological index; S3, analyzing and correcting the meteorological data and the meteorological comprehensive index based on the set meteorological data correction algorithm to obtain meteorological correction data; S4. Construct a meteorological power forecast model based on the model averaging method, and use the meteorological correction data as input to output the predicted power.

2. The method for predicting electricity consumption based on meteorological factors according to claim 1, characterized in that: In S1, the original power data is separated based on the fluctuation of meteorological data to obtain separated power, including the following steps: Obtain the original electricity data, and divide the electricity sequence in the original electricity data into a short-term fluctuation sequence and a long-term trend sequence based on the fluctuation of meteorological data; Based on the data characteristics of short-term fluctuation series and long-term trend series, the original electricity data is separated to obtain meteorological electricity, trend electricity and random electricity, and the above electricity is recorded to obtain separated electricity.

3. The method for predicting electricity consumption based on meteorological factors according to claim 1, characterized in that: In S2, the coupling effect and cumulative effect of meteorological factors on the meteorological electricity in the separated electricity are analyzed to obtain a comprehensive meteorological index, including the following steps: Based on the influence of humidity and temperature on electricity, a temperature and humidity index is constructed. Based on the temperature and humidity index, the coupling effect of meteorological electricity in the separated electricity is analyzed to obtain the coupling factor of multiple meteorological factors. Based on the multi-day cumulative effect of meteorological factors on electricity consumption, the cumulative effect analysis of the meteorological electricity in the separated electricity is carried out to obtain the cumulative effect factor; the coupling factors of multiple meteorological factors and the cumulative effect factors are statistically analyzed to obtain the meteorological comprehensive index.

4. The method for predicting electricity consumption based on meteorological factors according to claim 1, characterized in that: The meteorological data correction algorithm analyzes meteorological data and meteorological comprehensive index based on Fisher information processing method, including the following steps: Separate the meteorological factors in the meteorological data to obtain single meteorological factors, and divide the meteorological comprehensive index corresponding to each single meteorological factor into a data window according to the sampling order; The meteorological comprehensive indexes corresponding to different single meteorological factors in the data window are analyzed, the types of single meteorological factors are boxed based on the random boxing method and the corresponding probability density functions are set, and the Fisher information value of each single meteorological factor is obtained based on the probability density function.

5. The method for predicting electricity consumption based on meteorological factors according to claim 1, characterized in that: In S3, the meteorological data and the meteorological comprehensive index are corrected based on the set meteorological data correction algorithm to obtain meteorological correction data, including the following steps: Extract the meteorological forecast data from the meteorological data to obtain the forecast time of the meteorological forecast data, and interpolate the actual meteorological data at several sampling times before the forecast time to obtain a discrete meteorological data set; Based on the meteorological data correction algorithm, the Fisher information value of the discrete meteorological data set is calculated and normalized to obtain meteorological quantitative data. Based on the meteorological quantitative data, the meteorological comprehensive index and meteorological forecast data, a meteorological forecast correction formula is constructed, and meteorological correction data is obtained based on the meteorological forecast correction formula.

6. The method for predicting electricity consumption based on meteorological factors according to claim 1, characterized in that: In S4, a meteorological power forecasting model is constructed based on the model averaging method, and the meteorological correction data is used as an input value to output the predicted power, including the following steps: The meteorological factors in the meteorological correction data are screened for their characteristics to obtain meteorological factors, and the important meteorological factors are screened based on the degree of influence of meteorological factors on electricity. For each important meteorological factor, a corresponding sub-combination model is constructed based on different variable combinations, and the weight of each sub-combination model is calculated using the corresponding weight calculation criteria based on the variable combination of each sub-combination model; The variable combination corresponding to the sub-combination model corresponding to the maximum weight value of each sub-combination model is selected as the prediction influencing variable to predict the power consumption, and the error between the power prediction value and the actual power value is fitted to optimize the power prediction value to obtain the predicted power consumption.

7. The method for predicting electricity consumption based on meteorological factors according to claim 6, characterized in that: Based on the degree of influence of meteorological factors on electricity, various meteorological factors are screened to obtain important meteorological factors, including the following steps: a meteorological factor evaluation index system is constructed based on the degree of influence of meteorological factors on electricity, each meteorological factor is evaluated based on the meteorological factor evaluation index system to obtain a meteorological impact score, and meteorological factors corresponding to meteorological impact scores exceeding a set score threshold are screened to obtain important meteorological factors.

8. The method for predicting electricity consumption based on meteorological factors according to claim 6, characterized in that: The sub-combination model automatically selects the most suitable weight calculation criterion from the database based on the number of combinations to calculate the weight of each sub-combination model, and the weight calculation criterion at least includes AIC criterion, BIC criterion, Jackknife criterion and LASSO method.

9. A method for predicting electricity consumption based on meteorological factors according to claim 1, 3 or 5, characterized in that: The meteorological data at least includes meteorological forecast data, meteorological factors and actual meteorological data.

10. The method for predicting electricity consumption based on meteorological factors according to claim 9, characterized in that: The meteorological factors include at least temperature, humidity, wind speed and rainfall.

Images