Power system load prediction method and system based on meteorological evolution process

Abstract

The invention discloses an electric power system load prediction method based on a meteorological evolution process. The method comprises the following steps: constructing a target electric power system load prediction index set comprising meteorological factors; obtaining corresponding historical data and future prediction data to construct a historical data set and a future data set; performing threshold judgment on the temperature change of the day to be predicted; constructing a training data set and training the load prediction primary model to obtain a load prediction model; and adopting the load prediction model to carry out actual load prediction based on the meteorological evolution process on the target power system. The invention also discloses a system for realizing the power system load prediction method based on the meteorological evolution process. According to the invention, through processing and matching of historical meteorological data and meteorological data on a prediction day, a data set conforming to a prediction condition is constructed, and a load prediction model is adopted to learn the data set; therefore, the method not only can realize power system load prediction based on the meteorological evolution process, but also is higher in reliability and better in accuracy.

Description

Technical Field

[0001] The present invention belongs to the field of electrical automation, and in particular relates to a method and system for predicting load of an electric power system based on a meteorological evolution process. Background Art

[0002] With the development of economy and technology and the improvement of people's living standards, electricity has become an indispensable secondary energy source in people's production and life, bringing endless convenience to people's production and life. Therefore, ensuring the stable and reliable supply of electricity has become one of the most important tasks of the power system.

[0003] Power system load forecasting is one of the important tasks of the power system. Accurate and reliable load forecasting results can greatly improve the safety and stability of power system operation. Therefore, power system load forecasting has always been the focus of researchers.

[0004] The load forecast results of the power system are closely related to weather factors. At present, the traditional power system load forecasting scheme can obtain good forecasting accuracy when the weather conditions remain relatively stable. However, when faced with drastic changes in the weather in a short period of time, such as large-scale cold waves and large-scale heat waves, the existing power system load forecasting scheme is prone to large forecasting deviations, which greatly affects the safe and stable operation of the power system. Summary of the invention

[0005] One of the purposes of the present invention is to provide a method for forecasting power system load based on meteorological evolution process with high reliability and good accuracy.

[0006] A second object of the present invention is to provide a system for implementing the power system load forecasting method based on the meteorological evolution process.

[0007] The method for forecasting power system load based on meteorological evolution provided by the present invention comprises the following steps:

[0008] S1. Construct a set of load forecasting indicators for the target power system including meteorological factors;

[0009] S2. According to the indicator set obtained in step S1, obtain the corresponding historical data to construct a historical data set, and obtain the corresponding future prediction data to construct a future data set;

[0010] S3. Based on the meteorological data forecast information of the target power system on the forecast day, a threshold value is determined for the temperature change on the forecast day;

[0011] S4. Construct a training data set based on the threshold determination result obtained in step S3 and the historical data set obtained in step S2;

[0012] S5. Using the training data set constructed in step S4, the selected load forecasting primary model is trained to obtain a load forecasting model;

[0013] S6. Using the load forecasting model obtained in step S5, actual load forecasting based on the meteorological evolution process is performed on the target power system.

[0014] The step S1 of constructing a set of load forecasting indicators for the target power system including meteorological factors specifically includes the following steps:

[0015] The constructed load forecasting index set includes time data, calendar data, load data and meteorological factor data;

[0016] The time data includes a certain time in 24 hours a day, and the i-th time is represented by T i ;

[0017] The calendar data includes week data, and the kth day of a week is represented by W k ;

[0018] The load data includes the load data of the previous day, and the load data at the i-th moment of the previous day is represented by L i-24 ;

[0019] The meteorological factor data include temperature data, humidity data, wind speed data, rainfall data, cumulative effect data and meteorological transformation data; the temperature data include the temperature C at the ith moment i , the highest temperature in a day C max and the lowest temperature of the day min ; Humidity data includes humidity D at the i-th moment i ; Wind speed data includes the wind speed S at the i-th moment i ; The rainfall data includes the rainfall R at the i-th moment i ; Cumulative effect data include the cumulative number of days the temperature exceeds the set value G H , the cumulative number of days the temperature is below the set value G L and the cumulative number of extreme days G S ; Meteorological transformation data includes the temperature difference at the same time of the previous day The temperature difference from the previous moment Temperature difference at the same time of the previous day is the difference between the current temperature and the temperature at the same time of the previous day. is the temperature difference between the current moment and the previous moment.

[0020] According to the indicator set obtained in step S1, the step S2 obtains the corresponding historical data to construct a historical data set, and simultaneously obtains the corresponding future prediction data to construct a future data set, which specifically includes the following steps:

[0021] Obtain the corresponding historical data and build historical data set B L for in, is the independent variable historical data set, and The value of is the dependent variable historical data set, and The value of L i is the load at the i-th moment;

[0022] Get the corresponding future prediction data and build future data set B F for in is the future data set of independent variables, and The value of The superscript “'” indicates the future forecast data of the corresponding variable; is the dependent variable future dataset, and The value of L' i is the predicted load at the i-th moment in the future;

[0023] Calculate the meteorological evolution process indicators for the forecast day, including the mean temperature C' for the forecast day mean , the maximum temperature of the day to be predicted C' max , the average value of the temperature difference at the same time of the day before the forecast day C' min , the maximum value of the temperature difference at the same time of the day before the prediction day The minimum value of the temperature difference at the same time on the day before the forecast and the cumulative number of extreme days G' S .

[0024] The step S3 of performing threshold determination on the temperature change on the day to be predicted based on the meteorological data prediction information of the target power system on the day to be predicted specifically includes the following steps:

[0025] According to the meteorological evolution process index of the day to be predicted obtained in step S2, a 2N+1 day time window meteorological evolution process index data set Q is constructed as Q=[P d-N ,…,P d ,…,P d+N ], P d is the index data of meteorological evolution process on the dth day, P d-N is the meteorological evolution process index data of the N days before the dth day, P d+NThe meteorological evolution process index data for N days after the dth day;

[0026] In the constructed 2N+1 day time window, the following temperature drastic change indicators are calculated:

[0027] Calculate the first indicator

[0028] Calculate the second index for The value of d is [dN,d+N];

[0029] Calculate the third index

[0030] The following rules are used to judge the temperature change on the forecast day:

[0031] If the first indicator Not greater than the first threshold, the second index Not greater than the second threshold and the third index If the value is not greater than the third threshold, the temperature change on the forecast day is judged to be “not drastic”;

[0032] If the first indicator Greater than the first threshold, the second index Greater than the second threshold or the third indicator If the temperature change on the forecast day is greater than the third threshold, the temperature change on the forecast day is determined to be “dramatic”.

[0033] Step S4, based on the threshold determination result obtained in step S3 and the historical data set obtained in step S2, constructs a training data set, which specifically includes the following steps:

[0034] If the temperature change on the forecast day is determined to be “not drastic”, the historical data set obtained in step S2 is used as the training data set;

[0035] If the temperature change on the forecast day is judged as "dramatic", the following steps are used to construct the training data set:

[0036] Construct a 2N+1 day time window to represent the meteorological evolution process indicator dataset Q for the historical m days m , denoted as Q m =[P m-N ,…,P m ,…,P m+N ], P m is the meteorological evolution process index data of the mth day in history, P m-N is the meteorological evolution process index data of the N days before the mth day in history, P m+N The meteorological evolution process index data for the N days after the mth day in history;

[0037] The following formula is used to calculate the matching degree value D(Q m ,Q):

[0038]

[0039] In the formula is the meteorological evolution process index data of the kth day in the historical meteorological evolution process index data set of the mth day; P k is the meteorological evolution process index data of the kth day in the forecast daily meteorological evolution process index data set;

[0040] According to the obtained matching degree value D(Q m ,Q), select the meteorological evolution process indicator data sets corresponding to several groups of data with the highest matching degree values ​​to construct the training data set.

[0041] Step S5 uses the training data set constructed in step S4 to train the selected load forecasting primary model to obtain a load forecasting model, which specifically includes the following steps:

[0042] Use machine learning models or deep learning models as primary load forecasting models;

[0043] The training data set obtained in step S4 is used to train the load forecasting primary model to obtain a load forecasting model.

[0044] The step S5 specifically includes the following steps:

[0045] The LightGBM model is used as the selected primary model for load forecasting;

[0046] The following formula is used as the objective function of the primary load forecasting model:

[0047]

[0048] Where Obj (kk) is the objective function value at the kkth iteration; NN is the total number of variable factors; for y ii and The squared loss function is y ii The model prediction value for the current iteration; is the model prediction value at the kk-1th iteration; g ii is the first intermediate variable, and For Find the first-order derivative; f kk (x ii ) is the model fitting function added in the kkth step; h iiis the second intermediate variable, and Express Find the second-order derivative; Ω(f kk ) is the regular term corresponding to the kth iteration; constant is the set constant term;

[0049] The training data set obtained in step S4 is used to train the load forecasting primary model to obtain a load forecasting model.

[0050] The step S5 further comprises the following steps:

[0051] Get the test dataset;

[0052] Use the test data set to test the trained load forecasting model;

[0053] During the test, the following formula is used to calculate the accuracy R of the load forecasting model: 2 :

[0054]

[0055] In the formula is the predicted value of power load at time i; is the actual value of the power load at time i; is the average power load in the test data set; N t The number of days for testing;

[0056] According to the obtained accuracy R 2 , determine the load forecasting model:

[0057] If the accuracy R 2 If it is greater than the set accuracy threshold, the load forecasting model is judged to meet the accuracy requirements and the subsequent steps are carried out;

[0058] If the accuracy R 2 If the load forecasting model is less than or equal to the set accuracy threshold, it is determined that the load forecasting model does not meet the accuracy requirement, and step S5 is repeated to retrain the selected load forecasting primary model until the trained load forecasting model meets the accuracy requirement.

[0059] The present invention also provides a system for implementing the method for load forecasting of an electric power system based on a meteorological evolution process, comprising a data acquisition module, a data set construction module, a meteorological determination module, a training set construction module, a model training module and a load forecasting module; the data acquisition module, the data set construction module, the meteorological determination module, the training set construction module, the model training module and the load forecasting module are connected in series in sequence; the data acquisition module is used to construct a set of load forecasting indicators for a target electric power system including meteorological factors, and upload the data information to the data set construction module; the data set construction module is used to obtain corresponding historical data to construct a historical data set based on the received data information and the obtained indicator set, and simultaneously obtain corresponding future forecast data to construct a future data set, and upload the data information to the meteorological determination module; the meteorological The determination module is used to perform threshold determination on the temperature change on the day to be predicted based on the received data information and the meteorological data prediction information of the target power system on the day to be predicted, and upload the data information to the training set construction module; the training set construction module is used to construct a training data set based on the received data information and the obtained threshold determination result and based on the obtained historical data set, and upload the data information to the model training module; the model training module is used to train the selected load prediction primary model based on the received data information and the constructed training data set to obtain the load prediction model, and upload the data information to the load prediction module; the load prediction module is used to perform actual load prediction based on the meteorological evolution process for the target power system based on the received data information and the obtained load prediction model.

[0060] The method and system for power system load forecasting based on the meteorological evolution process provided by the present invention constructs a data set that meets the forecast situation by processing and matching historical meteorological data and forecast day meteorological data, and uses a load forecasting model to learn the data set; therefore, the present invention can not only realize power system load forecasting based on the meteorological evolution process, but also has higher reliability and better accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 The figure is a schematic diagram of the method flow of the present invention.

[0062] Figure 2 Schematic diagram of the functional modules of the system of the present invention. DETAILED DESCRIPTION

[0063] like Figure 1 The method flow chart of the method of the present invention is shown as follows: the power system load forecasting method based on the meteorological evolution process provided by the present invention comprises the following steps:

[0064] S1. Construct a set of load forecasting indicators for the target power system including meteorological factors; specifically, the following steps are included:

[0065] The constructed load forecasting index set includes time data, calendar data, load data and meteorological factor data;

[0066] The time data includes a certain time in 24 hours a day, and the i-th time is represented by T i ;

[0067] The calendar data includes week data, and the kth day of a week is represented by W k ;

[0068] The load data includes the load data of the previous day, and the load data at the i-th moment of the previous day is represented by L i-24 ;

[0069] The meteorological factor data include temperature data, humidity data, wind speed data, rainfall data, cumulative effect data and meteorological transformation data; the temperature data include the temperature C at the ith moment i , the highest temperature in a day C max and the lowest temperature of the day min ; Humidity data includes humidity D at the i-th moment i ; Wind speed data includes the wind speed S at the i-th moment i ; The rainfall data includes the rainfall R at the i-th moment i ; Cumulative effect data include the cumulative number of days the temperature exceeds the set value G H , the cumulative number of days the temperature is below the set value G L and the cumulative number of extreme days G S ; Meteorological transformation data includes the temperature difference at the same time of the previous day The temperature difference from the previous moment Temperature difference at the same time of the previous day is the difference between the current temperature and the temperature at the same time of the previous day. is the temperature difference between the current moment and the previous moment;

[0070] S2. According to the indicator set obtained in step S1, the corresponding historical data is obtained to construct a historical data set, and the corresponding future prediction data is obtained to construct a future data set; specifically, the steps include:

[0071] Obtain the corresponding historical data and build historical data set B L for in, is the independent variable historical data set, and The value of is the dependent variable historical data set, and The value of L iis the load at the i-th moment;

[0072] Get the corresponding future prediction data and build future data set B F for in is the future data set of independent variables, and The value of The superscript “'” indicates the future forecast data of the corresponding variable; is the dependent variable future dataset, and The value of L' i is the predicted load at the i-th moment in the future;

[0073] Calculate the meteorological evolution process indicators for the forecast day, including the mean temperature C' for the forecast day mean , the maximum temperature of the day to be predicted C' max , the average value of the temperature difference at the same time of the day before the forecast day C' min , the maximum value of the temperature difference at the same time of the day before the prediction day The minimum value of the temperature difference at the same time on the day before the forecast and the cumulative number of extreme days G' S ;

[0074] S3. According to the meteorological data forecast information of the target power system on the forecast day, a threshold value is determined for the temperature change on the forecast day; specifically, the steps include:

[0075] According to the meteorological evolution process index of the day to be predicted obtained in step S2, a 2N+1 day time window meteorological evolution process index data set Q is constructed as Q=[P d-N ,…,P d ,…,P d+N ], P d is the index data of meteorological evolution process on the dth day, P d-N is the meteorological evolution process index data of the N days before the dth day, P d+N The meteorological evolution process index data for N days after the dth day;

[0076] In the constructed 2N+1 day time window, the following temperature drastic change indicators are calculated:

[0077] Calculate the first indicator for

[0078] Calculate the second index for The value of d is [dN,d+N];

[0079] Calculate the third index for

[0080] The following rules are used to judge the temperature change on the forecast day:

[0081] If the first indicator Not greater than the first threshold (preferably 4°C), the second index Not greater than the second threshold (preferably 8°C) and the third indicator If the temperature change on the forecast day is not greater than the third threshold (preferably 7 days), the temperature change on the forecast day is determined to be “not drastic”;

[0082] If the first indicator Greater than the first threshold (preferably 4°C), the second indicator Greater than the second threshold (preferably 8°C) or the third indicator If the temperature change is greater than a third threshold (preferably 7 days), the temperature change on the forecast day is determined to be “dramatic”;

[0083] S4. According to the threshold determination result obtained in step S3, a training data set is constructed based on the historical data set obtained in step S2; specifically, the steps include:

[0084] If the temperature change on the forecast day is determined to be “not drastic”, the historical data set obtained in step S2 is used as the training data set;

[0085] If the temperature change on the forecast day is judged as "dramatic", the following steps are used to construct the training data set:

[0086] Construct a 2N+1 day time window to represent the meteorological evolution process indicator dataset Q for the historical m days m , denoted as Q m =[P m-N ,…,P m ,…,P m+N ], P m is the meteorological evolution process index data of the mth day in history, P m-N is the meteorological evolution process index data of the N days before the mth day in history, P m+N The meteorological evolution process index data for the N days after the mth day in history;

[0087] The following formula is used to calculate the matching degree value D(Q m ,Q):

[0088]

[0089] In the formula is the meteorological evolution process index data of the kth day in the historical meteorological evolution process index data set of the mth day; P kis the meteorological evolution process index data of the kth day in the forecast daily meteorological evolution process index data set;

[0090] According to the obtained matching degree value D(Q m ,Q), select the meteorological evolution process indicator data sets corresponding to several groups of data with the highest matching degree values ​​to construct the training data set;

[0091] S5. Using the training data set constructed in step S4, the selected load forecasting primary model is trained to obtain a load forecasting model; specifically comprising the following steps:

[0092] Use machine learning models or deep learning models as primary load forecasting models;

[0093] Using the training data set obtained in step S4, the primary load forecasting model is trained to obtain a load forecasting model;

[0094] Take the LightGBM model as an example:

[0095] The LightGBM model is used as the selected primary model for load forecasting;

[0096] The following formula is used as the objective function of the primary load forecasting model:

[0097]

[0098] Where Obj (kk) is the objective function value at the kkth iteration; NN is the total number of variable factors; for y ii and The squared loss function is y ii The model prediction value for the current iteration; is the model prediction value at the kk-1th iteration; g ii is the first intermediate variable, and For Find the first-order derivative; f kk (x ii ) is the model fitting function added in step kk (preferably, the least squares fitting function can be used); h ii is the second intermediate variable, and Express Find the second-order derivative; Ω(f kk ) is the regular term corresponding to the kth iteration; constant is the set constant term;

[0099] Using the training data set obtained in step S4, the primary load forecasting model is trained to obtain a load forecasting model;

[0100] In addition, after the training is completed, the trained load forecasting model can be used directly or after testing;

[0101] When conducting the test, you can use the following steps to conduct the test:

[0102] Get the test dataset;

[0103] Use the test data set to test the trained load forecasting model;

[0104] During the test, the following formula is used to calculate the accuracy R of the load forecasting model: 2 :

[0105]

[0106] In the formula is the predicted value of power load at time i; is the actual value of the power load at time i; is the average power load in the test data set; N t The number of days for testing;

[0107] According to the obtained accuracy R 2 , determine the load forecasting model:

[0108] If the accuracy R 2 If it is greater than the set accuracy threshold, the load forecasting model is judged to meet the accuracy requirements and the subsequent steps are carried out;

[0109] If the accuracy R 2 If the value is less than or equal to the set accuracy threshold, it is determined that the load forecasting model does not meet the accuracy requirement, and step S5 is repeated to train the selected load forecasting primary model again until the trained load forecasting model meets the accuracy requirement;

[0110] S6. Using the load forecasting model obtained in step S5, actual load forecasting based on the meteorological evolution process is performed on the target power system.

[0111] like Figure 2The functional module schematic diagram of the system of the present invention is shown as follows: the system disclosed in the present invention for implementing the method for load forecasting of an electric power system based on a meteorological evolution process comprises a data acquisition module, a data set construction module, a meteorological determination module, a training set construction module, a model training module and a load forecasting module; the data acquisition module, the data set construction module, the meteorological determination module, the training set construction module, the model training module and the load forecasting module are connected in series in sequence; the data acquisition module is used to construct a set of load forecasting indicators for a target electric power system including meteorological factors, and upload the data information to the data set construction module; the data set construction module is used to obtain corresponding historical data to construct a historical data set based on the received data information and the obtained indicator set, and simultaneously obtain corresponding future forecast data to construct a future data set, and upload the data information to the meteorological determination module. The meteorological judgment module is used to perform threshold judgment on the temperature change on the forecast day according to the received data information and the meteorological data forecast information of the target power system on the forecast day, and upload the data information to the training set construction module; the training set construction module is used to construct a training data set based on the received data information and the obtained threshold judgment result and the obtained historical data set, and upload the data information to the model training module; the model training module is used to train the selected load forecasting primary model according to the received data information and the constructed training data set to obtain the load forecasting model, and upload the data information to the load forecasting module; the load forecasting module is used to perform actual load forecasting based on the meteorological evolution process for the target power system according to the received data information and the obtained load forecasting model.

Claims

1. A method for predicting power system load based on meteorological evolution process, comprising the following steps: S1. Construct a set of load forecasting indicators for the target power system including meteorological factors; S2. According to the indicator set obtained in step S1, obtain the corresponding historical data to construct a historical data set, and obtain the corresponding future prediction data to construct a future data set; S3. Based on the meteorological data forecast information of the target power system on the forecast day, a threshold value is determined for the temperature change on the forecast day; S4. Construct a training data set based on the threshold determination result obtained in step S3 and the historical data set obtained in step S2; S5. Using the training data set constructed in step S4, the selected load forecasting primary model is trained to obtain a load forecasting model; S6. Using the load forecasting model obtained in step S5, actual load forecasting based on the meteorological evolution process is performed on the target power system.

2. The method for predicting power system load based on meteorological evolution according to claim 1 is characterized in that The step S1 of constructing a set of load forecasting indicators for the target power system including meteorological factors specifically includes the following steps: The constructed load forecasting index set includes time data, calendar data, load data and meteorological factor data; The time data includes a certain time in 24 hours a day, and the i-th time is represented by T i ; The calendar data includes week data, and the kth day of a week is represented by W k ; The load data includes the load data of the previous day, and the load data at the i-th moment of the previous day is represented by L i-24 ; The meteorological factor data include temperature data, humidity data, wind speed data, rainfall data, cumulative effect data and meteorological transformation data; the temperature data include the temperature C at the ith moment i , the highest temperature in a day C max and the lowest temperature of the day min ; Humidity data includes humidity D at the i-th moment i ; Wind speed data includes the wind speed S at the i-th moment i ; The rainfall data includes the rainfall R at the i-th moment i ; Cumulative effect data include the cumulative number of days the temperature exceeds the set value G H , the cumulative number of days the temperature is below the set value G L and the cumulative number of extreme days G S ; Meteorological transformation data includes the temperature difference at the same time of the previous day The temperature difference from the previous moment Temperature difference at the same time of the previous day is the difference between the current temperature and the temperature at the same time of the previous day. is the temperature difference between the current moment and the previous moment.

3. The method for predicting power system load based on meteorological evolution according to claim 2 is characterized in that According to the indicator set obtained in step S1, the step S2 obtains the corresponding historical data to construct a historical data set, and simultaneously obtains the corresponding future prediction data to construct a future data set, which specifically includes the following steps: Obtain the corresponding historical data and build historical data set B L for in, is the independent variable historical data set, and The value of is the dependent variable historical data set, and The value of L i is the load at the i-th moment; Get the corresponding future prediction data and build a future data set in is the future data set of the independent variable, and The value of The superscript "'" indicates the future forecast data of the corresponding variable; is the dependent variable future dataset, and The value of L' i is the predicted load at the i-th moment in the future; Calculate the meteorological evolution process indicators for the forecast day, including the mean temperature C' for the forecast day mean , the maximum temperature of the day to be predicted C' max , the average value of the temperature difference at the same time of the day before the forecast day C' min , the maximum value of the temperature difference at the same time of the day before the prediction day The minimum value of the temperature difference at the same time on the day before the forecast and the cumulative number of extreme days G' S .

4. The method for predicting power system load based on meteorological evolution according to claim 3 is characterized in that Step S3, based on the meteorological data forecast information of the target power system on the forecast day, performs threshold determination on the temperature change on the forecast day, specifically including the following steps: According to the meteorological evolution process index of the day to be predicted obtained in step S2, a 2N+1 day time window meteorological evolution process index data set Q is constructed as Q=[P d-N ,…,P d ,…,P d+N ], P d is the index data of meteorological evolution process on the dth day, P d-N is the meteorological evolution process index data of the N days before the dth day, P d+N The meteorological evolution process index data for N days after the dth day; In the constructed 2N+1 day time window, the following temperature drastic change indicators are calculated: Calculate the first indicator for Calculate the second index for The value of d is [dN,d+N]; Calculate the third index for The following rules are used to judge the temperature change on the forecast day: If the first indicator Not greater than the first threshold, the second index Not greater than the second threshold and the third index If the value is not greater than the third threshold, the temperature change on the forecast day is judged to be "not severe"; If the first indicator Greater than the first threshold, the second index Greater than the second threshold or the third indicator If it is greater than the third threshold, the temperature change on the predicted day is determined to be "dramatic".

5. The method for predicting power system load based on meteorological evolution according to claim 4 is characterized in that Step S4, based on the threshold determination result obtained in step S3 and the historical data set obtained in step S2, constructs a training data set, which specifically includes the following steps: If the temperature change on the forecast day is determined to be "not severe", the historical data set obtained in step S2 is used as the training data set; If the temperature change on the forecast day is judged as "dramatic", the following steps are used to construct the training data set: Construct a 2N+1 day time window to represent the meteorological evolution process indicator dataset Q for the historical m days m , denoted as Q m =[P m-N ,…,P m ,…,P m+N ], P m is the meteorological evolution process index data of the mth day in history, P m-N is the meteorological evolution process index data of the N days before the mth day in history, P m+N The meteorological evolution process index data for the N days after the mth day in history; The following formula is used to calculate the matching degree value D(Q m ,Q): In the formula is the meteorological evolution process index data of the kth day in the historical meteorological evolution process index data set of the mth day; P k is the meteorological evolution process index data of the kth day in the forecast daily meteorological evolution process index data set; According to the obtained matching degree value D(Q m ,Q), select the meteorological evolution process indicator data sets corresponding to several groups of data with the highest matching degree values ​​to construct the training data set.

6. The method for predicting power system load based on meteorological evolution according to claim 5 is characterized in that Step S5 uses the training data set constructed in step S4 to train the selected load forecasting primary model to obtain a load forecasting model, which specifically includes the following steps: Use machine learning models or deep learning models as primary load forecasting models; The training data set obtained in step S4 is used to train the load forecasting primary model to obtain a load forecasting model.

7. The method for predicting power system load based on meteorological evolution according to claim 6 is characterized in that The step S5 specifically includes the following steps: The LightGBM model is used as the selected primary model for load forecasting; The following formula is used as the objective function of the primary load forecasting model: Where Obj (kk) is the objective function value at the kkth iteration; NN is the total number of variable factors; for y ii and The squared loss function is y ii The model prediction value for the current iteration; is the model prediction value at the kk-1th iteration; g ii is the first intermediate variable, and For Find the first-order derivative; f kk (x ii ) is the model fitting function added in the kkth step; h ii is the second intermediate variable, and Express Find the second-order derivative; Ω(f kk ) is the regular term corresponding to the kth iteration; constant is the set constant term; The training data set obtained in step S4 is used to train the load forecasting primary model to obtain a load forecasting model.

8. The method for predicting power system load based on meteorological evolution according to claim 6 is characterized in that The step S5 further comprises the following steps: Get the test dataset; Use the test data set to test the trained load forecasting model; During the test, the following formula is used to calculate the accuracy R of the load forecasting model: 2 : In the formula is the predicted value of power load at time i; is the actual value of the power load at time i; is the average power load in the test data set; N t The number of days for testing; According to the obtained accuracy R 2 , determine the load forecasting model: If the accuracy R 2 If it is greater than the set accuracy threshold, the load forecasting model is judged to meet the accuracy requirements and the subsequent steps are carried out; If the accuracy R 2 If the load forecasting model is less than or equal to the set accuracy threshold, it is determined that the load forecasting model does not meet the accuracy requirement, and step S5 is repeated to retrain the selected load forecasting primary model until the trained load forecasting model meets the accuracy requirement.

9. A system for implementing the method for forecasting power system load based on meteorological evolution process as claimed in any one of claims 1 to 8, characterized in that It includes a data acquisition module, a data set construction module, a meteorological determination module, a training set construction module, a model training module and a load forecasting module; the data acquisition module, the data set construction module, the meteorological determination module, the training set construction module, the model training module and the load forecasting module are connected in series in sequence; the data acquisition module is used to construct a set of load forecasting indicators for the target power system including meteorological factors, and upload the data information to the data set construction module; the data set construction module is used to obtain corresponding historical data to construct a historical data set based on the received data information and the obtained indicator set, and simultaneously obtain corresponding future forecast data to construct a future data set, and upload the data information to the meteorological determination module; The meteorological determination module is used to perform threshold determination on the temperature change on the forecast day according to the received data information and the forecast information of the meteorological data of the target power system on the forecast day, and upload the data information to the training set construction module; The training set construction module is used to construct a training data set based on the received data information, the obtained threshold judgment result, and the obtained historical data set, and upload the data information to the model training module; The model training module is used to train the selected load forecasting primary model according to the received data information and the constructed training data set to obtain the load forecasting model, and upload the data information to the load forecasting module; The load forecasting module is used to carry out actual load forecasting based on the meteorological evolution process for the target power system according to the received data information and the obtained load forecasting model.

Images