Summer load forecasting methods and systems for power systems

Abstract

This invention discloses a method for predicting summer load in a power system, comprising: acquiring historical load data of the target power system and processing it to obtain training and testing datasets; constructing an initial summer load prediction model and training and testing it to obtain a summer load prediction model; constructing future load datasets of the target power system under different scenarios and using the summer load prediction model to complete the prediction of the target power system's future summer load. This invention also discloses a system for implementing the aforementioned method for predicting summer load in a power system. This invention obtains the final probabilistic load prediction result by constructing influencing variables of power system load and a load prediction model based on LightGBM, and by deeply mining and utilizing historical meteorological data from the hottest days of summer; therefore, this invention can not only achieve summer load prediction for power systems over a longer period, but also has high reliability, good accuracy, and is objective and scientific.

Description

Technical Field

[0001] This invention belongs to the field of electrical automation, and specifically relates to a method and system for predicting summer loads in power systems. Background Technology

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

[0003] Currently, the summer heatwave is intensifying, posing a significant load challenge to the power system. To ensure a stable power supply during the summer, the power system needs to make comprehensive and accurate forecasts of the load throughout the season, allowing for the early development of corresponding production and operation plans. Therefore, summer load forecasting is of paramount importance for the power system.

[0004] Currently, traditional load forecasting schemes for power systems can generally only predict load data for the next 40 days, while summer load forecasting requires predicting load data for at least 60 days. This makes existing load forecasting schemes unsuitable for direct application to summer load forecasting. Moreover, existing load forecasting schemes suffer from poor accuracy and reliability when forecasting highly time-sensitive power loads such as summer loads. Summary of the Invention

[0005] One of the objectives of this invention is to provide a reliable, accurate, and objective scientific method for predicting summer loads in power systems.

[0006] The second objective of this invention is to provide a system for implementing the aforementioned power system summer load forecasting method.

[0007] The power system summer load forecasting method provided by this invention includes the following steps:

[0008] S1. Obtain historical load data information of the target power system;

[0009] S2. Process the historical load data information obtained in step S1 to construct the training dataset and the test dataset;

[0010] S3. Construct an initial model for summer load forecasting based on the LightGBM model;

[0011] S4. Use the training dataset constructed in step S2 to train the initial summer load forecasting model constructed in step S3 to obtain the summer load forecasting test model.

[0012] S5. The summer load forecast test model obtained in step S4 is tested using the test dataset constructed in step S3 to obtain the summer load forecast model;

[0013] S6. Based on the historical load data of the target power system obtained in step S1, construct a future load dataset of the target power system under different scenarios;

[0014] S7. Based on the future load dataset constructed in step S6, use the summer load forecasting model obtained in step S5 to complete the future summer load forecast of the target power system.

[0015] Step S1, which involves obtaining historical load data information of the target power system, specifically includes the following steps:

[0016] The acquired data includes historical meteorological data, historical calendar data, and historical load data;

[0017] Historical meteorological data includes real-time temperature data for the past 24 hours. i The highest temperature in history within a day Historical low temperature The highest temperature on the day before the historical day

[0018] Historical calendar data includes historical day time data T i Historical day of the week data W k ;

[0019] Historical load data includes hourly load data from the day before the historical date. i-24 ;

[0020] Where i represents the i-th moment in a historical day; k represents the k-th day of a week.

[0021] Step S2 involves processing the historical load data obtained in step S1 to construct the training dataset and the test dataset, specifically including the following steps:

[0022] The historical load data obtained in step S1 is cleaned and interpolated to obtain a complete set of historical load data.

[0023] Construct the dataset H as H = [H x H y ], where H x The independent variable is a historical dataset and H y For the historical dataset of the dependent variable and H y =[L i ], L i This is the load data for historical days.

[0024] The constructed dataset H is divided into a training dataset and a test dataset according to a set ratio.

[0025] Step S3, which involves constructing an initial model for summer load forecasting based on the LightGBM model, specifically includes the following steps:

[0026] The following formula is used as the objective function of the initial model for summer load forecasting:

[0027]

[0028] In the formula Obj (kk) Let be the objective function value at the k-th iteration; NN is the total number of variable factors; y ii This is the model prediction value in the current iteration; This is the model prediction value at step kk-1; It is a squared loss function and g ii The first intermediate variable and Indicates to Find the first derivative; f kk (x ii ) represents the model fitting function added at step k; h ii It is the second intermediate variable and Indicates to Find the second derivative; Ω(f) kk ) represents the regularization term corresponding to the k-th iteration; constant is the set constant term.

[0029] Step S5 involves testing the summer load forecasting test model obtained in step S4 using the test dataset constructed in step S3 to obtain the summer load forecasting model. This process specifically includes the following steps:

[0030] The summer load forecast test model obtained in step S4 was tested using the test dataset constructed in step S3.

[0031] The accuracy R of the summer load forecasting test model is calculated using the following formula. 2 :

[0032]

[0033] In the formula N t L represents the number of days of testing. t j Let j be the predicted power load value at time j; Let j be the actual power load value at time j; The average power load in the test dataset;

[0034] Evaluation of the summer load forecasting test model:

[0035] If the precision value R 2 If the value is greater than or equal to the set threshold, it means that the summer load forecast test model meets the set requirements; the corresponding summer load forecast test model will be used as the final summer load forecast model.

[0036] If the precision value R 2 If the value is less than the set threshold, repeat steps S4 to S5 until the accuracy value R of the summer load forecast test model is reached. 2 Greater than or equal to the set threshold.

[0037] Step S6, which involves constructing a future load dataset for the target power system under different scenarios based on the historical load data information of the target power system obtained in step S1, specifically includes the following steps:

[0038] The constructed future dataset F is F = [F x ,F y ], where F x For the independent variable, the future dataset and Among them L' i-24 To predict the time-based load data for the day before, T i 'W' provides the time data for the predicted day. k 'For the weekday data of the forecast date, C i 'This is the real-time temperature data for the 24 hours of the predicted day.' To predict the day's highest temperature, To predict the day's lowest temperature, To predict the highest temperature of the day before; F y For the dependent variable, the future dataset and F y =[L' i ], L' i To predict the power load at time i on day i;

[0039] Align the dates of the hottest days of summer (Sanfu) for each year with the dates of the hottest days of summer (Sanfu) for the target year;

[0040] Construct summer load forecast datasets for different historical years; where the meteorological forecast datasets for different scenarios are represented as C. case The total number of scenes is represented by N. case The weather forecast dataset for the nth scenario is represented as

[0041] Obtain the weather forecast dataset and represent the data for the previous N days as follows: Replace the top N meteorological data points in all meteorological forecast datasets for different years with... Obtain new weather forecast datasets for different scenarios;

[0042] Ultimately, we obtain future load datasets for the target power system under different scenarios.

[0043] Step S7, based on the future load dataset constructed in step S6, uses the summer load forecasting model obtained in step S5 to complete the future summer load forecasting of the target power system, specifically including the following steps:

[0044] Based on the future load dataset constructed in step S6, the summer load forecasting model obtained in step S5 is used to forecast loads for different year scenarios, and the final load forecast results are obtained. Represented as This represents the load prediction result at time i in the nth scenario;

[0045] The load forecast results for each time step in all scenarios are sorted, and the sorted load forecast results are represented as P. pre The number of load forecast results at each time point is the same as the total number of scenarios;

[0046] Based on the sorted load forecast results P pre The words are processed using the following formula for probabilistic processing:

[0047]

[0048] In the formula The load prediction result at time i with probability q; The sorted load forecast results P pre The m-th load forecast result at time i; The sorted load forecast results P pre The (m+1)th load forecast result at time i; g is the third intermediate variable and g = N case *pm,N case Let p be the total number of scenes, and p be the fourth intermediate variable. q is a probability value ranging from 1 to 99, and m is the fifth intermediate variable. This is the floor symbol;

[0049] Finally, what was obtained The forecast results for the future summer load of the target power system.

[0050] This invention also provides a system for implementing the aforementioned power system summer load forecasting method, comprising a data acquisition module, a dataset construction module, a prediction model construction module, a model training module, a model testing module, a future data construction module, and a load forecasting module; the data acquisition module, dataset construction module, prediction model construction module, model training module, model testing module, future data construction module, and load forecasting module are connected in series; the data acquisition module is used to acquire historical load data information of the target power system and upload the data to the dataset construction module; the dataset construction module is used to process the acquired historical load data information according to the received data, construct training datasets and test datasets, and upload the data to the prediction model construction module; the prediction model construction module is used to construct a summer load forecasting model based on the received data and the LightGBM model. The system first generates an initial summer load forecasting model and uploads the data to the model training module. The model training module trains the initial summer load forecasting model using a training dataset based on the received data, obtaining a summer load forecasting test model, and uploads the data to the model testing module. The model testing module tests the summer load forecasting test model using a test dataset based on the received data, obtaining a summer load forecasting model, and uploads the data to the future data construction module. The future data construction module constructs future load datasets for the target power system under different scenarios based on the received data and historical load data, and uploads the data to the load forecasting module. The load forecasting module, based on the received data and the future load dataset, uses the summer load forecasting model to complete the future summer load forecast for the target power system.

[0051] The power system summer load forecasting method and system provided by this invention constructs the influencing variables of power system load and a load forecasting model based on LightGBM, and obtains the final probabilistic load forecasting results through in-depth mining and utilization of historical meteorological data information during the hottest days of summer. Therefore, this invention can not only achieve power system summer load forecasting for a long period of time, but also has high reliability, good accuracy, and is objective and scientific. Attached Figure Description

[0052] Figure 1 This is a schematic diagram of the method flow of the present invention.

[0053] Figure 2 This is a schematic diagram of the functional modules of the system of the present invention. Detailed Implementation

[0054] like Figure 1 The diagram shown is a flowchart of the method of the present invention: The power system summer load forecasting method disclosed in this invention includes the following steps:

[0055] S1. Obtain historical load data of the target power system; specifically including the following steps:

[0056] The acquired data includes historical meteorological data, historical calendar data, and historical load data;

[0057] Historical meteorological data includes real-time temperature data for the past 24 hours. i The highest temperature in history within a day Historical low temperature The highest temperature on the day before the historical day

[0058] Historical calendar data includes historical day time data T i Historical day of the week data W k ;

[0059] Historical load data includes hourly load data from the day before the historical date. i-24 ;

[0060] Where i represents the i-th moment in a historical day; k represents the k-th day of a week;

[0061] S2. Process the historical load data obtained in step S1 to construct the training dataset and the test dataset; specifically, this includes the following steps:

[0062] The historical load data obtained in step S1 is cleaned and interpolated to obtain a complete set of historical load data.

[0063] Construct the dataset H as H = [H x H y ], where H x The independent variable is a historical dataset and H y For the historical dataset of the dependent variable and H y =[L i ], L i This is the load data for historical days.

[0064] The constructed dataset H is divided into a training dataset and a test dataset according to a set ratio;

[0065] S3. Construct an initial model for summer load forecasting based on the LightGBM model; specifically including the following steps:

[0066] The following formula is used as the objective function of the initial model for summer load forecasting:

[0067]

[0068] In the formula Obj (kk)Let be the objective function value at the k-th iteration; NN is the total number of variable factors; y ii This is the model prediction value in the current iteration; This is the model prediction value at step kk-1; It is a squared loss function and g ii The first intermediate variable and Indicates to Find the first derivative; f kk (x ii ) represents the model fitting function added at step k; h ii It is the second intermediate variable and Indicates to Find the second derivative; Ω(f) kk ) represents the regularization term corresponding to the k-th iteration; constant is the set constant term;

[0069] After simplification, the objective function depends only on each variable x. i The first and second derivatives g and h on the error function, combined with the core algorithm of the LightGBM model, can be used to quickly achieve optimized solutions;

[0070] S4. Use the training dataset constructed in step S2 to train the initial summer load forecasting model constructed in step S3 to obtain the summer load forecasting test model.

[0071] S5. The summer load forecasting test model obtained in step S4 is tested using the test dataset constructed in step S3 to obtain the summer load forecasting model; specifically, the following steps are included:

[0072] The summer load forecast test model obtained in step S4 was tested using the test dataset constructed in step S3.

[0073] The accuracy R of the summer load forecasting test model is calculated using the following formula. 2 :

[0074]

[0075] In the formula N t The number of days for testing; Let j be the predicted power load value at time j; Let j be the actual power load value at time j; The average power load in the test dataset;

[0076] Evaluation of the summer load forecasting test model:

[0077] If the precision value R 2If the value is greater than or equal to the set threshold, it means that the summer load forecast test model meets the set requirements; the corresponding summer load forecast test model will be used as the final summer load forecast model.

[0078] If the precision value R 2 If the value is less than the set threshold, repeat steps S4 to S5 until the accuracy value R of the summer load forecast test model is reached. 2 Greater than or equal to the set threshold;

[0079] S6. Based on the historical load data of the target power system obtained in step S1, construct a future load dataset of the target power system under different scenarios; specifically including the following steps:

[0080] The constructed future dataset F is F = [F x ,F y ], where F x For the independent variable, the future dataset and Among them L' i-24 To predict the time-based load data for the day before, T i 'W' provides the time data for the predicted day. k 'For the weekday data of the forecast date, C i 'This is the real-time temperature data for the 24 hours of the predicted day.' To predict the day's highest temperature, To predict the day's lowest temperature, To predict the highest temperature of the day before; F y For the dependent variable, the future dataset and F y =[L' i ], L' i To predict the power load at time i on day i;

[0081] Align the dates of the hottest days of summer (Sanfu) for each year with the dates of the hottest days of summer (Sanfu) for the target year;

[0082] Construct summer load forecast datasets for different historical years; where the meteorological forecast datasets for different scenarios are represented as C. case The total number of scenes is represented by N. case The weather forecast dataset for the nth scenario is represented as

[0083] Obtain the weather forecast dataset and represent the data for the previous N days as follows: Replace the top N meteorological data points in all meteorological forecast datasets for different years with... Obtain new weather forecast datasets for different scenarios;

[0084] Ultimately, we obtain future load datasets for the target power system under different scenarios;

[0085] S7. Based on the future load dataset constructed in step S6, and using the summer load forecasting model obtained in step S5, complete the future summer load forecasting for the target power system; specifically, this includes the following steps:

[0086] Based on the future load dataset constructed in step S6, the summer load forecasting model obtained in step S5 is used to forecast loads for different year scenarios, and the final load forecast results are obtained. Represented as This represents the load prediction result at time i in the nth scenario;

[0087] The load forecast results for each time step in all scenarios are sorted, and the sorted load forecast results are represented as P. pre The number of load forecast results at each time point is the same as the total number of scenarios;

[0088] Based on the sorted load forecast results P pre The words are processed using the following formula for probabilistic processing:

[0089]

[0090] In the formula The load prediction result at time i with probability q; The sorted load forecast results P pre The m-th load forecast result at time i; The sorted load forecast results P pre The (m+1)th load forecast result at time i; g is the third intermediate variable and g = N case *pm,N case Let p be the total number of scenes, and p be the fourth intermediate variable. q is a probability value ranging from 1 to 99, and m is the fifth intermediate variable. This is the floor symbol;

[0091] Finally, what was obtained The forecast results for the future summer load of the target power system.

[0092] like Figure 2The diagram shows the functional modules of the system of this invention: The system for implementing the summer load forecasting method for power systems disclosed in this invention includes a data acquisition module, a dataset construction module, a prediction model construction module, a model training module, a model testing module, a future data construction module, and a load forecasting module; these modules are connected in series. The data acquisition module acquires historical load data information of the target power system and uploads the data to the dataset construction module. The dataset construction module processes the acquired historical load data information based on the received data to construct training and testing datasets, and uploads the data to the prediction model construction module. The prediction model construction module, based on the received data, performs a prediction model construction based on LightG... The BM model is used to construct an initial summer load forecasting model and upload the data to the model training module. The model training module trains the initial summer load forecasting model using the received data and a training dataset to obtain a summer load forecasting test model, which is then uploaded to the model testing module. The model testing module tests the summer load forecasting test model using the received data and a test dataset to obtain a summer load forecasting model, which is then uploaded to the future data construction module. The future data construction module constructs future load datasets for the target power system under different scenarios based on the received data and historical load data, and uploads this data to the load forecasting module. The load forecasting module, based on the received data and the future load dataset, uses the summer load forecasting model to complete the future summer load forecast for the target power system.

Claims

1. A method for predicting summer load in a power system, comprising the following steps: S1. Obtain historical load data information of the target power system; S2. Process the historical load data information obtained in step S1 to construct the training dataset and the test dataset; S3. Construct an initial model for summer load forecasting based on the LightGBM model; S4. Use the training dataset constructed in step S2 to train the initial summer load forecasting model constructed in step S3 to obtain the summer load forecasting test model; S5. The summer load forecast test model obtained in step S4 is tested using the test dataset constructed in step S3 to obtain the summer load forecast model; S6. Based on the historical load data of the target power system obtained in step S1, construct a future load dataset of the target power system under different scenarios; specifically including the following steps: The constructed future dataset F is ,in For the independent variable, the future dataset and ,in To predict the hourly load data for the day before, For the time data of the predicted day, For the weekday data of the forecast day, To predict the real-time temperature data for the next 24 hours, To predict the day's highest temperature, To predict the day's lowest temperature, To predict the highest temperature of the day before; For the dependent variable, the future dataset and , To predict the power load at time i on day i; Align the dates of the hottest days of summer (Sanfu) for each year with the dates of the hottest days of summer (Sanfu) for the target year; Construct a summer load forecast dataset for different historical years; among which... Weather forecast datasets under different scenarios are represented as follows The total number of scenes is expressed as The weather forecast dataset for the nth scenario is represented as ; Obtain the weather forecast dataset and represent the data for the previous N days as follows: Replace the top N meteorological data points in the meteorological forecast dataset for all different years with... This yields new meteorological forecast datasets for different scenarios; Ultimately, we obtain future load datasets for the target power system under different scenarios; S7. Based on the future load dataset constructed in step S6, and using the summer load forecasting model obtained in step S5, complete the future summer load forecasting for the target power system; specifically, this includes the following steps: Based on the future load dataset constructed in step S6, the summer load forecasting model obtained in step S5 is used to forecast loads for different year scenarios, and the final load forecast results are obtained. Represented as , This represents the load prediction result at time i in the nth scenario; The load forecast results for each time step in all scenarios are sorted, and the sorted load forecast results are represented as follows: The number of load forecast results at each time point is the same as the total number of scenarios; Based on the sorted load forecast results The following formula is used for probabilistic processing: In the formula The load prediction result at time i with probability q; The sorted load forecast results The m-th load forecast result at time i; The sorted load forecast results The (m+1)th load forecast result at time i; It is the third intermediate variable and , The total number of scenes, It is the fourth intermediate variable and , The value is a probability value and ranges from 1 to 99. It is the fifth intermediate variable and , This is the floor symbol; Finally, what was obtained The forecast results for the future summer load of the target power system.

2. The power system summer load forecasting method according to claim 1, characterized in that... Step S1, which involves obtaining historical load data information of the target power system, specifically includes the following steps: The acquired data includes historical meteorological data, historical calendar data, and historical load data; Historical meteorological data includes real-time temperature data for the past 24 hours. The highest temperature in history within a day The lowest temperature in history within a day The highest temperature on the day before the historical day ; Historical calendar data includes historical date times. Historical data on the day of the week ; Historical load data includes hourly load data from the day before the historical date. ; Where i represents the i-th moment in a historical day; k represents the k-th day of a week.

3. The power system summer load forecasting method according to claim 2, characterized in that... Step S2 involves processing the historical load data obtained in step S1 to construct the training dataset and the test dataset, specifically including the following steps: The historical load data obtained in step S1 is cleaned and interpolated to obtain a complete set of historical load data. Construct dataset H as ,in The independent variable is a historical dataset and ; For the historical dataset of the dependent variable and , This is the load data for historical days. The constructed dataset H is divided into a training dataset and a test dataset according to a set ratio.

4. The power system summer load forecasting method according to claim 3, characterized in that... Step S3, which involves constructing an initial model for summer load forecasting based on the LightGBM model, specifically includes the following steps: The following formula is used as the objective function of the initial model for summer load forecasting: In the formula Let be the objective function value at the kkth iteration; NN is the total number of variable factors; This is the model prediction value in the current iteration; This is the model prediction value at step kk-1; It is a squared loss function and ; The first intermediate variable and , Indicates to Find the first derivative; The model fitting function added at step kk; It is the second intermediate variable and , Indicates to Find the second derivative; This refers to the regularization term corresponding to the k-th iteration; This is a constant term that is set.

5. The power system summer load forecasting method according to claim 4, characterized in that... Step S5 involves testing the summer load forecasting test model obtained in step S4 using the test dataset constructed in step S3 to obtain the summer load forecasting model. This process specifically includes the following steps: The summer load forecast test model obtained in step S4 was tested using the test dataset constructed in step S3. The accuracy of the summer load forecasting test model is calculated using the following formula. : In the formula The number of days for testing; Let j be the predicted power load value at time j; Let j be the actual power load value at time j; The average power load in the test dataset; Evaluation of the summer load forecasting test model: If precision value If the value is greater than or equal to the set threshold, it means that the summer load forecast test model meets the set requirements; the corresponding summer load forecast test model will be used as the final summer load forecast model. If precision value If the value is less than the set threshold, repeat steps S4-S5 until the accuracy value of the summer load forecast test model is reached. Greater than or equal to the set threshold.

6. A system for implementing the power system summer load forecasting method according to any one of claims 1 to 5, characterized in that... It includes a data acquisition module, a dataset construction module, a prediction model construction module, a model training module, a model testing module, a future data construction module, and a load forecasting module; these modules are connected in series; the data acquisition module is used to acquire historical load data information of the target power system and upload the data to the dataset construction module. The dataset construction module is used to process the acquired historical load data based on the received data, construct the training dataset and the test dataset, and upload the data to the prediction model construction module; the prediction model construction module is used to construct the initial summer load prediction model based on the LightGBM model based on the received data, and upload the data to the model training module. The model training module is used to train the initial summer load forecasting model using the training dataset based on the received data, obtain the summer load forecasting test model, and upload the data to the model testing module. The model testing module is used to test the summer load forecasting test model using a test dataset based on the received data, obtain the summer load forecasting model, and upload the data to the future data construction module; the future data construction module is used to construct future load datasets of the target power system under different scenarios based on the received data and the historical load data information of the target power system, and upload the data to the load forecasting module. The load forecasting module is used to forecast the future summer load of the target power system based on the received data and the future load dataset, using a summer load forecasting model.

Images