Regional power grid short-term load prediction method and system

Abstract

The invention discloses a regional power grid short-term load prediction method. The method comprises the steps of dividing time periods, obtaining historical load data and meteorological data information of a target regional power grid in different time periods and processing the historical load data and the meteorological data information, and constructing a training data set according to the time periods; constructing a short-term load prediction initial model for each time period, and training to obtain a short-term load prediction model of each time period; acquiring data information of a to-be-predicted day and performing load prediction of each time period of the to-be-predicted day; selecting a historical day most similar to the data information of the day to be predicted, and obtaining load data of each corresponding time period; and in combination with the load prediction result of each time period and the load data of each time period of the corresponding historical day, completing short-term load prediction of the target regional power grid on the to-be-predicted day. The invention also discloses a system for realizing the regional power grid short-term load prediction method. According to the method, time-phased prediction of the short-term load of the regional power grid can be realized, the reliability is higher, and the accuracy is better.

Description

Technical Field

[0001] The invention belongs to the field of electrical automation, and in particular relates to a method and system for predicting short-term load of a regional power grid. 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 of great significance to the power system. The short-term load forecasting of the regional power grid directly affects the operation plan formulation and regulation plan formulation of the regional power grid. Therefore, accurate and reliable short-term load forecasting results can effectively improve the economy, safety and stability of the regional power grid operation. At the same time, in recent years, environmental problems have become more and more serious, so more and more new energy power generation systems have begun to connect to the regional power grid and generate electricity. The randomness and intermittency of the output of the new energy power generation system have brought great challenges to the load forecasting of the regional power grid.

[0004] However, the currently commonly used regional power grid load forecasting scheme basically outputs the load forecast results of the forecasted day as a whole, without considering the load in a specific time period; this situation is no longer applicable to today's refined regional power grid operation strategy. Moreover, the existing regional load forecasting scheme does not take into account the impact of new energy power generation systems on regional power grid loads. Summary of the invention

[0005] One of the purposes of the present invention is to provide a method for short-term load forecasting of a regional power grid with high reliability and good accuracy.

[0006] A second object of the present invention is to provide a system for implementing the regional power grid short-term load forecasting method.

[0007] The method for short-term load forecasting of a regional power grid provided by the present invention comprises the following steps:

[0008] S1. Divide the time period according to the load characteristics of the target area power grid, and obtain the historical load data information and historical meteorological data information of the target area power grid in different time periods;

[0009] S2. Process the data information obtained in step S1 and construct training data sets according to time periods;

[0010] S3. Constructing an initial short-term load forecasting model for each time period, and using the training data set of each time period for training to obtain a short-term load forecasting model for each time period;

[0011] S4. Obtain data information of the day to be predicted, and use the short-term load forecasting model for each time period obtained in step S3 to perform load forecasting for each time period of the day to be predicted;

[0012] S5. Selecting a historical day that is most similar to the data information of the day to be predicted from the historical data, and obtaining the load data of each time period of the historical day;

[0013] S6. Combine the load forecast results of each time period obtained in step S4 and the load data of each time period of the corresponding historical day obtained in step S5 to complete the short-term load forecast of the target area power grid on the day to be predicted.

[0014] Step S1, dividing the time periods according to the load characteristics of the target area power grid, and acquiring the historical load data information and historical meteorological data information of the target area power grid in different time periods, specifically includes the following steps:

[0015] The following rules are used to divide the time periods:

[0016] Low load: 00:00~06:00;

[0017] Morning peak load: 08:00-11:30;

[0018] Waist load: 12:00-16:00;

[0019] Evening peak load: 17:00-23:00;

[0020] The historical load data information includes historical daily power load data, historical daily holiday type data, historical daily week data and historical economic data;

[0021] The historical meteorological data information includes historical daily weather data.

[0022] Step S2 processes the data information obtained in step S1 and constructs training data sets according to time periods, which specifically includes the following steps:

[0023] Encode historical daily weather data with numbers;

[0024] Encode the historical daily weekly data with numbers;

[0025] Historical daily holiday type data, coded with numbers;

[0026] For each divided time period, the correlation coefficient between each weather data and load data, the correlation coefficient between each economic data and load data, the correlation coefficient between each weekly data and load data, and the correlation coefficient between each holiday type data and load data are calculated, and the first several weather data with the largest correlation coefficient, economic data with a correlation coefficient greater than a first set value, weekly data with a correlation coefficient greater than a second set value, and holiday type data with a correlation coefficient greater than a third set value are selected to construct a training data set.

[0027] The correlation coefficient specifically includes the following steps:

[0028] The correlation coefficient is calculated using the following formula:

[0029]

[0030] Where ρ X,Y is the correlation coefficient; cov(X,Y) is the covariance between data X and data Y; X is the characteristic data to be analyzed, including weather data, economic data, week data and holiday type data; Y is the load data; ρ X is the standard deviation of the characteristic data to be analyzed; ρ Y is the standard deviation of the load data.

[0031] The step S3 constructs the initial short-term load forecasting model for each time period, and uses the training data set of each time period for training to obtain the short-term load forecasting model for each time period, which specifically includes the following steps:

[0032] For each time period, the LSTM model is used to build the initial model for short-term load forecasting;

[0033] The training data sets of each time period are used to train the corresponding LSTM model to obtain the short-term load forecasting model of each time period.

[0034] The step S4 of acquiring data information of the day to be predicted and using the short-term load forecasting model of each time period obtained in step S3 to perform load forecasting for each time period of the day to be predicted specifically includes the following steps:

[0035] The weather data information, week data and holiday type data of the day to be predicted are obtained and input into the short-term load prediction model of each time period obtained in step S3 to obtain the short-term load prediction results of each time period of the day to be predicted.

[0036] Step S5 of selecting a historical day that is most similar to the data information of the day to be predicted from the historical data, and obtaining the load data of each time period of the historical day, specifically includes the following steps:

[0037] For the tth day in history, construct the historical weather data feature set D for day t t :

[0038]

[0039] In the formula is the nth weather data of the tith day; is the nth weather data of the tth day;

[0040] The weather data D of the day to be predicted is calculated using the following formula: o and historical weather data feature set D t The similarity value S(D o ,D t ):

[0041]

[0042] Where d euc (D o ,D t ) is D o and D t The Euclidean distance between o ;D t ) is D o and D t The mutual information between

[0043] Select the similarity value S(D o ,D t ) The largest historical day t is taken as the historical day with the data information most similar to the day to be predicted;

[0044] Get the load per unit value curve of the historical day.

[0045] Step S6 combines the load forecast results of each time period obtained in step S4 with the load data of each time period of the corresponding historical day obtained in step S5 to complete the short-term load forecast of the target area power grid on the forecast day, which specifically includes the following steps:

[0046] The short-term load forecast value of the target area power grid on the forecast day is calculated using the following formula:

[0047]

[0048] In the formula The load forecast value of the target area power grid on the forecast day; is the corresponding historical day t obtained in step S5 i The load of the moment; t iThe time period corresponding to the time is the load of the corresponding historical day obtained in step S5, and the value of T is valley, early peak, mid-load or late peak; is t obtained in step S4 i The short-term load forecast result for the time period corresponding to the time on the day to be predicted.

[0049] The present invention also provides a system for implementing the regional power grid short-term load forecasting method, comprising a data acquisition module, a data processing module, a model training module, a preliminary forecasting module, a similarity selection module and a load forecasting module; the data acquisition module, the data processing module, the model training module, the preliminary forecasting module, the similarity selection module and the load forecasting module are connected in series in sequence; the data acquisition module is used to divide the time period according to the load characteristics of the target regional power grid, and obtain the historical load data information and historical meteorological data information of the target regional power grid in each time period, and upload the data information to the data processing module; the data processing module is used to process the data information according to the received data information, and construct a training data set according to the time period, and upload the data information to the model training module; the model training module is used to construct a preliminary forecasting module for short-term load for each time period according to the received data information The initial model is used, and the training data sets of each time period are used for training to obtain the short-term load forecasting model of each time period, and the data information is uploaded to the preliminary forecasting module; the preliminary forecasting module is used to obtain the data information of the day to be predicted according to the received data information, and use the short-term load forecasting model of each time period to perform load forecasting for each time period of the day to be predicted, and upload the data information to the similarity selection module; the similarity selection module is used to select the historical day that is most similar to the data information of the day to be predicted in the historical data according to the received data information, and obtain the load data of each time period of the historical day, and upload the data information to the load forecasting module; the load forecasting module is used to complete the short-term load forecasting of the target area power grid on the day to be predicted according to the received data information, combined with the load forecasting results of each time period and the load data of each time period of the corresponding historical day.

[0050] The method and system for regional power grid short-term load forecasting provided by the present invention classifies loads by time period, and performs preliminary forecasts of the short-term load of the regional power grid by time period in combination with historical data and meteorological data, and finally corrects the preliminary forecast results by time period based on the load data of the most similar day; therefore, the present invention can not only realize the time period forecast of the short-term load of the regional power grid, but also has higher reliability and better accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 The figure is a schematic diagram of the method flow of the present invention.

[0052] Figure 2 Schematic diagram of the prediction effect of the method embodiment of the present invention.

[0053] Figure 3 Schematic diagram of the functional modules of the system of the present invention. DETAILED DESCRIPTION

[0054] like Figure 1 The method flow chart of the method of the present invention is shown as follows: The regional power grid short-term load forecasting method disclosed by the present invention comprises the following steps:

[0055] S1. Divide the time period according to the load characteristics of the target area power grid, and obtain the historical load data information and historical meteorological data information of the target area power grid in different time periods; specifically, the steps include:

[0056] The following rules are used to divide the time periods:

[0057] Low load: 00:00~06:00;

[0058] Morning peak load: 08:00-11:30;

[0059] Waist load: 12:00-16:00;

[0060] Evening peak load: 17:00-23:00;

[0061] The historical load data information includes historical daily power load data, historical daily holiday type data, historical daily week data and historical economic data;

[0062] The historical meteorological data information includes historical daily weather data;

[0063] In specific implementation, the historical period is preferably at least 1 year;

[0064] S2. Process the data information obtained in step S1 and construct training data sets according to time periods; specifically, the steps include:

[0065] Encode historical daily weather data with numbers;

[0066] In specific implementation, the encoding method shown in Table 1 may be used:

[0067] Table 1 Weather code diagram

[0068] weather coding weather coding Blizzard 0 heavy rain 8 heavy snow 1 moderate rain 9 Moderate Snow 2 light rain 10 Light Snow 3 shower 11 Sleet 4 Negative 12 Freezing Rain 5 partly cloudy 13 Heavy rain 6 fog 14 rainstorm 7 clear 15

[0069] Encode the historical daily weekly data with numbers;

[0070] In specific implementation, the encoding method shown in Table 2 may be used:

[0071] Table 2 Weekly data coding diagram

[0072] Sunday Type coding Sunday Type coding on Monday 0 Tuesday 1 Wednesday 2 Thursday 3 Friday 4 Saturday 5 Sunday 6

[0073] Historical daily holiday type data, coded with numbers;

[0074] In specific implementation, the encoding method shown in Table 3 may be used:

[0075] Table 3 Holiday type coding diagram

[0076] Holiday Type coding Holiday Type coding New Year 0 Spring Festival 1 Qingming 2 May Day 3 Dragon Boat Festival 4 Mid-Autumn Festival 5 National Day 6

[0077] For each divided time period, calculate the correlation coefficient between each weather data and load data, the correlation coefficient between each economic data and load data, the correlation coefficient between each weekly data and load data, and the correlation coefficient between each holiday type data and load data, and select the first several (preferably the first three) weather data with the largest correlation coefficient, the economic data with a correlation coefficient greater than a first set value (preferably 0.2), the weekly data with a correlation coefficient greater than a second set value (preferably 0.2), and the holiday type data with a correlation coefficient greater than a third set value (preferably 0.2) to construct a training data set;

[0078] In specific implementation, the correlation coefficient is calculated using the following formula:

[0079]

[0080] Where ρ X,Y is the correlation coefficient; cov(X,Y) is the covariance between data X and data Y; X is the characteristic data to be analyzed, including weather data, economic data, week data and holiday type data; Y is the load data; ρ X is the standard deviation of the characteristic data to be analyzed; ρ Y is the standard deviation of the load data;

[0081] S3. Constructing an initial short-term load forecasting model for each time period, and using the training data set of each time period for training to obtain a short-term load forecasting model for each time period; specifically comprising the following steps:

[0082] For each time period, the LSTM model is used to build the initial model for short-term load forecasting;

[0083] The training data sets of each time period are used to train the corresponding LSTM model to obtain the short-term load forecasting model of each time period;

[0084] S4. Obtain data information of the day to be predicted, and use the short-term load forecasting model of each time period obtained in step S3 to perform load forecasting for each time period of the day to be predicted; specifically comprising the following steps:

[0085] Obtain weather data information, week data and holiday type data of the day to be predicted, and input them into the short-term load forecasting model of each time period obtained in step S3 to obtain the short-term load forecasting results of each time period of the day to be predicted;

[0086] S5. Selecting a historical day that is most similar to the data information of the day to be predicted from the historical data, and obtaining the load data of each time period of the historical day; specifically comprising the following steps:

[0087] For the tth day in history, construct the historical weather data feature set D for day t t :

[0088]

[0089] In the formula is the nth weather data of the tith day; is the nth weather data of the tth day; in specific implementation, the value of i is preferably 6, corresponding to a week of weather data;

[0090] The weather data D of the day to be predicted is calculated using the following formula: o and historical weather data feature set D t The similarity value S(D o ,D t ):

[0091]

[0092] Where d euc (D o ,D t ) is D o and D t The Euclidean distance between o ;D t ) is D o and D t The mutual information between

[0093] Select the similarity value S(D o ,D t ) The largest historical day t is taken as the historical day with the data information most similar to the day to be predicted;

[0094] Obtain the load per unit value curve of the historical day; in specific implementation, it is recommended that the screening range of historical days be at least 2 years;

[0095] S6. Combining the load forecast results of each time period obtained in step S4 and the load data of each time period of the corresponding historical day obtained in step S5, the short-term load forecast of the target regional power grid on the day to be predicted is completed; specifically, the following steps are included:

[0096] The short-term load forecast value of the target area power grid on the forecast day is calculated using the following formula:

[0097]

[0098] In the formula The load forecast value of the target area power grid on the forecast day; is the corresponding historical day t obtained in step S5 i The load of the moment; t i The time period corresponding to the time is the load of the corresponding historical day obtained in step S5, and the value of T is valley, early peak, mid-load or late peak; is t obtained in step S4 i The short-term load forecast result for the time period corresponding to the time on the day to be predicted.

[0099] The method of the present invention is used to predict the short-term load of a certain province in four seasons. The prediction results are shown in Figure 2. Figure 2 (a) to (d) correspond to the short-term load forecast results of spring, summer, autumn and winter respectively; Figure 2 It can be seen that the method of the present invention has good prediction accuracy and reliability.

[0100] like Figure 3The functional module schematic diagram of the system of the present invention is shown as follows: the system for implementing the short-term load forecasting method of the regional power grid disclosed in the present invention comprises a data acquisition module, a data processing module, a model training module, a preliminary forecasting module, a similarity selection module and a load forecasting module; the data acquisition module, the data processing module, the model training module, the preliminary forecasting module, the similarity selection module and the load forecasting module are connected in series in sequence; the data acquisition module is used to divide the time period according to the load characteristics of the target regional power grid, and obtain the historical load data information and historical meteorological data information of the target regional power grid in each time period, and upload the data information to the data processing module; the data processing module is used to process the data information according to the received data information, and construct a training data set according to the time period, and upload the data information to the model training module; the model training module is used to construct a training data set for each time period according to the received data information An initial model for short-term load forecasting is established, and training data sets of various time periods are used for training to obtain short-term load forecasting models for various time periods, and the data information is uploaded to the preliminary forecasting module; the preliminary forecasting module is used to obtain data information of the day to be predicted based on the received data information, and use the short-term load forecasting models of various time periods to perform load forecasts for various time periods of the day to be predicted, and upload the data information to the similarity selection module; the similarity selection module is used to select the historical day that is most similar to the data information of the day to be predicted in the historical data based on the received data information, and obtain the load data for various time periods of the historical day, and upload the data information to the load forecasting module; the load forecasting module is used to complete the short-term load forecast of the target area power grid on the day to be predicted based on the received data information, combined with the load forecast results for each time period and the load data for each time period of the corresponding historical day.

Claims

1. A method for short-term load forecasting of a regional power grid, comprising the following steps: S1. Divide the time period according to the load characteristics of the target area power grid, and obtain the historical load data information and historical meteorological data information of the target area power grid in different time periods; S2. Process the data information obtained in step S1 and construct training data sets according to time periods; S3. Constructing an initial short-term load forecasting model for each time period, and using the training data set of each time period for training to obtain a short-term load forecasting model for each time period; S4. Obtain data information of the day to be predicted, and use the short-term load forecasting model for each time period obtained in step S3 to perform load forecasting for each time period of the day to be predicted; S5. Selecting a historical day that is most similar to the data information of the day to be predicted from the historical data, and obtaining the load data of each time period of the historical day; S6. Combine the load forecast results of each time period obtained in step S4 and the load data of each time period of the corresponding historical day obtained in step S5 to complete the short-term load forecast of the target area power grid on the day to be predicted.

2. The method for regional power grid short-term load forecasting according to claim 1, characterized in that Step S1, dividing the time periods according to the load characteristics of the target area power grid, and acquiring the historical load data information and historical meteorological data information of the target area power grid in different time periods, specifically includes the following steps: The following rules are used to divide the time periods: Low load: 00:00~06:00; Morning peak load: 08:00-11:30; Waist load: 12:00-16:00; Evening peak load: 17:00-23:00; The historical load data information includes historical daily power load data, historical daily holiday type data, historical daily week data and historical economic data; The historical meteorological data information includes historical daily weather data.

3. The method for short-term load forecasting of a regional power grid according to claim 2, characterized in that Step S2 processes the data information obtained in step S1 and constructs training data sets according to time periods, which specifically includes the following steps: Encode historical daily weather data with numbers; Encode the historical daily weekly data with numbers; Historical daily holiday type data, coded with numbers; For each divided time period, the correlation coefficient between each weather data and load data, the correlation coefficient between each economic data and load data, the correlation coefficient between each weekly data and load data, and the correlation coefficient between each holiday type data and load data are calculated, and the first several weather data with the largest correlation coefficient, economic data with a correlation coefficient greater than a first set value, weekly data with a correlation coefficient greater than a second set value, and holiday type data with a correlation coefficient greater than a third set value are selected to construct a training data set.

4. The method for regional power grid short-term load forecasting according to claim 3 is characterized in that The correlation coefficient specifically includes the following steps: The correlation coefficient is calculated using the following formula: Where ρ X,Y is the correlation coefficient; cov(X,Y) is the covariance between data X and data Y; X is the characteristic data to be analyzed, including weather data, economic data, week data and holiday type data; Y is the load data; ρ X is the standard deviation of the characteristic data to be analyzed; ρ Y is the standard deviation of the load data.

5. The method for regional power grid short-term load forecasting according to claim 4, characterized in that The step S3 constructs the initial short-term load forecasting model for each time period, and uses the training data set of each time period for training to obtain the short-term load forecasting model for each time period, which specifically includes the following steps: For each time period, the LSTM model is used to build the initial model for short-term load forecasting; The training data sets of each time period are used to train the corresponding LSTM model to obtain the short-term load forecasting model of each time period.

6. The method for regional power grid short-term load forecasting according to claim 5, characterized in that The step S4 of acquiring data information of the day to be predicted and using the short-term load forecasting model of each time period obtained in step S3 to perform load forecasting for each time period of the day to be predicted specifically includes the following steps: The weather data information, week data and holiday type data of the day to be predicted are obtained and input into the short-term load prediction model of each time period obtained in step S3 to obtain the short-term load prediction results of each time period of the day to be predicted.

7. The method for regional power grid short-term load forecasting according to claim 6, characterized in that Step S5 of selecting a historical day that is most similar to the data information of the day to be predicted from the historical data, and obtaining the load data of each time period of the historical day, specifically includes the following steps: For the tth day in history, construct the historical weather data feature set D for day t t : In the formula is the nth weather data of the tith day; is the nth weather data of the tth day; The weather data D of the day to be predicted is calculated using the following formula: o and historical weather data feature set D t The similarity value S(D o ,D t ): Where d euc (D o ,D t ) is D o and D t The Euclidean distance between o ;D t ) is D o and D t The mutual information between Select the similarity value S(D o ,D t ) The largest historical day t is taken as the historical day with the data information most similar to the day to be predicted; Get the load per unit value curve of the historical day.

8. The method for regional power grid short-term load forecasting according to claim 7, characterized in that Step S6 combines the load forecast results of each time period obtained in step S4 with the load data of each time period of the corresponding historical day obtained in step S5 to complete the short-term load forecast of the target area power grid on the forecast day, which specifically includes the following steps: The short-term load forecast value of the target area power grid on the forecast day is calculated using the following formula: In the formula The load forecast value of the target area power grid on the forecast day; is the corresponding historical day t obtained in step S5 i The load of the moment; t i The time period corresponding to the time is the load of the corresponding historical day obtained in step S5, and the value of T is valley, early peak, mid-load or late peak; is t obtained in step S4 i The short-term load forecast result for the time period corresponding to the time on the day to be predicted.

9. A system for implementing the method for regional power grid short-term load forecasting according to any one of claims 1 to 8, characterized in that It includes a data acquisition module, a data processing module, a model training module, a preliminary prediction module, a similar selection module and a load prediction module; the data acquisition module, the data processing module, the model training module, the preliminary prediction module, the similar selection module and the load prediction module are connected in series in sequence; the data acquisition module is used to divide the time period according to the load characteristics of the target area power grid, and obtain the historical load data information and historical meteorological data information of the target area power grid in different time periods, and upload the data information to the data processing module; The data processing module is used to process the data information according to the received data information, construct training data sets according to time periods, and upload the data information to the model training module; The model training module is used to construct the initial model of short-term load forecasting for each time period according to the received data information, and use the training data set of each time period for training to obtain the short-term load forecasting model of each time period, and upload the data information to the preliminary forecasting module; the preliminary forecasting module is used to obtain the data information of the day to be predicted according to the received data information, and use the short-term load forecasting model of each time period to perform load forecasting for each time period of the day to be predicted, and upload the data information to the similarity selection module; The similarity selection module is used to select the historical day that is most similar to the data information of the day to be predicted in the historical data according to the received data information, obtain the load data of each time period of the historical day, and upload the data information to the load prediction module; The load forecasting module is used to complete the short-term load forecast of the target area power grid on the forecasted day based on the received data information, combined with the load forecast results of each time period and the load data of each time period of the corresponding historical day.

Images