Power planning method and device based on hourly predicted load correction, terminal equipment and storage medium
By obtaining historical data and using load prediction models for monthly and hourly load prediction, the problem of difficult to consider power imbalance in the current technology of monthly particle size prediction is solved, and the hourly particle size load prediction of the new power system is achieved, which satisfies the power System planning requirements.
Patent Information
- Application Number
- CN202411156953.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-22
- Publication Date
- 2024-12-03
AI Technical Summary
The load prediction based on monthly particle size in the prior art is difficult to take into account the power imbalance problem for many consecutive days in advance, and cannot meet the planning needs of new power systems.
By obtaining historical load data, historical weather data and calendar data, using a preset load prediction model to predict monthly load and hourly loads, and correcting the hourly load prediction results to obtain hourly predicted loads to meet the hourly particle size of the power system Planning requirements.
Accurate prediction of hourly loads within a continuous multi-day period is achieved, which meets the planning needs of the power system on hourly granularity and improves the reliability and accuracy of predictions.
Smart Images

Figure CN119067474A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of power systems, and in particular to a power planning method, device, terminal equipment and storage medium based on hourly forecast load correction. Background Art
[0002] With the development of social economy and technology, electric energy has become one of the indispensable energy sources in people's production and life. Therefore, ensuring the stable supply of electric energy is crucial to the power system. In the power system, planning the power system based on load forecasting is an important technology to ensure power balance. Load forecasting can be divided into short-term load forecasting (one day to one week), medium-term load forecasting (one month to twelve months) and long-term load forecasting (more than one year) according to the forecast period. In traditional power grids, the existing forecast load is mainly for the monthly average load and monthly maximum load for the next month or the next twelve months. However, in the new power system, due to the increasing proportion of renewable energy year by year and the increasing uncertainty on the power generation side of the power system, the forecast load based on monthly granularity is difficult to take into account the power imbalance problem for consecutive days in advance, and cannot meet the planning needs of the power system. Summary of the invention
[0003] The embodiments of the present invention provide a power planning method, apparatus, terminal device and storage medium based on hourly forecast load correction, which can effectively solve the problem that the prior art forecast load based on monthly granularity is difficult to consider the power imbalance problem for consecutive days in advance and cannot meet the power system planning needs.
[0004] An embodiment of the present invention provides a power planning method based on hourly forecast load correction, comprising:
[0005] Obtain historical load data, historical weather data, historical temperature data of the area to be planned, and calendar data of the time period to be planned;
[0006] Inputting the historical load data, the historical temperature data and the calendar data into a preset first load forecasting model to perform monthly load forecasting, and obtaining a first load forecasting result for representing monthly granularity;
[0007] Inputting the historical load data, the historical weather data and the calendar data into a preset second load forecasting model to perform hourly load forecasting, and obtaining a second load forecasting result representing hourly granularity;
[0008] According to the first load forecast result and the second load forecast result, the second load forecast result is corrected to obtain an hourly forecast load;
[0009] According to the hourly forecast load, power planning is performed for the planned area.
[0010] Furthermore, the training of the first load forecasting model includes:
[0011] According to the historical load data, the historical temperature data and the calendar data, respectively calculating a first average load, a first maximum load, a monthly average temperature, a monthly maximum temperature, monthly historical load data and monthly historical temperature data for representing monthly granularity;
[0012] According to the first average load, the first maximum load, the monthly average temperature, the monthly maximum temperature, the monthly historical load data, the monthly historical temperature data and the calendar data, a first test set for representing the data within the preset time period and a first training set for representing the data within the preset time period are constructed; wherein the monthly historical load data in the first test set is used as label data;
[0013] Taking the first training set as input and the first predicted load, the first average load and the first maximum load used to represent the monthly granularity load within the planning time period as output, the first load forecasting model to be trained is trained until the error between the first predicted load and the monthly historical load data in the first test set is less than the preset first prediction error, and the first load forecasting result is obtained.
[0014] Furthermore, the training of the second load forecasting model includes:
[0015] According to the historical load data, the historical weather data and the calendar data, a second test set for representing hourly granularity within a preset time period and a second training set for representing hourly granularity within a historical time period that is not within the preset time period are constructed; wherein the historical load data in the second test set is used as label data;
[0016] Taking the second training set as input and the second predicted load used to represent the hourly granularity load in the planning time period as output, the second load prediction model to be trained is trained until the error between the second predicted load and the historical load data in the second test set is less than the preset second prediction error, and a second load prediction result is obtained.
[0017] Further, according to the first load forecast result and the second load forecast result, the second load forecast result is corrected to obtain the hourly forecast load, including:
[0018] Taking the first load forecast result and the second load forecast result as input and the revised second load forecast result as output, repeatedly performing the revision operation until the preset loss function converges, obtaining the final second load forecast result, and using the final second load forecast result as the hourly load forecast;
[0019] The correction operation includes:
[0020] The current alignment coefficient is calculated based on the first average load, the first maximum load and the current second load prediction result; the current second load prediction result of the initial correction operation is the second load prediction result
[0021] Aligning the current second load forecast result according to the first load forecast result, the current second load forecast result, the first average load and the current alignment coefficient to obtain an aligned second load forecast result;
[0022] According to a preset load sliding length, sliding adjustment is performed on the aligned second load prediction result according to the load sliding length to obtain an updated second load prediction result;
[0023] The current alignment coefficient is updated according to the updated second load prediction result.
[0024] Further, according to the first load forecast result, the current second load forecast result, the first average load and the current alignment coefficient, the current second load forecast result is aligned to obtain the aligned second load forecast result, including:
[0025] According to the current second load forecast result, a second average load of the current second load forecast result is calculated;
[0026] The translation coefficient is calculated by dividing the second average load and the first average load;
[0027] According to the current second load prediction result and the translation coefficient, an initial translation load prediction result is obtained;
[0028] Alignment is performed according to the initial translation load prediction result and the current alignment coefficient to obtain an aligned second load prediction result.
[0029] Furthermore, according to the hourly predicted load, power planning is performed for the planned area, including:
[0030] Counting the number of time periods without output within a preset time length according to the hourly forecast load;
[0031] According to the number of time periods and the hourly forecast load, the new energy output of the area to be planned within the time period to be planned is calculated at a preset confidence level;
[0032] The total installed capacity of the planned area within the planned time period is planned according to the new energy output and the hourly predicted load.
[0033] As an improvement of the above solution, another embodiment of the present invention provides a power planning device based on hourly predicted load correction, including:
[0034] A data acquisition module is used to acquire historical load data, historical weather data, historical temperature data of the area to be planned, and calendar data of the time period to be planned; wherein the historical time periods corresponding to the historical load data, the historical weather data, and the historical temperature data are all consistent; the historical load data is hourly data; and the length of the time period to be planned is consistent with that of the historical time period;
[0035] A first load forecasting module, used for inputting the historical load data, the historical temperature data and the calendar data into a preset first load forecasting model to perform monthly load forecasting, and obtain a first load forecasting result for representing a monthly granularity;
[0036] A second load forecasting module, used for inputting the historical load data, the historical weather data and the calendar data into a preset second load forecasting model to perform hourly load forecasting, and obtain a second load forecasting result representing an hourly granularity;
[0037] A load correction module, configured to correct the second load forecast result according to the first load forecast result and the second load forecast result to obtain an hourly forecast load;
[0038] The power planning module is used to carry out power planning for the planned area according to the hourly predicted load.
[0039] Furthermore, the power planning module is used to carry out power planning for the planned area according to the hourly predicted load;
[0040] The performing power planning for the area to be planned within the planning time period according to the hourly predicted load includes:
[0041] Counting the number of time periods without output within a preset time length according to the hourly forecast load;
[0042] According to the number of time periods and the hourly forecast load, the new energy output of the area to be planned within the time period to be planned is calculated at a preset confidence level;
[0043] The total installed capacity of the planned area within the planned time period is planned according to the new energy output and the hourly predicted load.
[0044] Another embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, it implements a power planning method based on hourly predicted load correction as described in the above embodiment.
[0045] Another embodiment of the present invention provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the power planning method based on hourly predicted load correction described in the above embodiment.
[0046] By implementing the present invention, at least the following beneficial effects are achieved:
[0047] The present invention provides an electric power planning method, apparatus, terminal equipment and storage medium based on hourly predicted load correction. The method can obtain historical load data, historical weather data, historical temperature data and calendar data of a to-be-planned area and a to-be-planned time period; input the historical load data, the historical temperature data and the calendar data into a preset first load prediction model to perform monthly load prediction, and obtain a first load prediction result for representing monthly granularity; input the historical load data, the historical weather data and the calendar data into a preset second load prediction model to perform hourly load prediction, and obtain a second load prediction result for representing hourly granularity; according to the first load prediction result and the second load prediction result, correct the second load prediction result to obtain an hourly predicted load; and according to the hourly predicted load, perform electric power planning for the to-be-planned area. Through historical load data, historical weather data, historical temperature data and calendar data, the first load forecast result and the second load forecast result are obtained, and then the first load forecast result is used to correct the second load forecast result to make the second load forecast result with hourly granularity more accurate. Compared with the monthly granularity forecast result, the hourly granularity forecast result can take into account the power imbalance problem for consecutive days, that is, the second load forecast result takes into account the load of each hour for consecutive days to meet the planning requirements of the power system at the hourly granularity. BRIEF DESCRIPTION OF THE DRAWINGS
[0048] Figure 1 It is a flow chart of a power planning method based on hourly forecast load correction provided by an embodiment of the present invention;
[0049] Figure 2 It is a structural schematic diagram of a power planning device based on hourly forecast load correction provided by an embodiment of the present invention;
[0050] Figure 3It is another flow chart of a power planning method based on hourly forecast load correction provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0051] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0052] See also Figure 1 , is a flow chart of a power planning method based on hourly forecast load correction provided by an embodiment of the present invention, comprising:
[0053] S1. Obtain historical load data, historical weather data, historical temperature data of the area to be planned, and calendar data of the time period to be planned;
[0054] S2, inputting the historical load data, the historical temperature data and the calendar data into a preset first load forecasting model to perform monthly load forecasting, and obtaining a first load forecasting result for representing monthly granularity;
[0055] S3, inputting the historical load data, the historical weather data and the calendar data into a preset second load forecasting model to perform hourly load forecasting, and obtaining a second load forecasting result representing hourly granularity;
[0056] S4. According to the first load forecast result and the second load forecast result, the second load forecast result is corrected to obtain an hourly forecast load;
[0057] S5. Perform power planning for the area to be planned based on the hourly predicted load.
[0058] Specifically, historical weather data includes humidity, wind speed and rainfall; calendar data for the time period to be planned includes months, weeks and days. The historical time periods corresponding to the historical load data, the historical weather data and the historical temperature data are all consistent, ensuring the correlation between the data, and load forecasting can be performed in the same time dimension. The length of the time period to be planned is consistent with that of the historical time period, which not only improves data comparability and model accuracy, but also enhances forecast reliability. The historical load data, historical weather data and historical temperature data are all hourly data.
[0059] In a preferred embodiment of the present invention, the historical load data, historical weather data, historical temperature data and calendar data of the area to be planned are first obtained; wherein the historical time periods corresponding to the historical load data, the historical weather data and the historical temperature data are all consistent; the historical load data, the historical weather data and the historical temperature data are all hourly data; the length of the time period to be planned is consistent with that of the historical time period; then the historical load data, the historical temperature data and the calendar data are input into a preset first load forecasting model for monthly load forecasting to obtain a first load forecasting result for representing monthly granularity; and the historical load data, the historical weather data and the calendar data are input into a preset second load forecasting model for hourly load forecasting to obtain a second load forecasting result for representing hourly granularity; then, according to the first load forecasting result and the second load forecasting result, the second load forecasting result is corrected to obtain the hourly forecast load; finally, according to the hourly forecast load, the power planning is performed for the area to be planned. The first load forecasting result uses historical temperature data because temperature is one of the main factors affecting long-term power demand. For example, seasonal changes lead to temperature changes that affect heating and cooling loads, and seasonal and monthly trend changes need to be captured. Historical temperature data can effectively reflect such changes. The second load forecast result uses historical weather data. Changes in weather data in a short period of time have a greater impact on hourly load fluctuations, so it is necessary to pay attention to short-term meteorological changes.
[0060] Preferably, the training of the first load forecasting model includes:
[0061] According to the historical load data, the historical temperature data and the calendar data, respectively calculating a first average load, a first maximum load, a monthly average temperature, a monthly maximum temperature, monthly historical load data and monthly historical temperature data for representing monthly granularity;
[0062] According to the first average load, the first maximum load, the monthly average temperature, the monthly maximum temperature, the monthly historical load data, the monthly historical temperature data and the calendar data, a first test set for representing the data within the preset time period and a first training set for representing the data within the preset time period are constructed; wherein the monthly historical load data in the first test set is used as label data;
[0063] Taking the first training set as input and the first predicted load, the first average load and the first maximum load used to represent the monthly granularity load within the planning time period as output, the first load forecasting model to be trained is trained until the error between the first predicted load and the monthly historical load data in the first test set is less than the preset first prediction error, and the first load forecasting result is obtained.
[0064] Specifically, because the historical load data is hourly granular, the monthly historical load data must be calculated first. The historical temperature data is also hourly data, so the monthly historical temperature data must be calculated first, and then the monthly average temperature and the monthly maximum temperature are obtained. The first average load (monthly average load) and the first maximum load (the largest monthly load in the monthly historical load data) are calculated. Then, according to the first average load, the first maximum load, the monthly average temperature, the monthly maximum temperature, the monthly historical load data, the monthly historical temperature data and the calendar data, a first test set for representing the preset time period length and a first training set for representing the historical time period not within the preset time period length are constructed; wherein the monthly historical load data in the first test set is used as label data; the first training set is used as input, and the first predicted load, the first average load and the first maximum load used to represent the monthly granular load in the planned time period are used as output, and the first load forecasting model to be trained is trained until the error between the first predicted load and the monthly historical load data in the first test set is less than the preset first prediction error, and the first load forecasting result is obtained.
[0065] In a preferred embodiment of the present invention, a) collect hourly historical load data L of the area to be planned within a historical time period T. real = {l real,t}, t∈T, the corresponding historical temperature data is Temp={temp t}, t∈T, calendar data S={s t}, t∈T can be directly obtained through t. b) The historical load data is averaged on a monthly basis and organized into L ave = {l ave ,m},m∈M,M is the month in the time period T; calculate the maximum value L of the historical load data according to the monthly max = {l max,m}, m∈M; calculate the monthly average Temp of historical temperature data ave ={temp ave,m}, m∈M; calculate the maximum value Temp of the historical temperature data according to the monthly max ={temp max,m}, m∈M, the calendar data only retains the month information S={s m}, m∈M. c) Construct a monthly training dataset D2 to predict the average monthly load for the next twelve months, where each sample is {l ave ,temp ave ,s,r}, l, temp, s are the input features of the first load forecasting model. l is a vector, each of which is the first average load, l ave =(l ave,m-1 ,lave,m-2 ,l ave,m-3 ,......), m refers to the time of the first month to be predicted; temp ave is a vector, each of which is the temperature data of a historical month, temp ave =(temp ave,m+11 ,...,temp ave,m ,temp ave,m-1 ,temp ave,m-2 ,...); s is a vector, each item of which is the calendar time information of the month to be predicted, that is, calendar data, s = (s m+11 ,s m+10 ,...,s m ), r is the true value label, which is a vector, each of which is the monthly historical load data, r = (l ave,m+11 ,l ave,m+10 ,...,l ave,m ). Divide the monthly training data set D2 into the first training set D 2,train With the first test set D 2,test , D 2,test is the last 12 samples in the data set, i.e. the historical load data of the previous 12 months closest to the current moment. d) Use D 2,train Train the first load forecasting model and set D 2,train The characteristic part ave ,temp ave ,s} as the input of the model, D train The label part r is used as the prediction target of the model to minimize the prediction load. The root mean square error with label r is the optimization target, and the first load forecasting model is trained. The root mean square error formula is: Where i is the i-th training sample in the first training set, n2 is D 2,train e) Use D 2,test To evaluate the accuracy of the first load forecasting model, D 2,test The characteristic part ave ,temp ave ,s} as the input of the model to get the predicted value Calculate predicted values When the error between the first predicted load and the monthly historical load data in the first test set is less than the preset first prediction error, the first load prediction result is obtained. The first load prediction result includes the predicted maximum monthly load and average monthly load. The mean absolute percentage error (MAPE) is usually used to evaluate the accuracy of the model. The formula is as follows: Where i is the i-th test sample in the first test set, n2 is D 2,testThe first load forecasting model to be trained can use any machine learning or statistical model as a forecasting model.
[0066] Preferably, the training of the second load forecasting model includes:
[0067] According to the historical load data, the historical weather data and the calendar data, a second test set for representing hourly granularity within a preset time period and a second training set for representing hourly granularity within a historical time period that is not within the preset time period are constructed; wherein the historical load data in the second test set is used as label data;
[0068] Taking the second training set as input and the second predicted load used to represent the hourly granularity load in the planning time period as output, the second load prediction model to be trained is trained until the error between the second predicted load and the historical load data in the second test set is less than the preset second prediction error, and a second load prediction result is obtained.
[0069] In a preferred embodiment of the present invention, the second load forecast result includes the load for 8760 consecutive hours in the future year. a) Collect the hourly load historical data L of the area to be planned in the historical time period T real = {l real,t}, t∈T, the corresponding historical weather data is W={w t}, t∈T, calendar data S={s t}, t∈T can be directly obtained through t; in all the collected data, if there is a leap year, the data on February 29th is removed. b) Construct an hourly training dataset D, where each sample is {l,w,s,r t}, l, w, s are the input features of the second load forecasting model. l is a vector, each of which is the load at the same time of the previous year for the hour to be forecasted, l = (l t-8760 ,l t-17520 ,l t-26280 ,......), t refers to the time of the last hour to be predicted in each sample; w is a vector, each item of which is the historical weather data of the hour to be predicted and the same time in previous years. The historical weather data are represented by superscripts a, b...n. s is a vector, each of which is the calendar data of the hour to be predicted, including month m, day of the week (Monday to Sunday) d, hour of the day (1 to 24 hours) h, s = (m t ,d t ,h t );r t is the label, which is the historical real load l t, which is the historical hourly load, is a scalar. The hourly training data set D is divided into the second training set D train With the second test set D test , D test There are 8760 samples in the last year of the second test set. c) Using D train Train the second load forecasting model and set D train The feature part {l,w,s} is used as the input of the model, D train Middle label part r t As the prediction target of the model; to minimize the prediction load With label r t The root mean square error is taken as the optimization target, and the second load forecasting model is trained. The root mean square error formula is: Where i is the i-th training sample in the second training set, and n is D train d) Use D test To evaluate the accuracy of the second load forecasting model, D test The feature part {l,w,s} is used as the input of the model to get the predicted value r t ; Calculate the predicted value r t and the true load label r t When the error between the second predicted load and the historical load data in the second test set is less than the preset second prediction error, the second load prediction result is obtained. The mean absolute percentage error (MAPE) is usually used to evaluate the accuracy of the model, and the formula is as follows: Where i is the i-th test sample in the second test set, and n is D test The total number of samples in the above second load prediction model to be trained can use any machine learning method as a prediction model. The "second load prediction model" proposed in this embodiment can be established using any machine learning model, and the "first load prediction model" used can be established using any statistical model or machine learning model. Therefore, an appropriate model can be selected for prediction according to the load characteristics of the area to be planned, showing a high degree of applicability and flexibility.
[0070] Specifically, according to the first load forecast result and the second load forecast result, the second load forecast result is corrected to obtain the hourly forecast load, including:
[0071] Taking the first load forecast result and the second load forecast result as input and the revised second load forecast result as output, repeatedly performing the revision operation until the preset loss function converges, obtaining the final second load forecast result, and using the final second load forecast result as the hourly load forecast;
[0072] The correction operation includes:
[0073] The current alignment coefficient is calculated according to the first average load, the first maximum load and the current second load prediction result; the current second load prediction result of the first correction operation is the second load prediction result;
[0074] Aligning the current second load forecast result according to the first load forecast result, the current second load forecast result, the first average load and the current alignment coefficient to obtain an aligned second load forecast result;
[0075] According to a preset load sliding length, sliding adjustment is performed on the aligned second load prediction result according to the load sliding length to obtain an updated second load prediction result;
[0076] The current alignment coefficient is updated according to the updated second load prediction result.
[0077] Preferably, the current second load forecast result is aligned according to the first load forecast result, the current second load forecast result, the first average load and the current alignment coefficient to obtain the aligned second load forecast result, including:
[0078] According to the current second load forecast result, a second average load of the current second load forecast result is calculated;
[0079] The translation coefficient is calculated by dividing the second average load and the first average load;
[0080] According to the current second load prediction result and the translation coefficient, an initial translation load prediction result is obtained;
[0081] Alignment is performed according to the initial translation load prediction result and the current alignment coefficient to obtain an aligned second load prediction result.
[0082] In a preferred embodiment of the present invention, Figure 3 As shown, a) the second load forecasting model is used to predict 8760 times to obtain the second load forecasting result; the hourly forecast load of the annual hourly granularity (8760 hours), {l pred,t , l pred,t+1 ,……,l pred,t+8759}, if the forecast year is a leap year, copy the forecast result of February 28 as the forecast result of February 29; use the first load forecasting model to forecast once respectively, and obtain the first average load and the first maximum load in the first load forecast result, which are recorded as {l pred_ave,m , l pred_ave,m+1 ,……,l pred_ave,m+11} and {l pred_max,m , l pred_max,m+1 ,……,l pred_max,m+11}. The forecast load for a month is revised each time, and the forecast result of the hourly granularity in the revised month is recorded as l pred = {l pred,t , l pred,t+1 , ...}, in the month to be corrected, the prediction results of the first average load and the first maximum load by the first load forecasting model are recorded as l pred_ave and l pred_max b) pred Find the average and get the second average load μ, and then pred_ave Divide by μ to get the translation coefficient α, α = l pred_ave / μ; pred Multiply all elements in by the translation coefficient α to obtain the initial translation load forecast result after the monthly average load is aligned, which is recorded as c) To make The maximum load is equal to l pred_max , other elements are greater than l pred_ave The part is stretched proportionally. First find l pred The first maximum load is denoted as λ, and the alignment coefficient β1 is calculated as: Will All the values greater than l pred_ave The element is denoted as A, and A is stretched according to the alignment coefficient β1: A: = l ped_ave +β1(Al pred_ave ), the total amount of alignment is recorded as P, P = ∑(β-1)·(Al prred_ave , the second load prediction result after the first stretch is recorded as d) Less than l pred_ave The loads are aligned so that the average value of the second load prediction result after alignment is equal to l pred_ave .Will All the values less than l pred_ave The elements of are denoted as B, and the total amount of less than S is calculated, S = ∑(l pred_ave -B). It is necessary to reduce the pred_ave The part of P is also aligned to make the average value of the second load forecast result equal to l pred_ave , so the stretch coefficient β2 can be calculated, Stretch B according to the stretching coefficient β2: = l pred_ave -β2(l pred_ave -B), the second load prediction result after the second stretch is recorded as e) Performing steps b) to d) on the second load forecast results of each month, aligning the second load forecast results of all months, and recording the aligned second load forecast results as
[0083] In another preferred embodiment of the present invention, a) the second load forecast result after alignment is recorded as The first hour of the month to be adjusted for sliding average is tgap, and the corresponding load is The preset load sliding length is set to 73 hours, the initial position of the load sliding window is t_gap-72~t_gap, and the end position is t_gap-1~t_gap+71. b) The first time point in the load sliding window is recorded as w, and the last time point is w+72. The load sliding window is recorded as the load at the first time point each time. The following sliding average calculation is performed: Then the load sliding window moves one hour to the right of the time series, and the above calculation is performed again, and repeated until the load sliding window reaches its end position. c) Execute steps a) and b) for the hourly granularity forecast results of each month except January, and adjust the second load forecast results between all months. Repeat the correction operation until the preset loss function converges, that is, the stretching coefficient 1-ν≤β1≤1+ν, stop the correction operation, and ν can be set to 1%. After stopping the correction operation, the final hourly forecast load can be obtained.
[0084] Preferably, performing power planning for the area to be planned according to the hourly predicted load includes:
[0085] Counting the number of time periods without output within a preset time length according to the hourly forecast load;
[0086] According to the number of time periods and the hourly forecast load, the new energy output of the area to be planned within the time period to be planned is calculated at a preset confidence level;
[0087] The total installed capacity of the planned area within the planned time period is planned according to the new energy output and the hourly predicted load.
[0088] In a preferred embodiment of the present invention, through weather and historical data, the number of time periods in which new energy (such as photovoltaic) has no output for n consecutive hours / n days in different periods such as spring, summer, autumn and winter can be estimated, so the output of new energy at different confidence levels at the hourly granularity for the whole year of next year can be calculated. Then, according to the hourly granular load for the whole year of next year, and the output of new energy at different confidence levels at the hourly granularity for the whole year of next year, the required installed capacity of thermal power plants and energy storage capacity can be planned in advance. In the new power system, the proportion of renewable energy increases year by year, and the uncertainty on the power generation side of the power system increases. In this context, the medium-term load forecast results with monthly granularity cannot meet the planning requirements of the new power system. The hourly granular load forecasting method proposed in this embodiment can obtain the hourly granular load forecasting results for the whole year in the future. According to the forecast results, power planners can better plan the allocation of power generation resources and formulate power-related policies, showing a high practical value.
[0089] By implementing this embodiment, the historical load data, historical weather data, historical temperature data and calendar data of the area to be planned are obtained; the historical load data, the historical temperature data and the calendar data are input into the preset first load forecasting model for monthly load forecasting, and the first load forecasting result for representing the monthly granularity is obtained; the historical load data, the historical weather data and the calendar data are input into the preset second load forecasting model for hourly load forecasting, and the second load forecasting result for representing the hourly granularity is obtained; according to the first load forecasting result and the second load forecasting result, the second load forecasting result is corrected to obtain the hourly predicted load; according to the hourly predicted load, the power planning is performed in the area to be planned. The first load forecasting result and the second load forecasting result are obtained by the historical load data, the historical weather data, the historical temperature data and the calendar data, and then the first load forecasting result is used to correct the second load forecasting result, so that the second load forecasting result of the hourly granularity is more accurate. Compared with the monthly granularity forecasting result, the hourly granularity forecasting result can take into account the power imbalance problem of consecutive days, that is, the second load forecasting result takes into account the load of each hour in consecutive days, and meets the planning requirements of the power system at the hourly granularity.
[0090] See also Figure 2 , is a schematic diagram of a structure of a power planning device based on hourly forecast load correction provided by an embodiment of the present invention, comprising:
[0091] A data acquisition module is used to acquire historical load data, historical weather data, historical temperature data of the area to be planned, and calendar data of the time period to be planned;
[0092] A first load forecasting module, used for inputting the historical load data, the historical temperature data and the calendar data into a preset first load forecasting model to perform monthly load forecasting, and obtain a first load forecasting result for representing a monthly granularity;
[0093] A second load forecasting module, used for inputting the historical load data, the historical weather data and the calendar data into a preset second load forecasting model to perform hourly load forecasting, and obtain a second load forecasting result representing an hourly granularity;
[0094] A load correction module, configured to correct the second load forecast result according to the first load forecast result and the second load forecast result to obtain an hourly forecast load;
[0095] The power planning module is used to carry out power planning for the area to be planned within the planning time period according to the hourly predicted load.
[0096] Specifically, the power planning module is used to perform power planning for the planned area according to the hourly predicted load;
[0097] The performing power planning for the area to be planned within the planning time period according to the hourly predicted load includes:
[0098] Counting the number of time periods without output within a preset time length according to the hourly forecast load;
[0099] According to the number of time periods and the hourly forecast load, the new energy output of the area to be planned within the time period to be planned is calculated at a preset confidence level;
[0100] The total installed capacity of the planned area within the planned time period is planned according to the new energy output and the hourly predicted load.
[0101] The present invention provides an electric power planning device based on hourly predicted load correction, which obtains a first load forecast result and a second load forecast result through historical load data, historical weather data, historical temperature data and calendar data, and then corrects the second load forecast result with the first load forecast result to make the second load forecast result with hourly granularity more accurate. Compared with the monthly granularity forecast result, the hourly granularity forecast result can take into account the power imbalance problem for multiple consecutive days, that is, the second load forecast result takes into account the load of each hour in multiple consecutive days, thereby meeting the planning requirements of the electric power system at the hourly granularity.
[0102] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.
[0103] Those skilled in the art can clearly understand that, for the sake of convenience and brevity, the specific working process of the device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.
[0104] Another embodiment of the present invention further provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, a power planning method based on hourly forecast load correction as described in the above embodiment is implemented. The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, and a cloud server. The terminal device may include, but is not limited to, a processor and a memory.
[0105] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, and uses various interfaces and lines to connect various parts of the entire terminal device.
[0106] The memory can be used to store the computer program, and the processor realizes various functions of the terminal device by running or executing the computer program stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the mobile phone, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device or other volatile solid-state storage device.
[0107] Another embodiment of the present invention provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the power planning method based on hourly predicted load correction described in the above embodiment.
[0108] The storage medium is a computer-readable storage medium, and the computer program is stored in the computer-readable storage medium. When the computer program is executed by the processor, the steps of each of the above-mentioned method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium.
[0109] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.
Claims
1. A power planning method based on hourly forecast load correction, characterized in that: include: Obtain historical load data, historical weather data, historical temperature data of the area to be planned, and calendar data of the time period to be planned; Inputting the historical load data, the historical temperature data and the calendar data into a preset first load forecasting model to perform monthly load forecasting, and obtaining a first load forecasting result for representing monthly granularity; Inputting the historical load data, the historical weather data and the calendar data into a preset second load forecasting model to perform hourly load forecasting, and obtaining a second load forecasting result representing hourly granularity; According to the first load forecast result and the second load forecast result, the second load forecast result is corrected to obtain an hourly forecast load; According to the hourly forecast load, power planning is performed for the planned area.
2. A method for power planning based on hourly forecast load correction as claimed in claim 1, characterized in that: The training of the first load forecasting model includes: According to the historical load data, the historical temperature data and the calendar data, respectively calculating a first average load, a first maximum load, a monthly average temperature, a monthly maximum temperature, monthly historical load data and monthly historical temperature data for representing monthly granularity; According to the first average load, the first maximum load, the monthly average temperature, the monthly maximum temperature, the monthly historical load data, the monthly historical temperature data and the calendar data, a first test set for representing the data within the preset time period and a first training set for representing the data within the preset time period are constructed; wherein the monthly historical load data in the first test set is used as label data; Taking the first training set as input and the first predicted load, the first average load and the first maximum load used to represent the monthly granularity load within the planning time period as output, the first load forecasting model to be trained is trained until the error between the first predicted load and the monthly historical load data in the first test set is less than the preset first prediction error, and the first load forecasting result is obtained.
3. A method for power planning based on hourly forecast load correction as claimed in claim 2, characterized in that: The training of the second load forecasting model includes: According to the historical load data, the historical weather data and the calendar data, a second test set for representing hourly granularity within a preset time period and a second training set for representing hourly granularity within a historical time period that is not within the preset time period are constructed; wherein the historical load data in the second test set is used as label data; Taking the second training set as input and the second predicted load used to represent the hourly granularity load in the planning time period as output, the second load prediction model to be trained is trained until the error between the second predicted load and the historical load data in the second test set is less than the preset second prediction error, and a second load prediction result is obtained.
4. A method for power planning based on hourly forecast load correction as claimed in claim 3, characterized in that: According to the first load forecast result and the second load forecast result, the second load forecast result is corrected to obtain the hourly forecast load, including: Taking the first load forecast result and the second load forecast result as input and the revised second load forecast result as output, repeatedly performing the revision operation until the preset loss function converges, obtaining the final second load forecast result, and using the final second load forecast result as the hourly load forecast; The correction operation includes: The current alignment coefficient is calculated based on the first average load, the first maximum load and the current second load prediction result; the current second load prediction result of the initial correction operation is the second load prediction result Aligning the current second load forecast result according to the first load forecast result, the current second load forecast result, the first average load and the current alignment coefficient to obtain an aligned second load forecast result; According to a preset load sliding length, sliding adjustment is performed on the aligned second load prediction result according to the load sliding length to obtain an updated second load prediction result; The current alignment coefficient is updated according to the updated second load prediction result.
5. A method for power planning based on hourly forecast load correction as claimed in claim 4, characterized in that: According to the first load forecast result, the current second load forecast result, the first average load and the current alignment coefficient, the current second load forecast result is aligned to obtain an aligned second load forecast result, including: According to the current second load forecast result, a second average load of the current second load forecast result is calculated; The translation coefficient is calculated by dividing the second average load and the first average load; According to the current second load prediction result and the translation coefficient, an initial translation load prediction result is obtained; Alignment is performed according to the initial translation load prediction result and the current alignment coefficient to obtain an aligned second load prediction result.
6. The power planning method based on hourly forecast load correction according to claim 1, characterized in that: According to the hourly forecast load, power planning is carried out for the planned area, including: Counting the number of time periods without output within a preset time length according to the hourly forecast load; According to the number of time periods and the hourly forecast load, the new energy output of the area to be planned within the time period to be planned is calculated at a preset confidence level; The total installed capacity of the planned area within the planned time period is planned according to the new energy output and the hourly predicted load.
7. A power planning device based on hourly forecast load correction, characterized in that: include: A data acquisition module is used to acquire historical load data, historical weather data, historical temperature data of the area to be planned, and calendar data of the time period to be planned; A first load forecasting module, used for inputting the historical load data, the historical temperature data and the calendar data into a preset first load forecasting model to perform monthly load forecasting, and obtain a first load forecasting result for representing a monthly granularity; A second load forecasting module, used for inputting the historical load data, the historical weather data and the calendar data into a preset second load forecasting model to perform hourly load forecasting, and obtain a second load forecasting result representing an hourly granularity; A load correction module, configured to correct the second load forecast result according to the first load forecast result and the second load forecast result to obtain an hourly forecast load; The power planning module is used to carry out power planning for the area to be planned within the planning time period according to the hourly predicted load.
8. The power planning device based on hourly forecast load correction according to claim 7, characterized in that: The power planning module is used to perform power planning for the planned area according to the hourly predicted load; The performing power planning for the area to be planned within the planning time period according to the hourly predicted load includes: Counting the number of time periods without output within a preset time length according to the hourly forecast load; According to the number of time periods and the hourly forecast load, the new energy output of the area to be planned within the time period to be planned is calculated at a preset confidence level; The total installed capacity of the planned area within the planned time period is planned according to the new energy output and the hourly predicted load.
9. A terminal device, characterized in that: The invention comprises a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, an electric power planning method based on hourly predicted load correction as described in any one of claims 1 to 6 is implemented.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute a power planning method based on hourly predicted load correction as described in any one of claims 1 to 6.
