Method for predicting electric power load

Abstract

The present application relates to the technical field of electric power operation and maintenance, in particular to a kind of electric power load prediction method, comprising the following steps: step one: data acquisition;Step two: determine sequence periodicity parameter M, carry out periodicity accumulation of sampling sequence, obtain new periodicity accumulation sequence H1;Step three: generate load prediction model;Step four: model pending estimated parameter value;Step five: obtain new prediction model;Step six: carry out load model reliability verification, change the traditional load prediction series modeling mode, improve the prediction accuracy, reduce the generation of redundant series data information.

Description

Technical Field

[0001] This invention relates to the field of power operation and maintenance technology, and specifically to a power load forecasting method. Background Technology

[0002] Load forecasting is fundamental to ensuring a balance between power supply and demand. It provides information and a basis for the planning and construction of power grids and power sources, as well as for the operational decisions of power grid companies and users, and is directly related to the demand dispatching of the power system. Accurate load forecasting can effectively improve the planning and dispatching capabilities of the power grid and enhance its operational robustness. Existing load forecasting technologies include time series forecasting, which is easy to apply but requires high stability of the data. Neural network methods suffer from drawbacks such as being prone to getting trapped in local optima during learning, difficulty in determining the number of iterations, relatively large generalization errors, and difficulty in determining the hidden neurons.

[0003] Load forecasting is a typical type of forecasting based on a relatively small amount of data, random and uncertain information sequences, and these sequences often exhibit periodicity. However, current load forecasting models model solely based on the data sequence itself, ignoring its periodicity. This leads to data redundancy and periodic revisions of forecasts, increasing socioeconomic costs and creating security vulnerabilities in existing systems. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a method for predicting power load.

[0005] The technical solution adopted by this invention to solve its technical problem is: a power load forecasting method, comprising the following steps:

[0006] Step 1: Data Acquisition;

[0007] Step 2: Determine the periodicity parameter M of the sequence, perform periodic accumulation of the sampled sequence, and obtain a new periodic accumulation sequence H1;

[0008] Step 3: Generate the load forecasting model;

[0009] Step 4: Estimating undetermined parameter values ​​for the model;

[0010] Step 5: Obtain the new prediction model;

[0011] Step 6: Perform load model reliability verification.

[0012] In step one, an initial sampling sequence H0 = {h0(1),h0(2),…,h0(i),…,h0(N)} is obtained within no less than 4 days, where i is the sequence sampling point, i∈[1,N], N≥16, and represents the total number of load samples.

[0013] In step two, the new periodic accumulation sequence H1 is represented as follows:

[0014] H1={∑h0(1+j-1),∑h0(2+j-1),…,∑h0(i+j-1),…,∑h0(N+j-1)}, where j∈[1,M], M has an hour as the smallest period unit, and h1(1)=∑h0(1+j-1), h1(b)=∑h0(b+j-1), where b is the load sequence point to be predicted, b∈[1,L], L is the total number of elements in the periodic cumulative sequence H1, L≤N, and h1 as a whole represents the sequence code symbol generated by one data accumulation.

[0015] In step three, the load forecasting model is generated according to the following formula:

[0016] h1(b+1) = d1h1(b) + d2;

[0017] In the formula, d1 and d2 are the undetermined predicted parameter values ​​of the model, which are obtained from periodic sequences through the least squares fitting principle.

[0018] The formulas for calculating the undetermined predicted parameter values ​​d1 and d2 of the model are as follows:

[0019] [d1, d2] T =(R T R) -1 R T W;

[0020] In the formula, R = [h1(1), h1(2), ..., h1(LM); 1, 1, ..., 1] T ;

[0021] W=[hl(2), hl(3),..., hl(L-M+l)] T .

[0022] Step five includes the following sub-steps:

[0023] 5-1: Using the principle of taking the nearest value from the data sequence, a new prediction model is obtained, and its calculation is as follows:

[0024]

[0025] In the formula, u is the correction coefficient for the periodic series; u=∑[h1(k+1)-d1h1(k)-d2] / d1h1(k-1), k∈[1,b];

[0026] 5-2: Based on the existing sequence model parameters, the load forecasting model is obtained as follows:

[0027]

[0028] In step six, the reliability of the load model is verified based on the error coefficient P1.

[0029] If the error coefficient P1 is less than or equal to 5%, then load forecasting is performed; otherwise, return to step one to collect the next set of data and generate the model.

[0030] The formula for calculating the error coefficient P1 is as follows:

[0031]

[0032] In the formula, S represents the total number of predicted load points, S∈[1,N], p∈[1,S].

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] This invention provides a power load forecasting method that changes the traditional load forecasting series modeling approach. By extracting and analyzing the periodic features of the series, it improves forecasting accuracy, reduces the generation of redundant series data, lowers socioeconomic costs, and further enhances the robustness and accuracy of load forecasting by using appropriate correction coefficients, thereby significantly reducing the risk of the operating system. Attached Figure Description

[0035] Figure 1 This is a flowchart of the present invention.

[0036] Figure 2 This is a demonstration diagram of the operation of the present invention. Detailed Implementation

[0037] The embodiments of the present invention will be further described below with reference to the accompanying drawings:

[0038] Example 1

[0039] like Figure 1 and Figure 2 As shown, the power load forecasting method includes the following steps:

[0040] Step 1: Data Acquisition; In Step 1, an initial sampling sequence H0 = {h0(1),h0(2),…,h0(i),…,h0(N)} is obtained within no less than 4 days, where i is the sequence sampling point, i∈[1,N], N≥16, representing the total number of load samples. The power SCADA system database stores power load data, and the system terminal obtains the above data from the system database.

[0041] Step 2: Determine the periodicity parameter M of the sequence, and perform periodic accumulation of the sampled sequence to obtain a new periodic accumulation sequence H1; in Step 2, the new periodic accumulation sequence H1 is represented as follows:

[0042] H1 = {∑h0(1+j-1),∑h0(2+j-1),…,∑h0(i+j-1),…,∑h0(N+j-1)}, where j∈[1,M], M has an hour as the smallest periodic unit, and h1(1) = ∑h0(1+j-1), h1(b) = ∑h0(b+j-1), where b is the load sequence point to be predicted, b∈[1,L], L is the total number of elements in the periodic cumulative sequence H1, L≤N, and to ensure the robustness of the sequence, L is generally taken as N. h1 as a whole represents the sequence symbol generated by one data accumulation.

[0043] Step 3: Generate the load forecasting model; in Step 3, the load forecasting model is generated according to the following formula:

[0044] h1(b+1) = d1h1(b) + d2;

[0045] In the formula, d1 and d2 are the undetermined predicted parameter values ​​of the model, which are obtained from periodic sequences through the least squares fitting principle.

[0046] Step 4: Determine the model's undetermined predicted parameter values; the calculation formulas for the model's undetermined predicted parameter values ​​d1 and d2 are as follows:

[0047] [d1, d2] T =(R T R) -1 R T W;

[0048] In the formula, R = [h1(1), h1(2), ..., h1(LM); 1, 1, ..., 1] T ;

[0049] W=[hl(2), hl(3),..., hl(L-M+l)] T .

[0050] Step 5: Obtain a new prediction model; Step 5 includes the following sub-steps:

[0051] 5-1: Using the principle of taking the nearest value from the data sequence, a new prediction model is obtained, and its calculation is as follows:

[0052]

[0053] In the formula, u is the correction coefficient for the periodic series; u=∑[h1(k+1)-d1h1(k)-d2] / d1h1(k-1), k∈[1,b];

[0054] 5-2: Based on the existing sequence model parameters, the load forecasting model is obtained as follows:

[0055]

[0056] Step Six: Perform load model reliability verification. In Step Six, the load model reliability verification is performed based on the error coefficient P1.

[0057] If the error coefficient P1 is less than or equal to 5%, then load forecasting is performed; otherwise, return to step one to collect the next set of data and generate the model.

[0058] The formula for calculating the error coefficient P1 is as follows:

[0059]

[0060] In the formula, S represents the total number of predicted load points, S∈[1,N], p∈[1,S].

[0061] Example 2

[0062] Taking the active load sampling data of a power company in northern Shandong Province over 10 days, which is processed every 2 hours, as an example, we compare it with the load prediction results of a neural network self-learning model and an autoregressive model.

[0063] The specific historical operational data is shown in the table below:

[0064]

[0065] The running display is as follows Figure 1 As shown in the figure, the periodicity of historical operational data is mainly reflected in different time periods of each day. The operational data for 10 days is basically stable, with periodic changes occurring every 8 hours. The basic process of load forecasting using the method of this invention is as follows: Figure 1 As shown, the load forecasting model is implemented using the steps and methods of this invention. The comparison results of load forecasting results from combining other neural network self-learning models and autoregressive models are as follows: The comparison of predicted results for the next 8-hour operating data period is as follows (sampled every 2 hours):

[0066] Actual value of power load sampling The load model prediction value of this invention absolute error Relative error rate % 0.1621 0.1653 0.0032 1.974090068 0.2161 0.2205 0.0044 2.036094401 0.6159 0.5971 0.0188 3.052443579 0.7155 0.7136 0.0019 0.265548567

[0067] The relative error rate of load forecasting using the method of this invention is within 5%, and the model prediction meets the expected accuracy. Data analysis using MATLAB provides a comparison of the average relative error rates of other methods, as shown in the table below:

[0068] Method of the present invention Neural network self-learning model Autoregressive model Average prediction relative error rate 1.8％ 1.9％ 2.5％

[0069] It is evident that the method of the present invention has a low average relative error rate in load forecasting.

Claims

1. A method for predicting electricity load, characterized in that, Includes the following steps: Step 1: Data Acquisition; In Step 1, an initial sampling sequence H0={h0(1),h0(2),…,h0(i),…,h0(N)} is obtained within no less than 4 days, where i is the sequence sampling point, i∈[1,N], N≥16, representing the total number of load samples; Step 2: Determine the periodicity parameter M of the sequence, and perform periodic accumulation of the sampled sequence to obtain a new periodic accumulation sequence H1; in Step 2, the new periodic accumulation sequence H1 is represented as follows: H1={∑h0(1+j-1),∑h0(2+j-1),…,∑h0(i+j-1),…,∑h0(N+j-1)}, where j∈[1,M], M is the smallest period unit of hours, and h1(1)=∑h0(1+j-1), h1(b)=∑h0(b+j-1), where b is the load sequence point to be predicted, b∈[1,L], L is the total number of elements in the periodic cumulative sequence H1, L≤N, and h1 represents the sequence code symbol generated by one data accumulation; Step 3: Generate the load forecasting model; in Step 3, the load forecasting model is generated according to the following formula: h1(b+1)=d1h1(b)+d2; In the formula, d1 and d2 are the undetermined predicted parameter values ​​of the model, which are obtained from periodic sequences through the least squares fitting principle; Step 4: Estimate parameter values; Step 5: Obtain a new prediction model; Step 5 includes the following sub-steps: 5-1: Using the principle of taking the nearest value from the data sequence, a new prediction model is obtained, and its calculation is as follows: ； In the formula, u is the correction coefficient for the periodic series; u=∑[h1(k+1)-d1h1(k)-d2] / d1h1(k-1), k∈[1,b]; 5-2: Based on the existing sequence model parameters, the load forecasting model is obtained as follows: ,i∈[1,N]; Step 6: Verify the reliability of the load model based on the error coefficient P1.

2. The power load forecasting method according to claim 1, characterized in that, The formulas for calculating the undetermined predicted parameter values ​​d1 and d2 of the model are as follows: ； In the formula, R = [h1(1), h1(2), ..., h1(LM); 1, 1, ..., 1] T ; 。 3. The power load forecasting method according to claim 1, characterized in that, In step six, if the error coefficient P1 is less than or equal to 5%, load forecasting is performed; otherwise, return to step one to collect the next set of data and generate the model.

4. The power load forecasting method according to claim 3, characterized in that, The formula for calculating the error coefficient P1 is as follows: ； In the formula, S represents the total number of predicted load points, S∈[1,N], p∈[1,S].

Images