Novel power system load prediction method

Abstract

The invention relates to the technical field of electric power, and discloses a novel electric power system load prediction method, which comprises the following steps of: acquiring electric power and corresponding comprehensive energy system data in an area on site, performing feature matching classification, determining prediction indexes, expanding a prediction range, establishing a secondary prediction system, and performing GM (1, 2, 3, 4, 5, 6, 7, 8, 8, 9, 10) on the secondary prediction indexes, the method comprises: 1) performing grey prediction to obtain a predicted value sequence, and generating a grey prediction model; performing grey correlation degree analysis on the second-level prediction index and the first-level prediction index; constructing a Bayesian network structure; constructing a father node; sequentially setting the value domains of the father node and the child node; determining a probability value and probability distribution; and comparing a prediction result with an actual result for verification. According to the method, all factors which play a role in the future are included in the prediction model, the influence weight of each prediction index is accurately analyzed, and assistance is provided for future work key points of employees of electric power departments and related industries.

Description

Technical Field

[0001] The present invention relates to the field of electric power technology, and in particular to a novel method for predicting load of an electric power system. Background Art

[0002] In the power industry, as new energy power generation technology becomes increasingly mature, the proportion of new energy power generation methods such as wind power and photovoltaic power generation in the total power generation in the country has been steadily increasing. Unlike traditional power generation methods, which have the advantages of stable output and convenient grid management and dispatch, new energy power generation is greatly affected by weather and resource availability. At the same time, in terms of electricity consumption, with the rapid development of modern industry, the impact of large-capacity, nonlinear power impact loads such as arc furnaces, traction loads, and electric vehicle charging loads in the power grid has become more prominent.

[0003] The new power system is based on the premise of carrying the inherent requirements of achieving carbon peak and carbon neutrality, implementing the new development concept, building a new development pattern, and promoting high-quality development. It ensures energy and power security as the basic premise, meets the power demand of economic and social development as the primary goal, and maximizes the absorption of new energy as the main task. It takes a strong smart grid as the hub platform, and is supported by the interaction of source, grid, load and storage and multi-energy complementarity. It is a power system with the basic characteristics of clean and low-carbon, safe and controllable, flexible and efficient, intelligent and friendly, and open and interactive. At present, the new power system is far different from the traditional power system. The volatility and intermittency of power generation and the openness and uncertainty of electricity consumption all pose great challenges to the planning, operation, regulation and analysis of the new power system. To solve the above challenges, better predictions of grid load and power generation are needed to improve the "absorption" capacity of the grid.

[0004] Load forecasting is based on known regional electricity demand and comprehensively considers a variety of relevant influencing factors to make predictions, including demand forecasts and power consumption forecasts. Among them, demand forecasts determine the size of the new capacity of power generation, transmission, and distribution equipment; power consumption forecasts determine the types of equipment such as base load units and peak load units. Load forecasting provides the current status of load development, and at the same time determines the supply and consumption of electricity and the maximum load in each planned year, thereby determining the long-term planned power load composition of the new power system.

[0005] Traditional methods for grid load forecasting mainly include time series analysis, trend analysis, regression analysis, exponential smoothing, grey prediction model, etc. These methods are suitable for short-term power load forecasting, but they have the following disadvantages: huge workload, difficulty in reflecting the influence of various factors, inability to effectively deal with load changes and emergencies, inability to effectively deal with large data discreteness, low reliability of regional forecasting, inability to fully utilize existing system information, and uncertainty. In the prior art, although the introduction of BP neural network can make better predictions, BP neural network has the limitation of being too sensitive to initial weights, resulting in the inability to make more accurate predictions of the uncertainty of new power systems. Therefore, a new power system load forecasting method is proposed for the study of accurate prediction of user-side power load in new power systems. Summary of the invention

[0006] In order to solve the problems in the prior art of lacking effective, long-term and accurate prediction of power grid load in a complex power grid environment, resulting in low accuracy of power load prediction and affecting the prediction results, the present invention provides a novel power system load prediction method.

[0007] The present invention is implemented by the following technologies: The present invention provides a novel power system load forecasting method, comprising the following steps: S1. Establish a prediction indicator system Collect the power and corresponding integrated energy system data in the area on site, and perform feature matching and classification. Establish forecasting indicators, expand the forecasting scope and establish a secondary forecasting system. The secondary forecasting indicator system includes first-level forecasting indicators and second-level forecasting indicators.

[0008] Among them, the first-level prediction indicators include wind turbines F, photovoltaics P, charging piles CS, energy storage ES, refrigeration R, electric boilers EB, water tanks WT, street lights SL, and computer rooms MR.

[0009] The second-level prediction indicators include tree-type wind turbines F1, vertical-axis wind turbines F2, and horizontal-axis wind turbines F3 in wind turbines F; fixed-axis P1, single-axis P2, double-axis P3, monocrystalline P4, polycrystalline P5, single-sided P6, double-sided P7, and vertical P8 in photovoltaic P; AC CS1, AC-DC CS2, DC-DC CS3, V2G (Vehicle to Grid) CS4, fast charging CS5, and slow charging CS6 in charging piles CS; 100kW / 15s supercapacitor ES1, 500kW / 1.5MWh flywheel energy storage ES2, and 333kW / 30s electrochemical ES3 in energy storage ES; as well as refrigeration R, electric boiler EB, water tank WT, street lamp SL, and computer room MR.

[0010] S2, second level prediction indicator grey prediction Grey prediction is a prediction method based on grey system. When faced with many grey systems with incomplete information, in-depth research is needed to understand the actual behavior of the system and explore the potential mechanism of the system by analyzing the surface characteristics of a small amount of data and information, so as to discover the internal characteristics of the system and grasp the evolution law of the system.

[0011] The second-level prediction index is subjected to grey prediction using GM (1,1) to obtain the prediction value sequence and generate the grey prediction model.

[0012] The specific steps include: S21, firstly, the existing data samples need to be processed so that the grey system can better extract the information in the data. The present invention adopts the cumulative generation method to perform the grey generation method, and performs cumulative generation on the original sequence. While keeping the first column of the data unchanged, the remaining columns are added to themselves and the previous data to obtain the cumulative generation sequence. , Where: x (1) ( k ) represents the accumulation of sequence elements; x (0) ( i ) represents the original sequence element; S22, establishing a differential equation for the cumulatively generated sequence; , Where: a is the development coefficient, reflecting the development trend of the original sequence and the generated sequence; u To generate the control gray number; S23, introduction of vector Y and matrix B , determined by the least squares method a and u ; , S24, solving differential equations; , S25, yes Perform cumulative subtraction operations to obtain the predicted value sequence of the original data sequence. , Generate a grey prediction model for the second-level prediction indicators.

[0013] S3, Grey relational analysis of the second-level prediction indicators Since the causes of load are different, the significance and intensity of each factor are also different, so it is necessary to determine the correlation coefficient between each factor and the overall load to determine the key factors affecting the load forecast of the new power system. The gray correlation analysis method reflects the degree of correlation between curves and is suitable for the correlation analysis between the overall load and its related factors to understand the main characteristics of the new power system load and its main influencing factors. At the same time, the gray correlation analysis method makes up for the shortcomings caused by the use of mathematical statistics analysis, is applicable to any sample size, and can avoid the problem of inconsistency between quantitative results and qualitative analysis results.

[0014] Construct the initial value sequence of the predicted value sequence in the second-level prediction indicator; Calculate the absolute value of the difference between each forecast indicator sequence and the corresponding element of the reference sequence; Calculate the maximum difference and minimum difference between the two poles, and obtain the minimum absolute value and maximum absolute value of the two sequences; Calculate the correlation coefficient; Determine the grey correlation coefficient; calculate the correlation coefficient between the first-level prediction indicator containing the second-level prediction indicator and the second-level prediction indicator it contains, judge the influence of the second-level prediction indicator on the first-level prediction indicator according to the correlation coefficient, and select the indicator with significant influence.

[0015] The specific steps include: S31, construct the initial value sequence of the predicted value sequence in the second-level prediction index, make , in, X 0 is the reference series, which represents the loading of the first-level predictor including the second-level predictor; X 1 ，X 2 ，…，X i To compare the series, they represent the loadings of different second-level predictors; i is the number of second-level prediction indicators; S32, calculate the absolute value of the difference between each comparison sequence and the corresponding element of the reference sequence. ; S33, calculate the maximum difference and minimum difference between the two poles, and record the minimum absolute value of the two poles of the two sequences , let the maximum absolute value of the two sequences be ; S34, calculate the correlation coefficient as ξ, , Where: 0i(k) represents the correlation coefficient between the reference sequence and the ith comparison sequence at time k; p Represents the resolution coefficient, which is used to reduce the influence of excessive maximum value on the distortion of correlation coefficient, and can improve the resolution between correlation coefficients. p The value range is (0,1), usually 0.5; k =1,2, ,…, n ; i =1,2, ,…, n ; S35, determine the grey correlation coefficient γ , Where: γ 0i Represents the grey correlation coefficient between the comparison sequence and the reference sequence; Calculate the correlation coefficient between the first-level prediction indicators containing the second-level prediction indicators and the second-level prediction indicators they contain, and judge the influence of the second-level prediction indicators on the first-level prediction indicators based on the correlation coefficient. The larger the coefficient, the greater the influence of the second-level prediction indicators on the first-level prediction indicators. Eliminate indicators with insignificant influence and select indicators with significant influence.

[0016] S4, Grey relational analysis of the first-level prediction indicators Construct the initial value sequence of the predicted value sequence in the first-level prediction indicator; Calculate the absolute value of the difference between each forecast indicator sequence and the corresponding element of the reference sequence; Calculate the maximum difference and minimum difference between the two poles, and obtain the minimum absolute value and maximum absolute value of the two sequences; Calculate the correlation coefficient; Determine the grey correlation coefficient; calculate the correlation coefficients between all first-level prediction indicators, judge the impact of the first-level prediction indicators on the overall load of the new power system based on the correlation coefficients, and select the indicators with significant impact.

[0017] The specific steps include: S41, construct the initial value sequence of the prediction value sequence in the first-level prediction index, make , in, Y 0 is the reference sequence, which represents the overall load of the new power system; Y 1 ，Y 2 ，…，Y i For comparison series, they represent the loadings of different first-level predictors; i is the number of second-level prediction indicators; S42, calculate the absolute value of the difference between each comparison sequence and the corresponding element of the reference sequence. ; S43, calculate the maximum difference and minimum difference between the two poles, and record the minimum absolute value of the two poles of the two sequences , let the maximum absolute value of the two sequences be ; S44, calculate the correlation coefficient as ξ, , Where: 0i (k) represents the correlation coefficient between the reference sequence and the ith comparison sequence at time k; p Represents the resolution coefficient, which is used to reduce the influence of excessive maximum value on the distortion of correlation coefficient, and can improve the resolution between correlation coefficients. p The value range is (0,1), usually 0.5; k =1,2, ,…, n ; i =1,2, ,…, n ; S45, determine the grey correlation coefficient γ , Where: γ 0i Represents the grey correlation coefficient between the comparison sequence and the reference sequence; Calculate the correlation coefficient between all first-level prediction indicators, and judge the influence of the first-level prediction indicators on the overall load of the new power system based on the correlation coefficient. The larger the coefficient, the greater the influence of the first-level prediction indicators on the overall load of the new power system. Eliminate indicators with insignificant influence and select indicators with significant influence.

[0018] S5, Bayesian network reasoning Bayesian network is a probabilistic graph model consisting of a set of nodes and directed edges. It is used to represent the dependency between random variables and is convenient for modeling and reasoning about uncertainty problems. Nodes represent random variables and directed edges represent conditional dependencies. Joint probability distributions, conditional probability distributions, etc. can be easily calculated to make inferences and decisions. In the application of Bayesian network, the network structure and parameters are learned through training data, and then new data is input for inference to obtain probabilities, thereby improving the accuracy of predictions.

[0019] Construct a Bayesian network structure; construct a parent node; set the value range of the parent node and child node in turn; determine the probability value and probability distribution.

[0020] The specific steps include: S51, constructing Bayesian network structure A Bayesian network N (ξ, θ ) is composed of n variables Z =( Z 1 ，Z 2 ，…，Z n ), assuming that the nodes Z i have r i value, the parent node Pα( Z i )Total q i combinations; S52, build parent node According to the calculated grey correlation between the factors of each prediction index, the main factors affecting the load of the new power system are determined based on the size of the grey correlation, and the selected factors are classified and adjusted. The first-level prediction index containing the second-level prediction index is used as the parent node of the Bayesian network, and various factors are classified into the first-level prediction index according to the internal causal relationship; S53, set the value range The value range of the parent node is set to {0.33, 0.66, 0.99}, which respectively represent "abnormal", "fluctuation" and "normal" of the new power system load. The main factors affecting the new power system load are selected as child nodes, and the value range of the child nodes is set in the same way as the parent node. S54, Probability Distribution The probability value is determined through the experience of the expert team and expert parameter estimation of sample data, and the probability value is input into the probability table of the node to determine the probability distribution of the node.

[0021] S6, result comparison The predicted result is compared with the actual result for verification. If the deviation does not exceed 5%, it is proved that the result is accurate; otherwise, the step S2 is looped to add BP neural network prediction, and the arithmetic mean is taken with the gray prediction result, and then steps S3~S6 are performed.

[0022] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides a novel load forecasting method for a power system. Based on grey forecasting, 25 second-level forecasting indicators of power generation and power consumption are forecasted, and grey correlation analysis is performed to determine the influencing indicators, and then the grey correlation of 9 first-level forecasting indicators is analyzed. The Bayesian network is then used for reasoning, and the output is inferred through a probability model, making it more effective in processing uncertainty and incomplete data. Finally, the BP neural network is combined for forecasting, and the above process is cycled to greatly improve the accuracy of the forecast results.

[0023] The present invention effectively overcomes many deficiencies of existing prediction methods, solves the defects of time series method that have high requirements on the stability of original load data, ignores changes in load weights, and is difficult to grasp the characteristics of prediction rules; solves the defects that the exponential smoothing method is not suitable for sudden situations with large fluctuations in data and is difficult to adapt to continuous changes; solves the defects that the prediction accuracy of the gray system theory method depends on the discrete degree of the original data and has a small application range. Based on the accurate prediction and evaluation of the load and power generation conditions of new power systems, the present invention studies the load forms of new power systems on the user side, such as electric vehicle charging, avoids the influence of resource availability and large-capacity, nonlinear power shock loads, and reduces the prediction difficulties caused by the volatility and intermittency of power generation and the openness and uncertainty of electricity consumption.

[0024] The present invention includes all factors that will play a role in the future in the prediction model, and accurately analyzes the influence weights of various prediction indicators, providing assistance for the future work priorities of power departments and related industry practitioners, greatly saving manpower and material costs, and improving the load prediction technology level of new power systems. It is conducive to designing power management, reasonably arranging power grid operation modes and unit maintenance plans, saving coal and oil to reduce power generation costs, and formulating reasonable construction plans. While improving the "absorption" capacity of the power grid in the region, it enhances the stability of the new power system, greatly improves the utilization rate of new energy and the economic and social benefits of the new power system, and thus helps achieve my country's "dual carbon" goals. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a specific flow chart of the method of the present invention.

[0026] Figure 2 It is a prediction system diagram of the method of the present invention. DETAILED DESCRIPTION

[0027] The specific embodiments of the present invention are described in detail below.

[0028] A new power system load forecasting method, such as Figure 1 As shown, the following steps are included: S1. Establish a prediction indicator system Collect data on power and the corresponding integrated energy system in the region, perform feature matching and classification, establish prediction indicators, expand the prediction scope, and establish a secondary prediction system. The secondary prediction indicator system includes 9 first-level prediction indicators and 25 second-level prediction indicators. like Figure 2 As shown, the first-level prediction indicators include wind turbine F, photovoltaic P, charging pile CS, energy storage ES, refrigeration R, electric boiler EB, water tank WT, street lamp SL, and computer room MR; The second-level prediction indicators include tree-type wind turbines F1, vertical-axis wind turbines F2, and horizontal-axis wind turbines F3 in wind turbines F; fixed-axis P1, single-axis P2, double-axis P3, monocrystalline P4, polycrystalline P5, single-sided P6, double-sided P7, and vertical P8 in photovoltaic P; AC CS1, AC-DC CS2, DC-DC CS3, V2G (Vehicle to Grid) CS4, fast charging CS5, and slow charging CS6 in charging piles CS; 100kW / 15s supercapacitor ES1, 500kW / 1.5MWh flywheel energy storage ES2, and 333kW / 30s electrochemical ES3 in energy storage ES; as well as refrigeration R, electric boiler EB, water tank WT, street lamp SL, and computer room MR.

[0029] S2, second level prediction indicator grey prediction Grey prediction is a prediction method based on grey system. When faced with many grey systems with incomplete information, in-depth research is needed to understand the actual behavior of the system and explore the potential mechanism of the system by analyzing the surface characteristics of a small amount of data and information, so as to discover the internal characteristics of the system and grasp the evolution law of the system.

[0030] The second-level prediction index is subjected to grey prediction using GM (1,1) to obtain the prediction value sequence and generate the grey prediction model.

[0031] The specific steps include: S21, firstly, the existing data samples need to be processed so that the grey system can better extract the information in the data. The present invention adopts the cumulative generation method to perform the grey generation method, and performs cumulative generation on the original sequence. While keeping the first column of the data unchanged, the remaining columns are added to themselves and the previous data to obtain the cumulative generation sequence. , Where: x (1) ( k ) represents the accumulation of sequence elements; x (0) ( i ) represents the original sequence element; S22, establishing a differential equation for the cumulatively generated sequence; , Where: a is the development coefficient, reflecting the development trend of the original sequence and the generated sequence; u To generate the control gray number; S23, introduction of vector Y and matrix B , determined by the least squares method a and u ; , S24, solving differential equations; , S25, yes Perform cumulative subtraction operations to obtain the predicted value sequence of the original data sequence. , Generate a grey prediction model for 25 second-level prediction indicators.

[0032] S3, Grey relational analysis of the second-level prediction indicators Since the causes of load are different, the significance and intensity of each factor are also different, so it is necessary to determine the correlation coefficient between each factor and the overall load to determine the key factors affecting the load forecast of the new power system. The gray correlation analysis method reflects the degree of correlation between curves and is suitable for the correlation analysis between the overall load and its related factors to understand the main characteristics of the new power system load and its main influencing factors. At the same time, the gray correlation analysis method makes up for the shortcomings caused by the use of mathematical statistics analysis, is applicable to any sample size, and can avoid the problem of inconsistency between quantitative results and qualitative analysis results.

[0033] Construct the initial value sequence of the predicted value sequence in the second-level prediction indicator; Calculate the absolute value of the difference between each forecast indicator sequence and the corresponding element of the reference sequence; Calculate the maximum difference and minimum difference between the two poles, and obtain the minimum absolute value and maximum absolute value of the two sequences; Calculate the correlation coefficient; Determine the grey correlation coefficient; calculate the correlation coefficient between the first-level prediction indicator containing the second-level prediction indicator and the second-level prediction indicator it contains, judge the influence of the second-level prediction indicator on the first-level prediction indicator according to the correlation coefficient, and select the indicator with significant influence.

[0034] The specific steps include: S31, construct the initial value sequence of the predicted value sequence in the second-level prediction index, make , in, X 0 is the reference series, which represents the loading of the first-level predictor including the second-level predictor; X 1 ，X 2 ，…，X i To compare the series, they represent the loadings of different second-level predictors; i is the number of second-level prediction indicators; S32, calculate the absolute value of the difference between each comparison sequence and the corresponding element of the reference sequence. ; S33, calculate the maximum difference and minimum difference between the two poles, and record the minimum absolute value of the two poles of the two sequences , let the maximum absolute value of the two sequences be ; S34, calculate the correlation coefficient as ξ, , Where: 0i (k) represents the correlation coefficient between the reference sequence and the ith comparison sequence at time k; p Represents the resolution coefficient, which is used to reduce the influence of excessive maximum value on the distortion of correlation coefficient, and can improve the resolution between correlation coefficients. p The value range is (0,1), usually 0.5; k =1,2, ,…, n ; i =1,2, ,…, n ; S35, determine the grey correlation coefficient γ , Where: γ 0i Represents the grey correlation coefficient between the comparison sequence and the reference sequence; Calculate the correlation coefficient between the 4 first-level prediction indicators containing the second-level prediction indicators and the 20 second-level prediction indicators they contain. Judge the influence of the second-level prediction indicators on the first-level prediction indicators based on the correlation coefficient. The larger the coefficient, the greater the influence of the second-level prediction indicator on the first-level prediction indicator. Eliminate the indicators with insignificant influence and select the indicators with significant influence.

[0035] S4, Grey relational analysis of the first-level prediction indicators Construct the initial value sequence of the predicted value sequence in the first-level prediction indicator; Calculate the absolute value of the difference between each forecast indicator sequence and the corresponding element of the reference sequence; Calculate the maximum difference and minimum difference between the two poles, and obtain the minimum absolute value and maximum absolute value of the two sequences; Calculate the correlation coefficient; Determine the grey correlation coefficient; calculate the correlation coefficients between all first-level prediction indicators, judge the impact of the first-level prediction indicators on the overall load of the new power system based on the correlation coefficients, and select the indicators with significant impact.

[0036] The specific steps include: S41, construct the initial value sequence of the prediction value sequence in the first-level prediction index, make , in, Y 0 is the reference sequence, which represents the overall load of the new power system; Y 1 ，Y 2 ，…，Y i For comparison series, they represent the loadings of different first-level predictors; i is the number of second-level prediction indicators; S42, calculate the absolute value of the difference between each comparison sequence and the corresponding element of the reference sequence. ; S43, calculate the maximum difference and minimum difference between the two poles, and record the minimum absolute value of the two poles of the two sequences , let the maximum absolute value of the two sequences be ; S44, calculate the correlation coefficient as ξ, , Where: 0i (k) represents the correlation coefficient between the reference sequence and the ith comparison sequence at time k; p Represents the resolution coefficient, which is used to reduce the influence of excessive maximum value on the distortion of correlation coefficient, and can improve the resolution between correlation coefficients. p The value range is (0,1), usually 0.5; k =1,2, ,…, n ; i =1,2, ,…, n ; S45, determine the grey correlation coefficient γ , Where: γ 0i Represents the grey correlation coefficient between the comparison sequence and the reference sequence; The correlation coefficients among the nine first-level prediction indicators are calculated, and the influence of the first-level prediction indicators on the overall load of the new power system is judged according to the correlation coefficients. The larger the coefficient, the greater the influence of the first-level prediction indicator on the overall load of the new power system. The indicators with insignificant influence are eliminated, and the indicators with significant influence are selected.

[0037] S5, Bayesian network reasoning Bayesian network is a probabilistic graph model consisting of a set of nodes and directed edges. It is used to represent the dependency between random variables and is convenient for modeling and reasoning about uncertainty problems. Nodes represent random variables and directed edges represent conditional dependencies. Joint probability distributions, conditional probability distributions, etc. can be easily calculated to make inferences and decisions. In the application of Bayesian network, the network structure and parameters are learned through training data, and then new data is input for inference to obtain probabilities, thereby improving the accuracy of predictions.

[0038] Construct a Bayesian network structure; construct a parent node; set the value range of the parent node and child node in turn; determine the probability value and probability distribution.

[0039] The specific steps include: S51, constructing Bayesian network structure A Bayesian network N (ξ, θ ) is composed of n variables Z =( Z 1 ，Z 2 ，…，Z n ), assuming that the nodes Z i have r i value, the parent node Pα( Z i )Total q i combinations; S52, build parent node According to the calculated grey correlation between the factors of each prediction index, the main factors affecting the load of the new power system are determined based on the size of the grey correlation, and the selected factors are classified and adjusted. The first-level prediction index containing the second-level prediction index is used as the parent node of the Bayesian network, and various factors are classified into the first-level prediction index according to the internal causal relationship; S53, set the value range The value range of the parent node is set to {0.33, 0.66, 0.99}, which respectively represent "abnormal", "fluctuation" and "normal" of the new power system load. The main factors affecting the new power system load are selected as child nodes, and the value range of the child nodes is set in the same way as the parent node. S54, Probability Distribution The probability value is determined through the experience of the expert team and expert parameter estimation of sample data, and the probability value is input into the probability table of the node to determine the probability distribution of the node.

[0040] S6, result comparison The predicted result is compared with the actual result for verification. If the deviation does not exceed 5%, it is proved that the result is accurate; otherwise, the step S2 is looped to add BP neural network prediction, and the arithmetic mean is taken with the gray prediction result, and then steps S3~S6 are performed.

[0041] The scope of protection claimed by the present invention is not limited to the above specific embodiments. For those skilled in the art, the present invention may have various deformations and changes. Any modifications, improvements and equivalent substitutions made within the concept and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A novel power system load forecasting method, characterized in that: The following steps are involved: S1. Establish a prediction indicator system Collecting data on power and the corresponding integrated energy system in the region, performing feature matching and classification, establishing prediction indicators, expanding the prediction scope and establishing a secondary prediction system, wherein the secondary prediction indicator system includes first-level prediction indicators and second-level prediction indicators; S2, second level prediction indicator grey prediction The second-level prediction index is subjected to grey prediction using GM (1,1) to obtain the prediction value sequence and generate the grey prediction model; S3, Grey relational analysis of the second-level prediction indicators Construct the initial value sequence of the predicted value sequence in the second-level prediction indicator; Calculate the absolute value of the difference between each forecast indicator sequence and the corresponding element of the reference sequence; Calculate the maximum difference and minimum difference between the two poles, and obtain the minimum absolute value and maximum absolute value of the two sequences; Calculate the correlation coefficient; Determine the grey correlation coefficient; calculate the correlation coefficient between the first-level prediction indicator containing the second-level prediction indicator and the second-level prediction indicator it contains, judge the influence of the second-level prediction indicator on the first-level prediction indicator according to the correlation coefficient, and select the indicator with significant influence; S4, Grey relational analysis of the first-level prediction indicators Construct the initial value sequence of the predicted value sequence in the first-level prediction indicator; Calculate the absolute value of the difference between each forecast indicator sequence and the corresponding element of the reference sequence; Calculate the maximum difference and minimum difference between the two poles, and obtain the minimum absolute value and maximum absolute value of the two sequences; Calculate the correlation coefficient; Determine the grey correlation coefficient; calculate the correlation coefficients between all first-level prediction indicators, and judge the influence of the first-level prediction indicators on the overall load of the new power system according to the correlation coefficients, and select the indicators with significant influence; S5, Bayesian network reasoning Construct a Bayesian network structure; construct a parent node; set the value range of the parent node and the child node in turn; determine the probability value and probability distribution; S6, result comparison The predicted result is compared with the actual result for verification. If the deviation does not exceed 5%, it is proved that the result is accurate; otherwise, the step S2 is looped to add BP neural network prediction, and the arithmetic mean is taken with the gray prediction result, and then steps S3~S6 are performed.

2. A novel power system load forecasting method according to claim 1, characterized in that: In step S1, the first-level prediction indicators include wind turbine F, photovoltaic P, charging pile CS, energy storage ES, refrigeration R, electric boiler EB, water tank WT, street lamp SL, and machine room MR; The second-level prediction index includes the tree-type wind turbine F1, the vertical axis wind turbine F2, and the horizontal axis wind turbine F3 in the wind turbine F; Fixed axis P1, uniaxial P2, biaxial P3, monocrystalline P4, polycrystalline P5, single-sided P6, bifacial P7, vertical P8 in photovoltaic P; The charging pile CS includes AC CS1, AC-DC CS2, DC-DC CS3, V2G CS4, fast charging CS5, and slow charging CS6; The energy storage ES includes 100kW / 15s supercapacitor ES1, 500kW / 1.5MWh flywheel energy storage ES2, 333kW / 30s electrochemical ES3; as well as refrigeration R, electric boiler EB, water tank WT, street lamp SL, and computer room MR.

3. A novel power system load forecasting method according to claim 1, characterized in that: In step S2, the following steps are specifically included: S21, perform cumulative generation on the original sequence to obtain a cumulative generation sequence, ， Where: x (1) ( k ) represents the accumulation of sequence elements; x (0) ( i ) represents the original sequence element; S22, establishing a differential equation for the cumulatively generated sequence; ， Where: a is the development coefficient; u To generate the control gray number; S23, introduction of vector Y and matrix B , determined by the least squares method a and u ; ， S24, solving differential equations; ， S25, yes Perform cumulative subtraction operations to obtain the predicted value sequence of the original data sequence. ， Generate a grey prediction model for the second-level prediction indicators.

4. A novel power system load forecasting method according to claim 1, characterized in that: Step S3 specifically includes the following steps: S31, construct the initial value sequence of the predicted value sequence in the second-level prediction index, make , in, X 0 is the reference series, which represents the loading of the first-level predictor including the second-level predictor; X 1 ，X 2 ，…，X i To compare the series, they represent the loadings of different second-level predictors; i is the number of second-level prediction indicators; S32, calculate the absolute value of the difference between each comparison sequence and the corresponding element of the reference sequence. ; S33, calculate the maximum difference and minimum difference between the two poles, and record the minimum absolute value of the two poles of the two sequences , let the maximum absolute value of the two sequences be ; S34, calculate the correlation coefficient as ξ, ， Where: 0i (k) represents the correlation coefficient between the reference sequence and the ith comparison sequence at time k; p represents the resolution coefficient, p The value range is (0,1), usually 0.5; k =1,2, ，…， n ； i =1,2, ，…， n ； S35, determine the grey correlation coefficient γ ， Where: γ 0i Represents the grey correlation coefficient between the comparison sequence and the reference sequence; Calculate the correlation coefficient between the first-level prediction indicators containing the second-level prediction indicators and the second-level prediction indicators they contain, judge the influence of the second-level prediction indicators on the first-level prediction indicators based on the correlation coefficient, eliminate the indicators with insignificant influence, and select the indicators with significant influence.

5. A novel power system load forecasting method according to claim 1, characterized in that: Step S4 specifically includes the following steps: S41, construct the initial value sequence of the prediction value sequence in the first-level prediction index, make , in, Y 0 is the reference sequence, which represents the overall load of the new power system; Y 1 ，Y 2 ，…，Y i For comparison series, they represent the loadings of different first-level predictors; i is the number of second-level prediction indicators; S42, calculate the absolute value of the difference between each comparison sequence and the corresponding element of the reference sequence. ; S43, calculate the maximum difference and minimum difference between the two poles, and record the minimum absolute value of the two poles of the two sequences , let the maximum absolute value of the two sequences be ; S44, calculate the correlation coefficient as ξ, ， Where: 0i (k) represents the correlation coefficient between the reference sequence and the ith comparison sequence at time k; p represents the resolution coefficient, p The value range is (0,1), usually 0.5; k =1,2, ，…， n ； i =1,2, ，…， n ； S45, determine the grey correlation coefficient γ ， Where: γ 0i Represents the grey correlation coefficient between the comparison sequence and the reference sequence; Calculate the correlation coefficients between all first-level prediction indicators, and judge the impact of the first-level prediction indicators on the overall load of the new power system based on the correlation coefficients. Eliminate indicators with insignificant impact and select indicators with significant impact.

6. A novel power system load forecasting method according to claim 1, characterized in that: Step S5 specifically includes the following steps: S51, constructing Bayesian network structure A Bayesian network N (ξ, θ ) is composed of n variables Z =( Z 1 ，Z 2 ，…，Z n ), assuming that the nodes Z i have r i value, the parent node Pα( Z i )Total q i combinations; S52, build parent node According to the calculated grey correlation between the factors of each prediction index, the main factors affecting the load of the new power system are determined based on the size of the grey correlation, and the selected factors are classified and adjusted. The first-level prediction index containing the second-level prediction index is used as the parent node of the Bayesian network, and various factors are classified into the first-level prediction index according to the internal causal relationship; S53, set the value range The value range of the parent node is set to {0.33, 0.66, 0.99}, which respectively represent "abnormal", "fluctuation" and "normal" of the new power system load. The main factors affecting the new power system load are selected as child nodes, and the value range of the child nodes is set in the same way as the parent node. S54, Probability Distribution The probability value is determined through the experience of the expert team and expert parameter estimation of sample data, and the probability value is input into the probability table of the node to determine the probability distribution of the node.

Images