Power demand prediction method and device based on two-stage optimization

Abstract

The invention provides a power demand prediction method and device based on two-stage optimization, and is suitable for the technical field of power systems, and the method comprises the steps: dividing a to-be-predicted region based on the population density and the area, and obtaining a plurality of regions; acquiring historical power consumption data of each region, determining a first power matrix of each region through a first prediction model, and determining a special region based on all the first power matrixes; determining a second power matrix of the special area based on a second prediction model, and optimizing the corresponding first power matrix based on the second power matrix; and determining a first special value matrix of the special area based on the historical power consumption data of the special area and the plurality of adjacent areas, and optimizing the optimized matrix based on the first special value matrix to obtain a final power matrix. According to the method, the special region can be predicted for multiple times through multiple models, the special region is compared with the adjacent region, the prediction result is optimized twice, and the accuracy of the prediction result of the special region is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of power systems, and in particular relates to a method and device for predicting power demand based on two-stage optimization. Background Art

[0002] With the development of society, the demand for electricity is also growing continuously, which brings greater challenges to the stable operation of the power grid. At the same time, the development of sustainable energy has also reduced the stability of the power source in the power grid. However, when predicting power demand, it is mainly predicted by a single model, and the use of a single model is more dependent on the adaptive optimization of model parameters.

[0003] The optimization method of adaptive optimization of model parameters of a single model cannot take into account the impact of other complex factors. The prediction results are prone to inaccurate predictions in certain specific situations. For example, many new factories were built in a certain area this year, and the demand for electricity increased. It is difficult for existing prediction models to fully consider these factors when making short-term predictions. Summary of the invention

[0004] The embodiment of the present invention provides a method and device for predicting power demand based on two-stage optimization to solve the problem that the existing prediction model cannot fully consider the influence of complex factors when optimizing its own model parameters.

[0005] The present invention is achieved through the following technical solutions:

[0006] In a first aspect, an embodiment of the present invention provides a method for predicting power demand based on two-stage optimization, comprising:

[0007] The predicted area is divided into multiple regions based on population density and area;

[0008] For each area, historical power consumption data of the area is obtained, a first power matrix of the area is determined based on the historical power consumption data and a first prediction model, and a special area is determined based on all first power matrices;

[0009] Determine a second power matrix of the special area based on the second prediction model, and optimize the corresponding first power matrix based on the second power matrix to obtain a third power matrix;

[0010] A first special value matrix of the special area is determined based on historical power consumption data of the special area and a plurality of adjacent areas thereof, and a final power matrix of the special area is obtained by optimizing the third power matrix based on the first special value matrix.

[0011] In a possible implementation, the region to be predicted is divided into multiple regions based on population density and area, including:

[0012] Acquiring the area of ​​the region to be predicted, and initially dividing the region to be predicted based on the area to obtain a plurality of initial regions;

[0013] The population density of each initial area is obtained, and the initial area is adjusted based on the population density to obtain a plurality of areas to be predicted.

[0014] In a possible implementation, for each region, acquiring historical power consumption data of the region, and determining a first power matrix of the region based on the historical power consumption data and a first prediction model includes:

[0015] For each region, perform the following steps:

[0016] Obtaining historical electricity consumption data of the area within a preset period, and randomly selecting a prediction model from a prediction model library as the first prediction model; the prediction model library includes multiple prediction models;

[0017] Perform prediction based on the first prediction model and the historical electricity consumption data to obtain an initial power matrix of the area;

[0018] An influencing factor of each influencing factor of power demand forecasting is determined, and a correction matrix is ​​determined based on the influencing factor, and the initial power matrix is ​​corrected based on the correction matrix to obtain a first power matrix.

[0019] In a possible implementation manner, determining the special area based on all first power matrices includes:

[0020] determining a reference matrix based on all first power matrices;

[0021] Comparing each first power matrix with the reference matrix, and determining a special power value in each first power matrix that exceeds a preset range;

[0022] The number of special power values ​​in each of the first power matrices is obtained, and the area corresponding to the first power matrix where the number exceeds a preset value is taken as a special area.

[0023] In a possible implementation manner, determining a reference matrix based on all first power matrices includes:

[0024] Determine the area and population density of the region corresponding to each first power matrix, and determine the weight of each first power matrix based on the area and the population density;

[0025] A weighted average value of each type of power value in all first power matrices is calculated based on the weight value, and a reference matrix is ​​determined based on the weighted average value.

[0026] In a possible implementation manner, determining the second power matrix of the special area based on the second prediction model includes:

[0027] Randomly select a prediction model from other prediction models in the prediction model library except the first prediction model as the second prediction model;

[0028] Performing prediction based on the second prediction model and the historical power consumption data corresponding to the special area to obtain an initial power matrix of the special area;

[0029] The initial power matrix of the special area is corrected based on the correction matrix corresponding to the special area to obtain a second power matrix of the special area.

[0030] In a possible implementation manner, the optimizing the corresponding first power matrix based on the second power matrix to obtain a third power matrix includes:

[0031] Acquire the model characteristics of the second prediction model of the special area, the number of special power values ​​in the second power matrix corresponding to the special area, and the difference between each special power value and the corresponding value in the reference matrix;

[0032] Constructing a difference matrix of the second power matrix based on the difference between each of the special power values ​​and the corresponding value in the reference matrix;

[0033] Calculating a weight of the second power matrix based on the model characteristics and the number of special power values ​​in the second power matrix;

[0034] Optimizing the first power matrix based on the weight of the second power matrix to obtain a third power matrix;

[0035] The weight of the second power matrix is:

[0036]

[0037] Among them, Q j is the weight of the second power matrix corresponding to the jth special area, C is the computational complexity of the second prediction model, f is the frequency of use of the second prediction model in the area, k is the number of special power values ​​in the second power matrix, is the difference between the special power value in the ath row and bth column of the difference matrix and the corresponding value in the reference matrix, a is the ath row of the difference matrix, b is the bth column of the difference matrix, A is the total number of rows of the difference matrix, B is the total number of columns of the difference matrix, and k=AB;

[0038] The third power matrix is:

[0039]

[0040] Among them, H3 j is the three-power matrix of the jth special area, H2 j is the second power matrix of the jth special area, H1 j is the first power matrix of the jth special area, Q j is the weight value of the second power matrix corresponding to the j-th special area.

[0041] In a possible implementation manner, determining the first special value matrix of the special area based on the historical power consumption data of the special area and a plurality of adjacent areas includes:

[0042] Determine a plurality of adjacent areas of the special area, and determine a historical power matrix of the plurality of adjacent areas based on historical power consumption data of the plurality of adjacent areas;

[0043] Determining a historical power matrix of the special area based on the historical power consumption data of the special area;

[0044] Comparing the historical power matrix of the special area with the historical power matrices of the plurality of adjacent areas to obtain a plurality of historical difference matrices;

[0045] A first special value matrix for the special area is determined based on the multiple historical difference matrices.

[0046] In a possible implementation manner, the optimizing the third power matrix based on the first special value matrix to obtain a final power matrix for the special area includes:

[0047] Acquire first power matrices of a plurality of adjacent areas of the special area, and compare the third power matrix with the first power matrices of the plurality of adjacent areas to obtain a plurality of prediction difference matrices;

[0048] Determine a second special value matrix for the special area based on the multiple prediction difference matrices;

[0049] Acquire influencing factor change data, and determine the influence coefficient of each influencing factor based on the influencing factor change data;

[0050] If there is an influence coefficient out of the preset range among the influence coefficients, determining a third prediction model based on the influence factors corresponding to the influence coefficient out of the preset range;

[0051] Predicting the special area based on the third prediction model to obtain a fourth power matrix of the special area, and correcting the corresponding third power matrix based on the fourth power matrix to obtain a final power matrix of the special area;

[0052] If there is no influence coefficient beyond the preset range among the influence coefficients, optimizing the third power matrix based on the first special value matrix and the second special value matrix to obtain a final power matrix for the special area;

[0053]

[0054] Among them, H j is the final power matrix of the jth special area, a is the ath row element of the first special value matrix and the second special value matrix of the jth special area, b is the bth column element of the first special value matrix and the second special value matrix of the jth special area, A is the power load data, B is the time period, is the influence factor of the xth influencing factor, ΔG(x) ab is the estimated impact of the xth influencing factor on the power data in the ath row and bth column, is the element in the ath row and bth column of the first special value matrix of the jth special region, is the element in the ath row and bth column of the second special value matrix of the jth special region, H3 j is the third power matrix of the j-th special area.

[0055] In a second aspect, an embodiment of the present invention provides a power demand forecasting device based on two-stage optimization, comprising:

[0056] A division module is used to divide the predicted area into multiple regions based on population density and area;

[0057] A determination module, configured to obtain, for each region, historical power consumption data of the region, determine a first power matrix of the region based on the historical power consumption data and a first prediction model, and determine a special region based on all first power matrices;

[0058] an optimization module, configured to determine a second power matrix of the special area based on a second prediction model, and optimize the corresponding first power matrix based on the second power matrix to obtain a third power matrix;

[0059] The optimization module is also used to determine a first special value matrix of the special area based on historical electricity consumption data of the special area and its multiple adjacent areas, and optimize the third power matrix based on the first special value matrix to obtain a final power matrix of the special area.

[0060] The embodiment of the present invention provides a method and device for predicting power demand based on two-stage optimization, which divides the region by population density and area to obtain multiple regions, and determines a special region by the prediction result of each region, and then re-predicts the special region by different prediction methods, and optimizes the result of the first prediction by the result of the second prediction to obtain a new prediction result, calculates the difference between the historical data of the special region and the adjacent region and the difference in the prediction result, and optimizes the new prediction result of the special region for the second time to obtain the final prediction result. The special region is predicted multiple times by multiple models, and the prediction results are obtained, which ensures the accuracy of the prediction results. The special region is determined by the prediction result of each region, and the prediction results of the special region are optimized by multiple predictions to obtain the final prediction result, which fully considers the influence of all factors and ensures the accuracy of the prediction of the special region. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or related technical descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0062] Figure 1 It is a flow chart of a method for predicting power demand based on two-stage optimization provided by an embodiment of the present invention;

[0063] Figure 2 is a diagram of adjacent regions provided by an embodiment of the present invention;

[0064] Figure 3 It is a structural schematic diagram of a power demand forecasting device based on two-stage optimization provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0065] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present invention. However, it should be clear to those skilled in the art that the present invention may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present invention.

[0066] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0067] It should also be understood that the term "and / or" used in the present description and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0068] As used in the present specification and the appended claims, the term "if" may be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" may be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]", depending on the context.

[0069] In addition, in the description of the present specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0070] References to "one embodiment" or "some embodiments" etc. described in the present specification mean that one or more embodiments of the present invention include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0071] Figure 1 is a flow chart of a method for predicting power demand based on two-stage optimization provided by an embodiment of the present invention, referring to Figure 1 , the power demand forecasting method based on two-stage optimization is described as follows:

[0072] S110: Divide the area to be predicted into multiple regions based on population density and area.

[0073] Among them, population density can be obtained through permanent population statistics.

[0074] In an optional embodiment, the area to be predicted is divided into multiple regions based on population density and area, including:

[0075] The area of ​​the region to be predicted is obtained, and the region to be predicted is initially divided based on the area to obtain multiple initial regions;

[0076] The population density of each initial area is obtained, and the initial area is adjusted based on the population density to obtain multiple areas to be predicted.

[0077] Among them, treating the predicted area based on the area can be to determine an initial area, and divide the predicted area into several parts through the initial surface, and the preliminary divided area can be obtained by different division standards. For example, there are 15 small areas in District A of a certain city, among which the area of ​​the smallest small area is B㎡, then B㎡ can be used as the preliminary divided area. It can also be to determine the different administrative divisions in the area, and divide the predicted area according to the areas of these administrative divisions. For example, there are 3 administrative divisions in District C, and the area of ​​each administrative division is C1, C2, and C3 respectively, then District C can be divided into three initial areas, and the area of ​​each initial area is C1, C2, and C3 respectively. The predicted area can also be divided in other ways, and this embodiment does not limit the specific division method.

[0078] Optionally, the initial area is adjusted by population density to obtain multiple areas to be predicted by obtaining the population density of each initial area and the overall population density of the area. For example, the total area of ​​area 1 is C, the total population density is X, and area 1 is divided into three initial areas. The area of ​​each initial area is C1, C2, and C3, and the population density of each initial area is X1, X2, and X3, respectively. Then the population of each initial area is approximately C1X1, C2X2, and C3X3. If Then the area of ​​the first and second initial areas can be increased accordingly, and the area of ​​the third initial area can be reduced, so that the number of people in each area is balanced. The specific calculation method for adjusting the area of ​​the area is not limited here.

[0079] Dividing the predicted area into multiple regions by population density and area can reduce the difference in population size in each region and maximize the similarity of each region.

[0080] S120: For each area, obtain historical power consumption data of the area, determine a first power matrix of the area based on the historical power consumption data and a first prediction model, and determine a special area based on all first power matrices.

[0081] Optionally, the amount of historical electricity consumption data is not specifically limited here, but since there will be large differences in electricity consumption in different seasons, the historical electricity consumption data in this embodiment is the electricity consumption data of the region in the corresponding season in the past few years. For example, if it is necessary to predict the electricity consumption from August 3, 2023 to August 10, 2023, the selected historical data should be the historical data of the region in early August from 2020 to 2022.

[0082] Among them, special areas refer to areas with significantly different electricity consumption patterns from other areas. For example, if the electricity consumption of area 1 from 8:00 to 12:00 in the morning is significantly higher than that of other areas, but the electricity consumption from 13:00 to 16:00 in the afternoon is significantly lower than that of other areas, then area 1 is a special area. Each region can have one or more special areas. The reason for the formation of special areas is related to the type of buildings. For example, if half of the area in area 1 is factory buildings, then area 1 is likely to become a special area.

[0083] In an optional embodiment, for each region, historical power consumption data of the region is obtained, and a first power matrix of the region is determined based on the historical power consumption data and a first prediction model, including:

[0084] For each region, perform the following steps:

[0085] Obtain historical electricity consumption data of the area within a preset period, and randomly select a prediction model from a prediction model library as a first prediction model; the prediction model library includes multiple prediction models;

[0086] Based on the first prediction model and the historical electricity consumption data, a prediction is performed to obtain an initial power matrix of the area;

[0087] An influencing factor of each influencing factor of power demand forecasting is determined, and a correction matrix is ​​determined based on the influencing factor. The initial power matrix is ​​corrected based on the correction matrix to obtain a first power matrix.

[0088] In this embodiment, the prediction model library is pre-set, and the prediction model library contains a variety of pre-trained prediction models, such as neural network prediction models, SARIMA (Seasonal Autoregressive Integrated Moving Average) models, regression prediction models, etc., some of which can be predicted through multiple factors and historical electricity consumption data, and some can be predicted through one factor and historical electricity consumption data.

[0089] In this embodiment, the first prediction model is determined by randomly selecting from a prediction model database, but the first prediction model can also be specified according to prediction requirements. For example, if a more accurate prediction is required, a prediction model with greater computational complexity but higher accuracy can be selected as the first prediction model.

[0090] In this embodiment, the different power consumption in different time periods is taken into consideration during the prediction process. When making predictions, different time periods are predicted separately to obtain an initial power matrix.

[0091] The initial power matrix is:

[0092]

[0093] Among them, E1 i is the initial load matrix of the ith region, A is the power data, B is the time period, is the first-category power value corresponding to the first period after correction in the i-th region, is the second type of power value corresponding to the first period of the i-th region, is the A-th class power value corresponding to the first period of the i-th region, is the first type of power value corresponding to the second period of the i-th region, is the second type of power value corresponding to the second period of the i-th region, is the Class A power value corresponding to the second period of the ith region, is the first type of power value corresponding to the Bth period of the ith region, is the second type of power value corresponding to the Bth period of the i-th region, It is the Class A power value corresponding to the first period of the Bth area.

[0094] In this embodiment, the power value may be information related to power demand, such as power consumption, power load, and power purchase price, and each type of power value is different.

[0095] Among them, the influencing factor of each influencing factor of power demand forecast is determined by the degree of influence of the influencing factor, and the influencing factor of each time period may be different. For example, the influencing factors of area 1 in time period 1 include temperature, humidity, and large-scale event factors, among which the degree of influence of temperature is greater, the degree of influence of humidity is second, and the degree of influence of large-scale events is the lowest. The influencing factors can be 0.7, 0.2, and 0.1. Each correction coefficient in the correction matrix is ​​determined by the influencing factor and the value of the influencing factor. For example, the temperature of area 1 in time period 1 is A, the humidity is B, and the large-scale event may be C. The correction coefficient of area 1 in time period 1 can be obtained by unifying A, B, and C numerically, and then adding the influencing factor to the unified numerical value. For example, the unified results are A1, B1, and C1, and the correction coefficient is 0.7A1+0.2B1+0.1C1.

[0096] The first power matrix is ​​obtained by correcting the initial power matrix through the correction matrix, and the correction matrix is:

[0097]

[0098] Among them, E2 i is the time period-correction coefficient matrix of the ith region, is the correction coefficient corresponding to the first period of the i-th region, is the correction coefficient corresponding to the second period of the ith region, is the correction coefficient corresponding to the Bth time period in the ith region.

[0099] The first power matrix is:

[0100]

[0101] Among them, H1 i is the first power matrix of the ith region, (E1 i ) T is the transposed matrix of the initial power matrix of the ith region, A is the power data, B is the time period, is the first type of power value corresponding to the first period of the ith region after correction, is the second type of power value corresponding to the first period after correction in the i-th region, is the Class A power value corresponding to the first period after correction in the i-th area, is the first type of power value corresponding to the second period after correction in the i-th region, is the second type of power value corresponding to the second period after correction in the i-th region, is the Class A power value corresponding to the second period after correction in the i-th area, is the first type of power value corresponding to the Bth period after correction in the i-th area, is the second type of power value corresponding to the Bth period after correction in the i-th area, It is the Class A power value corresponding to the first time period after correction for the Bth area.

[0102] The initial power matrix is ​​corrected by using a correction matrix to obtain a first power matrix, which can effectively improve the accuracy of the first power matrix.

[0103] In an optional embodiment, determining a special area based on all first power matrices includes:

[0104] determining a reference matrix based on all first power matrices;

[0105] Comparing each first power matrix with a reference matrix, and determining a special power value in each first power matrix that exceeds a preset range;

[0106] The number of special power values ​​in each first power matrix is ​​obtained, and the area corresponding to the first power matrix whose number exceeds a preset value is taken as a special area.

[0107] Optionally, the reference matrix is ​​obtained by weighted summing all first power matrices according to regional area and population density.

[0108] Among them, the preset range of each power data is different. For example, the preset range of power data 1 can be [15, 20], and the preset range of power data 2 can be [11, 18]. The specific situation needs to be set according to actual needs, and this embodiment is not limited here. The preset range is obtained by comprehensively considering the fluctuation range of power data under normal circumstances and the power data corresponding to the population. For example, a total of three years of historical power consumption data of the same period are used for prediction. During these three years, the annual growth rate of power data 1 is 15%. The population of the region this year conforms to the growth law and the other influencing factors this year have not changed significantly. The historical lowest data of power data 1 is a, the historical highest data is b, and the model prediction accuracy is 98%. Then the prediction range of power data 1 should be [0.98×1.15a, 1.02×1.15b]. By setting the preset range, the predicted data can be judged to obtain special values ​​that are inconsistent with the growth law and population.

[0109] Optionally, the power values ​​in the first power matrix that exceed the preset range are special power values. For example, the first power matrix of area 2 is a 3*3 matrix, wherein the first type of power value in the second row is 50, the first type of power value in the second row of the reference matrix is ​​15, the difference is 35, and the preset range is [-25, +18]. The first type of power value in the second row of the first power matrix exceeds the preset range, and 50 is used as a special power value. If the preset value is 3, and there are 5 special power values ​​in the first power matrix, the number of special power values ​​in the first power matrix exceeds the preset value, and area 2 is used as a special area.

[0110] In an optional embodiment, determining a reference matrix based on all first power matrices includes:

[0111] Determine the area and population density of the region corresponding to each first power matrix, and determine the weight of each first power matrix based on the area and population density;

[0112] A weighted sum of each type of power value in all first power matrices is calculated based on the weights, and a reference matrix is ​​determined based on the weighted sum.

[0113] In this embodiment, the weight is related to the area and population density of the region. If region 1 has three regions, the areas of the three regions are C1, C2, and C3, and the population densities are X1, X2, and X3, respectively, then the weight of each region is

[0114] Optionally, the reference matrix is ​​determined by the weight of each region and the first power matrix of each region. If region 1 is divided into three regions, the first power matrices of region 1, region 2, and region 3 are respectively, H1 1 、H1 2 、H1 3 , the weights are q1, q2, q3 respectively, then the reference matrix of region 1 is 1H1 1 +2H1 2 +3H1 3 .

[0115] The first power matrix of the region is obtained by predicting through historical electricity consumption data and the first prediction model, and all the first power matrices are comprehensively considered to determine a reference matrix, and the first power matrix of each region is compared with the reference matrix to determine one or more special regions, which can effectively determine the special regions and improve the accuracy of the calculation.

[0116] S130: Determine a second power matrix of the special area based on the second prediction model, and optimize the corresponding first power matrix based on the second power matrix to obtain a third power matrix.

[0117] The second prediction model is determined by randomly selecting prediction models other than the first prediction model from the prediction model database, but the second prediction model can also be determined by other methods, and this solution does not make specific restrictions here.

[0118] In an optional embodiment, determining a second power matrix of a special area based on a second prediction model includes:

[0119] Randomly select a prediction model from other prediction models in the prediction model library except the first prediction model as the second prediction model;

[0120] Based on the second prediction model and the historical power consumption data corresponding to the special area, the initial power matrix of the special area is obtained by prediction;

[0121] The initial power matrix of the special area is corrected based on the correction matrix corresponding to the special area to obtain a second power matrix of the special area.

[0122] Among them, the second power matrix is:

[0123]

[0124] Among them, H2 j is the second power matrix of the jth special area, is the first type of power value corresponding to the first time period after correction in the jth special area, is the second type of power value corresponding to the first period after correction in the jth special area, is the Class A power value corresponding to the first period after correction in the jth special area, is the first type of power value corresponding to the second period after correction in the jth special area, is the second type of power value corresponding to the second period after correction in the jth special area, is the Class A power value corresponding to the second period after correction in the jth special area, is the first type of power value corresponding to the Bth period after correction in the jth special area, is the second type of power value corresponding to the Bth period after correction in the jth special area, It is the A-th class power value corresponding to the B-th time period after correction for the j-th special area.

[0125] In an optional embodiment, the corresponding first power matrix is ​​optimized based on the second power matrix to obtain a third power matrix, including:

[0126] Obtaining model features of the second prediction model for the special area, as well as the number of special power values ​​in the second power matrix corresponding to the special area and the difference between each special power value and the corresponding value in the reference matrix;

[0127] constructing a difference matrix of the second power matrix based on the difference between each special power value and the corresponding value in the reference matrix;

[0128] Calculating a weight of the second power matrix based on the model characteristics and the number of special power values ​​in the second power matrix;

[0129] Optimizing the first power matrix based on the weight of the second power matrix to obtain a third power matrix;

[0130] The weights of the second power matrix are:

[0131]

[0132] Among them, Q j is the weight of the second power matrix corresponding to the jth special area, C is the computational complexity of the second prediction model, f is the frequency of use of the second prediction model in the region, k is the number of special power values ​​in the second power matrix, is the difference between the special power value in the ath row and bth column of the difference matrix and the corresponding value in the reference matrix, a is the ath row of the difference matrix, b is the bth column of the difference matrix, A is the total number of rows of the difference matrix, B is the total number of columns of the difference matrix, and k=AB;

[0133] The third power matrix is:

[0134]

[0135] Among them, H3 j is the three-power matrix of the jth special area, H2 j is the second power matrix of the jth special area, H1 j is the first power matrix of the jth special area, Q j is the weight value of the second power matrix corresponding to the j-th special area.

[0136] In this embodiment, the model characteristics are the computational complexity of the model and the usage frequency of the model.

[0137] The construction process of the difference matrix is ​​as follows:

[0138] First, the second power matrix is ​​processed, and other power values ​​except the special power values ​​in the second power matrix and the reference matrix are assigned 0;

[0139] Then, the row and column numbers of each special power value in the second power matrix are obtained as the marking numbers, and the reference matrix is ​​processed to retain the values ​​in the reference matrix except the values ​​whose row and column numbers are the marking numbers, and the remaining values ​​are assigned 0;

[0140] Finally, the processed reference matrix is ​​subtracted from the processed second power matrix to obtain a difference matrix.

[0141] S140: determining a first special value matrix of the special area based on historical power consumption data of the special area and a plurality of adjacent areas thereof, and optimizing the third power matrix based on the first special value matrix to obtain a final power matrix of the special area.

[0142] Among them, the adjacent area of ​​the special area refers to the area adjacent to the location of the special area, such as Figure 2 As shown, areas A1, A2, A3, A4, and A5 are all adjacent areas of area A.

[0143] In this embodiment, the first special value matrix of each special area is determined by the historical power consumption data of multiple adjacent areas and the historical power consumption data of the special area. The first special value matrix can reflect the difference between the special area and the adjacent areas in historical situations.

[0144] Multiple regions are obtained by population density and area division, and the historical electricity consumption data of each region is obtained. The power matrix and special regions are determined based on the randomly selected model and historical electricity consumption data. Then one is selected from other prediction models for re-prediction, and the results of the first prediction are optimized based on the results of the re-prediction. The historical power matrix and predicted power matrix of the special region are compared with the historical power matrix and predicted power matrix of the adjacent region to obtain two comparison results, and the optimized power matrix is ​​optimized again based on the two comparison results to obtain the final result. The prediction results are optimized twice based on the results obtained by different prediction models and the comparison with adjacent regions, ensuring the accuracy of the final power matrix.

[0145] In an optional embodiment, determining a first special value matrix of a special area based on historical electricity consumption data of the special area and a plurality of adjacent areas thereof includes:

[0146] Determine a plurality of adjacent areas of the special area, and determine a historical power matrix of the plurality of adjacent areas based on historical power consumption data of the plurality of adjacent areas;

[0147] Determine a historical power matrix for a special area based on historical power consumption data for the special area;

[0148] Compare the historical power matrix of a special area with the historical power matrices of multiple adjacent areas to obtain multiple historical difference matrices;

[0149] A first special value matrix for a special area is determined based on a plurality of historical difference value matrices.

[0150] For each special area, assuming that the special area A has five adjacent areas (A1, A2, A3, A4, A5), the first special value matrix of the special area is obtained as follows:

[0151] Step 1: Calculate the difference between the historical power matrix (A') of the special area and the historical power matrix (A1', A2', A3', A4', A5') of each adjacent area to obtain five historical difference matrices (ΔA1', ΔA2', ΔA3', ΔA4', ΔA5');

[0152] Step 2: Get the distance from the location center of each adjacent area to the location center of the special area, and determine the weight of the historical difference matrix of each adjacent area according to the distance:

[0153]

[0154] Where M is the number of adjacent regions of the region, m is the nth adjacent region of the special region, and ρ m is the weight of the historical difference matrix of the mth adjacent region, d m is the distance from the location center of the mth adjacent area to the location center of the special area;

[0155] Step 3: Superimpose each historical difference matrix according to the weight to obtain the first special value matrix of the special area. The first special value matrix of the special area is:

[0156]

[0157] Among them, T1 is the first special value matrix of the special area, ρ m is the weight of the mth adjacent region of the region, and ΔAm' is the historical difference matrix between the special region and the mth adjacent region.

[0158] It should be noted that the first special value matrix of the special area can also be obtained by other methods, and the specific obtaining method is not limited in this solution.

[0159] In an optional embodiment, optimizing the third power matrix based on the first special value matrix to obtain a final power matrix for the special area includes:

[0160] Acquire first power matrices of multiple adjacent areas of the special area, and compare the third power matrix with the first power matrices of the multiple adjacent areas to obtain multiple prediction difference matrices;

[0161] Determine a second special value matrix for the special area based on the plurality of prediction difference matrices;

[0162] Acquire the influencing factor change data, and determine the influence coefficient of each influencing factor based on the influencing factor change data;

[0163] If there is an influence coefficient beyond the preset range among the influence coefficients, determining a third prediction model based on the influence factors corresponding to the influence coefficients beyond the preset range;

[0164] Predicting the special area based on the third prediction model to obtain a fourth power matrix of the special area, and correcting the corresponding third power matrix based on the fourth power matrix to obtain a final power matrix of the special area;

[0165] If there is no influence coefficient beyond the preset range among the influence coefficients, the third power matrix is ​​optimized based on the first special value matrix and the second special value matrix to obtain a final power matrix for the special area;

[0166]

[0167] Among them, H j is the final power matrix of the jth special area, a is the ath row element of the first special value matrix and the second special value matrix of the jth special area, b is the bth column element of the first special value matrix and the second special value matrix of the jth special area, A is the power load data, B is the time period, is the influence factor of the xth influencing factor, ΔG(x) ab is the estimated impact of the xth influencing factor on the power data in the ath row and bth column, is the factor of the ath row and bth column in the first special value matrix of the jth special region, is the factor of the ath row and bth column in the second special value matrix of the jth special region, H3 j is the third power matrix of the j-th special area.

[0168] In this embodiment, the prediction difference matrix is ​​obtained by comparing the third power matrix of the special area with the first power matrix of each of its adjacent areas. If the special area has three adjacent areas, the special area has three prediction difference matrices. The prediction difference matrix can reflect the specific gap between the special area and each adjacent area in the predicted data.

[0169] In this embodiment, the method for obtaining the second special value matrix is ​​the same as the method for obtaining the first special value matrix. The second special value matrix represents the difference between the special region and the adjacent regions in the predicted data.

[0170] The change data of the influencing factors refers to the difference between the average value of the influencing factors in the historical power consumption data and the average value of the predicted influencing factors. For example, the average temperature in the historical power consumption data is 23.3°C, and the average future weather temperature in the forecast period is expected to be 25.2°C. The change data of the influencing factor (temperature) is 25.2-23.3=1.9°C. The change data of the influencing factors will have different impacts on different power data. The deep learning algorithm can be used to determine the estimated impact of the change data of each influencing factor on different power data in different time periods.

[0171] The influence coefficient of the influencing factor is determined by the change data of the influencing factor and the influencing factor of the influencing factor. The influencing factor refers to the importance of the corresponding influencing factor. The importance of the same influencing factor may be different in different time periods and seasons. For example, at noon 12:00-14:00, the importance of temperature is greater than humidity, but in the time period of 10:00-12:00 in the evening, the importance of humidity will be greater than temperature; in summer from 12:00-14:00, the influence of temperature is relatively large, but in winter from 12:00-14:00, the influence of temperature will be reduced accordingly. The importance of the same influencing factor will also vary in different regions due to different geographical environments.

[0172] The following describes the selection process of the third prediction model through a specific example:

[0173] There are three factors that will affect the electricity demand in region B. In time period 1, the influencing factors of the three influencing factors are: X1, X2, and X3. In time period 1, the change data of the three influencing factors are Y1, Y2, and Y3. The influence coefficients of the three influencing factors are Z1 (Z1 = X1 × Y1), Z2 (Z2 = X2 × Y2), and Z3 (Z3 = X3 × Y3). If the influence coefficient of the first influencing factor exceeds the preset range, a prediction model that can predict electricity through the first influencing factor and historical electricity consumption data is selected in the model database, and this prediction model is used as the third prediction model. If the influence coefficient of an influencing factor exceeds the preset range, it means that the influencing factor will have a greater impact on the predicted data. For example, a sudden rise in temperature in a certain area will cause the increase in electricity consumption to increase, exceeding the preset range.

[0174] In this embodiment, the weight of the fourth power matrix can be calculated through the computational complexity of the fourth power matrix and the usage frequency of the third prediction model corresponding to the fourth power matrix in the area, and the third power matrix is ​​corrected by the weight of the fourth power matrix to obtain the final power matrix.

[0175] In this embodiment, the first special value matrix is ​​obtained through historical electricity consumption data, and the second special value matrix is ​​obtained through predicted electricity consumption data. Comparing the first special value matrix with the second special value matrix can determine the gap between the special area and the adjacent areas under historical and predicted conditions. By comparing the historical situation with the current situation and optimizing the third power matrix based on the comparison results to obtain the final power matrix, the accuracy of the final power matrix can be effectively improved by considering the specific special conditions of the special area.

[0176] This scheme determines the influence coefficient of the influencing factor through the change data of the influencing factor, and then determines the specific optimization method through the influence coefficient. According to different optimization methods, the specific scheme is determined to optimize the third power matrix to obtain the final power matrix. The first optimization method selects a more suitable prediction model through special values ​​for prediction, and guarantees the accuracy of the final power matrix for the third power matrix based on the prediction results. The final power matrix is ​​obtained by comparing the special values ​​obtained from the historical power data with the special values ​​obtained from the predicted power data, thereby ensuring the accuracy of the final power matrix.

[0177] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0178] Corresponding to the above embodiment, a method for predicting power demand based on two-stage optimization is described. Figure 3 A schematic structural diagram of a power demand forecasting device based on two-stage optimization provided by an embodiment of the present invention is shown. For ease of explanation, only the parts related to the embodiment of the present invention are shown.

[0179] See also Figure 3 , a power demand forecasting device based on two-stage optimization in an embodiment of the present invention may include:

[0180] A division module 31, used for dividing the area to be predicted into multiple areas based on population density and area;

[0181] A determination module 32 is used to obtain historical power consumption data of each area, determine a first power matrix of the area based on the historical power consumption data and a first prediction model, and determine a special area based on all first power matrices;

[0182] An optimization module 33 is used to determine a second power matrix of the special area based on the second prediction model, and optimize the corresponding first power matrix based on the second power matrix to obtain a third power matrix;

[0183] The optimization module 33 is further used to determine a first special value matrix of the special area based on historical power consumption data of the special area and multiple adjacent areas, and optimize the third power matrix based on the first special value matrix to obtain a final power matrix of the special area.

[0184] In some embodiments, the partitioning module 31 is specifically configured to:

[0185] The area of ​​the region to be predicted is obtained, and the region to be predicted is initially divided based on the area to obtain multiple initial regions;

[0186] The population density of each initial area is obtained, and the initial area is adjusted based on the population density to obtain multiple areas to be predicted.

[0187] In some embodiments, the determination module 32 is specifically configured to:

[0188] For each region, perform the following steps:

[0189] Obtain historical electricity consumption data of the area within a preset period, and randomly select a prediction model from a prediction model library as a first prediction model; the prediction model library includes multiple prediction models;

[0190] Based on the first prediction model and the historical electricity consumption data, a prediction is performed to obtain an initial power matrix of the area;

[0191] An influencing factor of each influencing factor of power demand forecasting is determined, and a correction matrix is ​​determined based on the influencing factor. The initial power matrix is ​​corrected based on the correction matrix to obtain a first power matrix.

[0192] In some embodiments, the determination module 32 is specifically configured to:

[0193] determining a reference matrix based on all first power matrices;

[0194] Comparing each first power matrix with a reference matrix, and determining a special power value in each first power matrix that exceeds a preset range;

[0195] The number of special power values ​​in each first power matrix is ​​obtained, and the area corresponding to the first power matrix whose number exceeds a preset value is taken as a special area.

[0196] In some embodiments, the determination module 32 is specifically configured to:

[0197] Determine the area and population density of the region corresponding to each first power matrix, and determine the weight of each first power matrix based on the area and population density;

[0198] A weighted average value of each type of power value in all first power matrices is calculated based on the weight value, and a reference matrix is ​​determined based on the weighted average value.

[0199] In some embodiments, the optimization module 33 is specifically configured to:

[0200] Randomly select a prediction model from other prediction models in the prediction model library except the first prediction model as the second prediction model;

[0201] Based on the second prediction model and the historical power consumption data corresponding to the special area, the initial power matrix of the special area is obtained by prediction;

[0202] The initial power matrix of the special area is corrected based on the correction matrix corresponding to the special area to obtain a second power matrix of the special area.

[0203] In some embodiments, the optimization module 33 is specifically configured to:

[0204] Obtaining model features of the second prediction model for the special area, as well as the number of special power values ​​in the second power matrix corresponding to the special area and the difference between each special power value and the corresponding value in the reference matrix;

[0205] constructing a difference matrix of the second power matrix based on the difference between each special power value and the corresponding value in the reference matrix;

[0206] Calculating a weight of the second power matrix based on the model characteristics and the number of special power values ​​in the second power matrix;

[0207] Optimizing the first power matrix based on the weight of the second power matrix to obtain a third power matrix;

[0208] The weights of the second power matrix are:

[0209]

[0210] Among them, Q j is the weight of the second power matrix corresponding to the jth special area, C is the computational complexity of the second prediction model, f is the frequency of use of the second prediction model in the region, k is the number of special power values ​​in the second power matrix, is the difference between the special power value in the ath row and bth column of the difference matrix and the corresponding value in the reference matrix, a is the ath row of the difference matrix, b is the bth column of the difference matrix, A is the total number of rows of the difference matrix, B is the total number of columns of the difference matrix, and k=AB;

[0211] The third power matrix is:

[0212]

[0213] Among them, H3 j is the three-power matrix of the jth special area, H2 jis the second power matrix of the jth special area, H1 j is the first power matrix of the jth special area, Q j is the weight value of the second power matrix corresponding to the j-th special area.

[0214] In some embodiments, the optimization module 33 is specifically configured to:

[0215] Determine a plurality of adjacent areas of the special area, and determine a historical power matrix of the plurality of adjacent areas based on historical power consumption data of the plurality of adjacent areas;

[0216] Determine a historical power matrix for a special area based on historical power consumption data for the special area;

[0217] Compare the historical power matrix of a special area with the historical power matrices of multiple adjacent areas to obtain multiple historical difference matrices;

[0218] A first special value matrix for a special area is determined based on a plurality of historical difference value matrices.

[0219] In some embodiments, the optimization module 33 is specifically configured to:

[0220] Acquire first power matrices of multiple adjacent areas of the special area, and compare the third power matrix with the first power matrices of the multiple adjacent areas to obtain multiple prediction difference matrices;

[0221] Determine a second special value matrix for the special area based on the plurality of prediction difference matrices;

[0222] Acquire the influencing factor change data, and determine the influence coefficient of each influencing factor based on the influencing factor change data;

[0223] If there is an influence coefficient beyond the preset range among the influence coefficients, determining a third prediction model based on the influence factors corresponding to the influence coefficients beyond the preset range;

[0224] Predicting the special area based on the third prediction model to obtain a fourth power matrix of the special area, and correcting the corresponding third power matrix based on the fourth power matrix to obtain a final power matrix of the special area;

[0225] If there is no influence coefficient beyond the preset range among the influence coefficients, the third power matrix is ​​optimized based on the first special value matrix and the second special value matrix to obtain a final power matrix for the special area;

[0226]

[0227] Among them, H jis the final power matrix of the jth special area, a is the ath row element of the first special value matrix and the second special value matrix of the jth special area, b is the bth column element of the first special value matrix and the second special value matrix of the jth special area, A is the power load data, B is the time period, is the influence factor of the xth influencing factor, ΔG(x) ab is the estimated impact of the xth influencing factor on the power data in the ath row and bth column, is the factor of the ath row and bth column in the first special value matrix of the jth special region, is the factor of the ath row and bth column in the second special value matrix of the jth special region, H3 j is the third power matrix of the j-th special area.

[0228] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0229] Those of ordinary skill in the art will appreciate that the templates, units, and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0230] If the module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned embodiments of the power demand forecasting method based on two-stage optimization 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 that can carry the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory, random access memory, electrical carrier signal, telecommunication signal and software distribution medium.

[0231] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A method for predicting power demand based on two-stage optimization, characterized in that: include: The predicted area is divided into multiple regions based on population density and area; For each area, historical power consumption data of the area is obtained, a first power matrix of the area is determined based on the historical power consumption data and a first prediction model, and a special area is determined based on all first power matrices; Determine a second power matrix of the special area based on the second prediction model, and optimize the corresponding first power matrix based on the second power matrix to obtain a third power matrix; A first special value matrix of the special area is determined based on historical power consumption data of the special area and a plurality of adjacent areas thereof, and a final power matrix of the special area is obtained by optimizing the third power matrix based on the first special value matrix.

2. The method for predicting power demand based on two-stage optimization according to claim 1, characterized in that: The predicted area is divided into multiple regions based on population density and area, including: Acquiring the area of ​​the region to be predicted, and initially dividing the region to be predicted based on the area to obtain a plurality of initial regions; The population density of each initial area is obtained, and the initial area is adjusted based on the population density to obtain a plurality of areas to be predicted.

3. The method for predicting power demand based on two-stage optimization according to claim 1, characterized in that: The step of acquiring historical power consumption data of each region and determining a first power matrix of the region based on the historical power consumption data and a first prediction model includes: For each region, perform the following steps: Obtaining historical electricity consumption data of the area within a preset period, and randomly selecting a prediction model from a prediction model library as the first prediction model; the prediction model library includes multiple prediction models; Perform prediction based on the first prediction model and the historical electricity consumption data to obtain an initial power matrix of the area; An influencing factor of each influencing factor of power demand forecasting is determined, and a correction matrix is ​​determined based on the influencing factor, and the initial power matrix is ​​corrected based on the correction matrix to obtain a first power matrix.

4. The method for predicting power demand based on two-stage optimization according to claim 1, characterized in that: The determining of the special area based on all the first power matrices comprises: determining a reference matrix based on all first power matrices; Comparing each first power matrix with the reference matrix, and determining a special power value in each first power matrix that exceeds a preset range; The number of special power values ​​in each of the first power matrices is obtained, and the area corresponding to the first power matrix where the number exceeds a preset value is taken as a special area.

5. The method for predicting power demand based on two-stage optimization according to claim 4, characterized in that: The determining of the reference matrix based on all the first power matrices comprises: Determine the area and population density of the region corresponding to each first power matrix, and determine the weight of each first power matrix based on the area and the population density; A weighted average value of each type of power value in all first power matrices is calculated based on the weight value, and a reference matrix is ​​determined based on the weighted average value.

6. The method for predicting power demand based on two-stage optimization according to claim 1, characterized in that: The determining the second power matrix of the special area based on the second prediction model includes: Randomly select a prediction model from other prediction models in the prediction model library except the first prediction model as the second prediction model; Performing prediction based on the second prediction model and the historical power consumption data corresponding to the special area to obtain an initial power matrix of the special area; The initial power matrix of the special area is corrected based on the correction matrix corresponding to the special area to obtain a second power matrix of the special area.

7. The method for predicting power demand based on two-stage optimization according to claim 1, characterized in that: The step of optimizing the corresponding first power matrix based on the second power matrix to obtain a third power matrix includes: Acquire the model characteristics of the second prediction model of the special area, the number of special power values ​​in the second power matrix corresponding to the special area, and the difference between each special power value and the corresponding value in the reference matrix; Constructing a difference matrix of the second power matrix based on the difference between each of the special power values ​​and the corresponding value in the reference matrix; Calculating a weight of the second power matrix based on the model characteristics and the number of special power values ​​in the second power matrix; Optimizing the first power matrix based on the weight of the second power matrix to obtain a third power matrix; The weight of the second power matrix is: Among them, Q j is the weight of the second power matrix corresponding to the jth special area, C is the computational complexity of the second prediction model, f is the frequency of use of the second prediction model in the area, k is the number of special power values ​​in the second power matrix, is the difference between the special power value in the ath row and bth column of the difference matrix and the corresponding value in the reference matrix, a is the ath row of the difference matrix, b is the bth column of the difference matrix, A is the total number of rows of the difference matrix, B is the total number of columns of the difference matrix, and k=AB; The third power matrix is: Among them, H3 j is the three-power matrix of the jth special area, H2 j is the second power matrix of the jth special area, H1 j is the first power matrix of the jth special area, Q j is the weight value of the second power matrix corresponding to the j-th special area.

8. The method for predicting power demand based on two-stage optimization according to claim 1, characterized in that: The determining of a first special value matrix of the special area based on the historical electricity consumption data of the special area and a plurality of adjacent areas includes: Determine a plurality of adjacent areas of the special area, and determine a historical power matrix of the plurality of adjacent areas based on historical power consumption data of the plurality of adjacent areas; Determining a historical power matrix of the special area based on the historical power consumption data of the special area; Comparing the historical power matrix of the special area with the historical power matrices of the plurality of adjacent areas to obtain a plurality of historical difference matrices; A first special value matrix for the special area is determined based on the multiple historical difference matrices.

9. The method for predicting power demand based on two-stage optimization according to claim 1, characterized in that: The step of optimizing the third power matrix based on the first special value matrix to obtain a final power matrix for the special area includes: Acquire first power matrices of a plurality of adjacent areas of the special area, and compare the third power matrix with the first power matrices of the plurality of adjacent areas to obtain a plurality of prediction difference matrices; Determine a second special value matrix for the special area based on the multiple prediction difference matrices; Acquire influencing factor change data, and determine the influence coefficient of each influencing factor based on the influencing factor change data; If there is an influence coefficient out of the preset range among the influence coefficients, determining a third prediction model based on the influence factors corresponding to the influence coefficient out of the preset range; Predicting the special area based on the third prediction model to obtain a fourth power matrix of the special area, and correcting the corresponding third power matrix based on the fourth power matrix to obtain a final power matrix of the special area; If there is no influence coefficient beyond the preset range among the influence coefficients, optimizing the third power matrix based on the first special value matrix and the second special value matrix to obtain a final power matrix for the special area; Among them, H j is the final power matrix of the jth special area, a is the ath row element of the first special value matrix and the second special value matrix of the jth special area, b is the bth column element of the first special value matrix and the second special value matrix of the jth special area, A is the power load data, B is the time period, is the influence factor of the xth influencing factor, ΔG(x) ab is the estimated impact of the xth influencing factor on the power data in the ath row and bth column, is the element in the ath row and bth column of the first special value matrix of the jth special region, is the element in the ath row and bth column of the second special value matrix of the jth special region, H3 j is the third power matrix of the j-th special area.

10. A power demand forecasting device based on two-stage optimization, characterized in that: include: A division module is used to divide the predicted area into multiple regions based on population density and area; A determination module, configured to obtain, for each region, historical power consumption data of the region, determine a first power matrix of the region based on the historical power consumption data and a first prediction model, and determine a special region based on all first power matrices; an optimization module, configured to determine a second power matrix of the special area based on a second prediction model, and optimize the corresponding first power matrix based on the second power matrix to obtain a third power matrix; The optimization module is further used to determine a first special value matrix of the special area based on historical electricity consumption data of the special area and its multiple adjacent areas, and optimize the third power matrix based on the first special value matrix to obtain a final power matrix of the special area.

Images