Power load prediction method and system for regions with imperfect information

Through nonlinear least squares method and data from common alternative regions, a polynomial, exponential and logarithmic function power load prediction model is constructed, which solves the problem of insufficient accuracy of power load prediction in traditional methods and improves power load prediction in areas with incomplete information Accuracy and reliability.

CN119093316APending Publication Date: 2024-12-06CHINA AGRI UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410904080.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-08
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

In areas where information is not sound, traditional power load prediction methods ignore the influence of nonlinear data and cannot accurately capture the complex relationship between power demand and nonlinear changing data, resulting in insufficient prediction accuracy.

Method used

By using the nonlinear least squares method, the nonlinear sequence data of the unsound area and the preset nonlinear association database are obtained, the common alternative areas associated with it are determined, and the historical nonlinear sequence data and actual power load data of the common alternative areas are constructed. Power load prediction models, including polynomials, exponentials, and logarithmic function expressions, optimize model parameters to improve prediction accuracy.

Benefits of technology

By capturing the relationship between power demand and nonlinear data, the accuracy and reliability of power load prediction are improved, and the problem of low accuracy of power load prediction in areas with incomplete information is solved, ensuring the stability of power supply and resource allocation of power supply effectiveness.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119093316A_ABST
    Figure CN119093316A_ABST
Patent Text Reader

Abstract

The invention provides a power load prediction method and system for an information-imperfect region. The method comprises the following steps: acquiring nonlinear sequence data of the information-imperfect region; according to the non-linear sequence data of the information imperfect region and a preset non-linear association database, obtaining a common substitution region associated with the information imperfect region; obtaining to-be-predicted nonlinear sequence data of the common substitution region, and obtaining power load prediction data of the common substitution region by using a pre-constructed power load prediction model according to the to-be-predicted nonlinear sequence data, taking the power load prediction data of the common substitution region as the power load prediction data of the region with unsound information; according to the method and the device, the power load prediction result of the region with imperfect information can be indirectly obtained by performing power load prediction on the common replacement region, and the problem of low power load prediction accuracy caused by incomplete power load data in the region with imperfect information is solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of electric load forecasting, and in particular to an electric load forecasting method and system in areas with incomplete information. Background Art

[0002] At present, with the expansion of the scale of power systems and the increasing demand for electricity, accurate prediction of power load is crucial to ensure the stable operation of power systems and effective resource allocation. Especially in the field of power system management and planning, due to the inability to accurately obtain power data, traditional power load forecasting methods will no longer be applicable when facing power load forecasting in areas with incomplete information. Therefore, accurate power load forecasting is crucial to ensure the stability and reliability of power supply. However, due to various reasons such as insufficient power monitoring equipment or data sharing restrictions, there may be some areas where accurate power data cannot be obtained. In these areas with incomplete information, the use of traditional extrapolation methods for power load forecasting will no longer be applicable.

[0003] In order to overcome the above difficulties, it is considered to perform common substitution processing on areas close to areas with incomplete information, so as to indirectly obtain the power load forecast results of the target area. When considering the establishment of power load forecasting models for common substitution areas close to areas with incomplete information, traditional methods usually consider linear data for modeling, ignoring the impact of nonlinear data on power demand, and therefore cannot capture the complex relationship between power demand and nonlinear change data, which leads to the problem of insufficient accuracy in predicting future power loads in existing technologies. Summary of the invention

[0004] In order to solve the problem that in the prior art, when establishing a power load forecasting model for a common alternative region associated with an area with incomplete information, only linear data is considered for modeling, and the influence of nonlinear data on power demand is ignored, so the complex relationship between power demand and nonlinear change data cannot be captured, which leads to the problem of insufficient accuracy in predicting power load in the prior art, the present invention proposes a power load forecasting method for an area with incomplete information, comprising:

[0005] Acquiring nonlinear series data in areas with incomplete information;

[0006] According to the nonlinear sequence data of the region with incomplete information and a preset nonlinear association database, a common alternative region associated with the region with incomplete information is obtained;

[0007] Acquire the nonlinear sequence data to be predicted of the common substitution area, and obtain the power load prediction data of the common substitution area by using a pre-built power load prediction model according to the nonlinear sequence data to be predicted, and use the power load prediction data of the common substitution area as the power load prediction data of the area with incomplete information;

[0008] The power load forecasting model is constructed using the least squares method, with the historical nonlinear sequence data of the common substitution area as the input of training data and the actual power load data corresponding to the historical nonlinear sequence data as the output of training data.

[0009] Optionally, the power load forecasting model includes the following construction process:

[0010] Acquire historical nonlinear sequence data of the common substitution area and actual power load data corresponding to the historical nonlinear sequence data;

[0011] Using the historical nonlinear sequence data of the common replacement region as input of training data;

[0012] Using actual power load data corresponding to the historical nonlinear sequence data as output of training data;

[0013] According to the input and output of the training data, a power load prediction model is constructed using the least squares method.

[0014] Optionally, constructing a power load prediction model using a least squares method based on the input and output of the training data includes:

[0015] According to the input and output of the training data, a polynomial load forecasting model composed of a polynomial function expression is trained to obtain a trained polynomial load forecasting model; wherein the polynomial function expression includes at least one learnable parameter for fitting the relationship between the input and output of the training data in the polynomial load forecasting model;

[0016] According to the input and output of the training data, an exponential load forecasting model composed of an exponential function expression is trained to obtain a trained exponential load forecasting model; wherein the exponential function expression includes at least one network parameter for fitting the relationship between the input and output of the training data in the exponential load forecasting model;

[0017] According to the input and output of the training data, a logarithmic load forecasting model composed of a logarithmic function expression is trained to obtain a trained logarithmic load forecasting model; wherein the logarithmic function expression includes at least one model parameter for fitting the relationship between the input and output of the training data in the logarithmic load forecasting model;

[0018] Based on the trained polynomial load forecasting model, the trained exponential load forecasting model and the trained logarithmic load forecasting model, a power load forecasting model is determined using a least squares method.

[0019] Optionally, the determining of the power load forecasting model by using the least squares method based on the trained polynomial load forecasting model, the trained exponential load forecasting model and the trained logarithmic load forecasting model includes:

[0020] Input the input of the training data as an input item into the trained polynomial load forecasting model to obtain a corresponding polynomial load forecasting value; optimize the learnable parameter value in the polynomial load forecasting model with the goal of minimizing the residual sum of squares between the actual power load data corresponding to the input of the training data and the polynomial load forecasting value to obtain a final polynomial load forecasting model;

[0021] Input the input of the training data as an input item into the trained exponential load prediction model to obtain a corresponding exponential load prediction value; optimize the network parameter value in the exponential load prediction model with the goal of minimizing the residual sum of squares between the actual power load data corresponding to the input of the training data and the exponential load prediction value to obtain a final exponential load prediction model;

[0022] Input the input of the training data as an input item into the trained logarithmic load prediction model to obtain a corresponding logarithmic load prediction value; optimize the model parameter values ​​in the logarithmic load prediction model with the goal of minimizing the residual sum of squares between the actual power load data corresponding to the input of the training data and the logarithmic load prediction value to obtain a final logarithmic load prediction model;

[0023] An electric power load forecasting model is determined based on the final polynomial load forecasting model, the final exponential load forecasting model and the final logarithmic load forecasting model.

[0024] Optionally, determining the power load forecasting model based on the final polynomial load forecasting model, the final exponential load forecasting model and the final logarithmic load forecasting model includes:

[0025] Respectively calculating the fitting error values ​​corresponding to the final polynomial load forecasting model, the final exponential load forecasting model and the final logarithmic load forecasting model;

[0026] A load forecasting model with the smallest fitting error value is selected as the power load forecasting model from the fitting error value of the final polynomial load forecasting model, the fitting error value of the final exponential load forecasting model and the fitting error value of the final logarithmic load forecasting model.

[0027] Optionally, the historical nonlinear sequence data of the common substitution region includes the following acquisition process:

[0028] Acquire time series data of the common substitution area and actual power load data corresponding to the time series data;

[0029] Calculating the correlation between the time series data and the actual power load data using the Spearman rank correlation coefficient method;

[0030] Time series data with correlation greater than a preset threshold are selected as the historical nonlinear series data of the common replacement area.

[0031] Optionally, the expression corresponding to the Spearman rank correlation coefficient method is as follows:

[0032]

[0033] Where ρ represents the correlation between the time series data and the corresponding actual power load data; n represents the number of data points in the time series data; d i It represents the rank difference between the time series data corresponding to the i-th data point and the actual power load data; i=1…n; a represents the coefficient value.

[0034] Optionally, the nonlinear series data includes one or more of the following: climate data and economic indicator data;

[0035] The climate data includes one or more of the following: monthly average temperature data, monthly average humidity data, monthly average wind speed data and monthly average air pressure data;

[0036] The economic indicator data include one or more of the following: monthly GDP, monthly per capita GDP, monthly employment population growth rate and monthly total population growth rate.

[0037] Based on the same inventive concept, the present invention also provides a power load forecasting system for areas with incomplete information, comprising:

[0038] Data acquisition module: used to obtain nonlinear sequence data in areas with incomplete information;

[0039] Common substitution module: used for obtaining common substitution regions associated with the region with incomplete information according to the nonlinear sequence data of the region with incomplete information and a preset nonlinear association database;

[0040] Load forecasting module: used to obtain the nonlinear sequence data to be predicted of the common substitution area, and obtain the power load forecasting data of the common substitution area by using the pre-built power load forecasting model according to the nonlinear sequence data to be predicted, and use the power load forecasting data of the common substitution area as the power load forecasting data of the area with incomplete information;

[0041] Among them, the power load forecasting model in the load forecasting module is constructed using the least squares method, with the historical nonlinear sequence data of the common substitution area as the input of training data and the actual power load data corresponding to the historical nonlinear sequence data as the output of training data.

[0042] Optionally, the power load forecasting model in the load forecasting module includes the following construction process:

[0043] Acquire historical nonlinear sequence data of the common substitution area and actual power load data corresponding to the historical nonlinear sequence data;

[0044] Using the historical nonlinear sequence data of the common replacement region as input of training data;

[0045] Using actual power load data corresponding to the historical nonlinear sequence data as output of training data;

[0046] According to the input and output of the training data, a power load prediction model is constructed using the least squares method.

[0047] Optionally, the load forecasting module constructs a power load forecasting model using a least squares method according to the input and output of the training data, including:

[0048] According to the input and output of the training data, a polynomial load forecasting model composed of a polynomial function expression is trained to obtain a trained polynomial load forecasting model; wherein the polynomial function expression includes at least one learnable parameter for fitting the relationship between the input and output of the training data in the polynomial load forecasting model;

[0049] According to the input and output of the training data, an exponential load forecasting model composed of an exponential function expression is trained to obtain a trained exponential load forecasting model; wherein the exponential function expression includes at least one network parameter for fitting the relationship between the input and output of the training data in the exponential load forecasting model;

[0050] According to the input and output of the training data, a logarithmic load forecasting model composed of a logarithmic function expression is trained to obtain a trained logarithmic load forecasting model; wherein the logarithmic function expression includes at least one model parameter for fitting the relationship between the input and output of the training data in the logarithmic load forecasting model;

[0051] Based on the trained polynomial load forecasting model, the trained exponential load forecasting model and the trained logarithmic load forecasting model, a power load forecasting model is determined using a least squares method.

[0052] Optionally, the load forecasting module determines the power load forecasting model using the least squares method based on the trained polynomial load forecasting model, the trained exponential load forecasting model and the trained logarithmic load forecasting model, including:

[0053] Input the input of the training data as an input item into the trained polynomial load forecasting model to obtain a corresponding polynomial load forecasting value; optimize the learnable parameter value in the polynomial load forecasting model with the goal of minimizing the residual sum of squares between the actual power load data corresponding to the input of the training data and the polynomial load forecasting value to obtain a final polynomial load forecasting model;

[0054] Input the input of the training data as an input item into the trained exponential load prediction model to obtain a corresponding exponential load prediction value; optimize the network parameter value in the exponential load prediction model with the goal of minimizing the residual sum of squares between the actual power load data corresponding to the input of the training data and the exponential load prediction value to obtain a final exponential load prediction model;

[0055] Input the input of the training data as an input item into the trained logarithmic load prediction model to obtain a corresponding logarithmic load prediction value; optimize the model parameter values ​​in the logarithmic load prediction model with the goal of minimizing the residual sum of squares between the actual power load data corresponding to the input of the training data and the logarithmic load prediction value to obtain a final logarithmic load prediction model;

[0056] An electric power load forecasting model is determined based on the final polynomial load forecasting model, the final exponential load forecasting model and the final logarithmic load forecasting model.

[0057] Optionally, the load forecasting module determines a power load forecasting model based on the final polynomial load forecasting model, the final exponential load forecasting model and the final logarithmic load forecasting model, including:

[0058] Respectively calculating the fitting error values ​​corresponding to the final polynomial load forecasting model, the final exponential load forecasting model and the final logarithmic load forecasting model;

[0059] A load forecasting model with the smallest fitting error value is selected as the power load forecasting model from the fitting error value of the final polynomial load forecasting model, the fitting error value of the final exponential load forecasting model and the fitting error value of the final logarithmic load forecasting model.

[0060] Optionally, the historical nonlinear sequence data of the common substitution area in the load forecasting module includes the following acquisition process:

[0061] Acquire time series data of the common substitution area and actual power load data corresponding to the time series data;

[0062] Calculating the correlation between the time series data and the actual power load data using the Spearman rank correlation coefficient method;

[0063] Time series data with correlation greater than a preset threshold are selected as the historical nonlinear series data of the common replacement area.

[0064] Optionally, the expression corresponding to the Spearman rank correlation coefficient method in the load forecasting module is as follows:

[0065]

[0066] Where ρ represents the correlation between the time series data and the corresponding actual power load data; n represents the number of data points in the time series data; d i It represents the rank difference between the time series data corresponding to the i-th data point and the actual power load data; i=1…n; a represents the coefficient value.

[0067] Optionally, the nonlinear sequence data in the data acquisition module includes one or more of the following: climate data and economic indicator data;

[0068] The climate data includes one or more of the following: monthly average temperature data, monthly average humidity data, monthly average wind speed data and monthly average air pressure data;

[0069] The economic indicator data include one or more of the following: monthly GDP, monthly per capita GDP, monthly employment population growth rate and monthly total population growth rate.

[0070] Compared with the prior art, the present invention has the following beneficial effects:

[0071] The present invention provides a method and system for predicting electric load in areas with incomplete information, comprising: obtaining nonlinear sequence data of areas with incomplete information; obtaining common alternative areas associated with the areas with incomplete information according to the nonlinear sequence data of the areas with incomplete information and a preset nonlinear association database; obtaining nonlinear sequence data to be predicted of the common alternative areas, and obtaining electric load prediction data of the common alternative areas according to the nonlinear sequence data to be predicted by using a pre-constructed electric load prediction model; using the electric load prediction data of the common alternative areas as the electric load prediction data of the areas with incomplete information; wherein the electric load prediction model is constructed by using the least square method with the historical nonlinear sequence data of the common alternative areas as the input of training data and the actual electric load data corresponding to the historical nonlinear sequence data as the output of training data; the present application selects areas associated with the nonlinear sequence data of areas with incomplete information as common alternative areas, and predicts the electric load of the common alternative areas by capturing the relationship between electric demand and nonlinear data, so as to indirectly obtain the electric load prediction results of areas with incomplete information, which is conducive to solving the problem of low accuracy of electric load prediction caused by incomplete electric load data in areas with incomplete information. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 A schematic flow chart of a method for predicting power load in areas with incomplete information provided by the present invention;

[0073] Figure 2 A schematic diagram of the overall process of a method for predicting power load in areas with incomplete information provided by the present invention;

[0074] Figure 3 A schematic diagram of the structure of a power load forecasting system for areas with incomplete information provided by the present invention;

[0075] Figure 4 The present invention provides a schematic structural diagram of an electronic device. DETAILED DESCRIPTION

[0076] The present invention provides a method and system for predicting power load in areas with incomplete information. The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings.

[0077] Embodiment 1:

[0078] A method for predicting power load in areas with incomplete information, the flow chart is as follows Figure 1 As shown, including:

[0079] Step 1: Obtain nonlinear sequence data in areas with incomplete information;

[0080] Step 2: Based on the nonlinear sequence data of the area with incomplete information and the preset nonlinear association database, obtain the common alternative areas associated with the area with incomplete information;

[0081] Step 3: Obtain the nonlinear sequence data to be predicted in the common substitution area, and obtain the power load prediction data of the common substitution area by using the pre-built power load prediction model according to the nonlinear sequence data to be predicted, and use the power load prediction data of the common substitution area as the power load prediction data of the area with incomplete information;

[0082] Among them, the power load forecasting model uses the historical nonlinear sequence data of common substitution areas as the input of training data, and the actual power load data corresponding to the historical nonlinear sequence data as the output of training data, and is constructed using the least squares method.

[0083] Generally, the problem of power load forecasting refers to the analysis and modeling of historical power load data to predict the load demand of the power system in the future. Its essence is to establish a forecasting model through historical data, and then use the forecasting model to forecast the power load in the future. Accurate power load forecasting is crucial to ensure the stability and reliability of power supply. It not only facilitates power dispatchers to effectively allocate resources and plan power generation capacity, but also helps to ensure the stability of power system operation. Especially in the field of power system planning and operation, load forecasting is a key input for long-term planning of power systems. Its forecast accuracy is closely related to the success of power system planning and operation. When the accuracy of power load is high, it helps to better integrate intermittent renewable energy (such as wind and solar energy) and ensure the reliability and flexibility of the power system. However, due to various reasons such as insufficient power monitoring equipment or data sharing restrictions, it is impossible to accurately obtain historical power load data, and the fluctuation of power load is often affected by nonlinear change data (such as climate conditions and economic situation), resulting in the traditional power load forecasting method being no longer applicable when facing power load forecasting in areas with incomplete information. To solve the above problems, it is necessary to establish a nonlinear model to predict the changes in power load. For example, nonlinear sequence data related to power load data generally include: load fluctuation data, meteorological data, economic indicator data and equipment operation data, etc. The present invention takes climate data and economic indicator data that have a greater impact on power load fluctuations as an example, and predicts the power load according to the changes in economic situation and climate factors in areas with incomplete information.

[0084] First, it is necessary to obtain climate data and economic indicator data for areas where the information for power load forecasting is incomplete. For example, climate data may include one or more of the following: monthly average temperature data, monthly average humidity data, monthly average wind speed data and monthly average air pressure data, which can be obtained by visiting the official website of the meteorological bureau in the area or using the global climate data platform; economic indicator data may include one or more of the following: monthly GDP, monthly per capita GDP, monthly employment population growth rate and monthly total population growth rate, which can be obtained by visiting the official website of the statistical department of the area, market analysis reports of industry associations, etc.

[0085] After obtaining the climate data and economic indicator data of the area with incomplete information, it is necessary to perform common substitution processing with the economic, climate, and electricity data of the area with similar economy and climate to the area with incomplete information, and by introducing the nonlinear least squares method, combining the historical climate data, historical economic indicator data, historical power load and other multi-dimensional data of the common substitution area with similar economy and climate, a more accurate and flexible power load forecasting model is constructed, among which the historical power load data can choose to collect the monthly power consumption data of the common substitution area;

[0086] In one implementation, the above-mentioned power load forecasting model includes the following construction process:

[0087] Obtain historical nonlinear series data of common substitution areas and actual power load data corresponding to the historical nonlinear series data;

[0088] The historical nonlinear sequence data of the common replacement regions are used as the input of training data;

[0089] The actual power load data corresponding to the historical nonlinear sequence data is used as the output of the training data;

[0090] According to the input and output of training data, the least square method is used to build the power load forecasting model.

[0091] By establishing the above-mentioned power load forecasting model, the power load forecasting results in areas with incomplete target information can be indirectly obtained, solving the problem of low accuracy of power load forecasting of traditional forecasting models in areas with incomplete information and when considering economic and climatic factors. The model aims to overcome the limitations of traditional methods, improve the accuracy and reliability of power load forecasting, and enable power system planners and operators to better respond to future changes in power demand.

[0092] In this implementation, the historical nonlinear sequence data of the common replacement region includes the following acquisition process:

[0093] Obtain the time series data of common substitution areas and the actual power load data corresponding to the time series data;

[0094] The Spearman rank correlation coefficient method is used to calculate the correlation between time series data and actual power load data;

[0095] Time series data with correlation greater than a preset threshold are selected as historical nonlinear series data of common alternative regions.

[0096] For example, the expression corresponding to the above Spearman rank correlation coefficient method is as follows:

[0097]

[0098] Where ρ represents the correlation between the time series data and the corresponding actual power load data; n represents the number of data points in the time series data; d i represents the rank difference between the time series data corresponding to the i-th data point and the actual power load data; i = 1…n; a represents the coefficient value. Taking the calculation of the rank difference of two variables X and Y as an example, the observations of the two variables X and Y are sorted and a rank is assigned to each observation. For example, the rank of X after sorting is The rank after sorting Y is By observing each pair of observations, the rank difference is expressed as:

[0099]

[0100] Among them, d i represents the rank difference of the i-th data point; represents the rank of the i-th data point in variable X; Represents the rank of the i-th data point in variable Y.

[0101] According to the definition of the Spearman rank correlation coefficient, its value range is between -1 and 1, where -1 indicates a complete negative correlation, 1 indicates a complete positive correlation, and 0 indicates no rank relationship. Based on the Spearman rank correlation coefficient between each indicator and the power load, the indicators with a high correlation with the power load among the climate factors and economic indicators are screened out, and the historical data of these indicators and the historical power load data are used as backup data sets for the power load prediction model.

[0102] In this implementation, the present invention uses the Spearman rank correlation coefficient method to calculate the correlation between the time series data of the common substitution area and the power load data. By correctly reflecting the correlation between the two variables when a nonlinear relationship is presented, and by the principle of ranking the correlation, it is not very sensitive to the obvious outliers in the data set, which is conducive to reducing the impact of outliers on the results. In addition, the Spearman correlation coefficient method is a non-parametric statistical method, which means that it does not require the data to satisfy the normal distribution or other specific distribution assumptions. This is especially important for power load data. In the relationship between power load and other factors such as weather and economic activities, this relationship may be nonlinear, and this type of data is often not normally distributed, but exhibits the characteristics of a peak and a thick tail. Therefore, by using the Spearman rank correlation coefficient method to obtain the correlation between the time series data of the common substitution area and the power load data, it is possible to identify the climate data and economic indicator data that are highly correlated with the power load, thereby screening out the most valuable features for model prediction, avoiding the introduction of noise by irrelevant variables, and predicting the power load based on the correlation results, which is conducive to the optimization and scheduling of the power system, improving the prediction accuracy of the model, and ensuring the stability and reliability of the power supply.

[0103] In this implementation, the historical climate and economic indicator data selected by the Spearman rank correlation coefficient method are used as independent variables (also called input of training data), and the historical actual power load data are used as dependent variables (also called output of training data). The nonlinear least squares method is used to select polynomial function expressions, exponential function expressions, and logarithmic function expressions to establish a climate-economic power load forecasting model. The specific process may include:

[0104] According to the input and output of the training data, a polynomial load forecasting model composed of a polynomial function expression is trained to obtain a trained polynomial load forecasting model; wherein the polynomial function expression includes at least one learnable parameter for fitting the relationship between the input and output of the training data in the polynomial load forecasting model;

[0105] According to the input and output of the training data, an exponential load forecasting model composed of an exponential function expression is trained to obtain a trained exponential load forecasting model; wherein the exponential function expression includes at least one network parameter for fitting the relationship between the input and output of the training data in the exponential load forecasting model;

[0106] According to the input and output of the training data, a logarithmic load forecasting model composed of a logarithmic function expression is trained to obtain a trained logarithmic load forecasting model; wherein the logarithmic function expression includes at least one model parameter for fitting the relationship between the input and output of the training data in the logarithmic load forecasting model;

[0107] Based on the trained polynomial load forecasting model, the trained exponential load forecasting model and the trained logarithmic load forecasting model, the power load forecasting model is determined using the least squares method.

[0108] For example, the above polynomial function expression can generally be expressed as:

[0109] f(x)=a m x m +a m-1 x m-1 +…+a p x p +…+a 2 x 2 +a 1 x+a 0

[0110] Where x represents the independent variable; f(x) represents the corresponding dependent variable; x m represents the m-th order function; a m Represents the coefficient value of the m-th function term; m represents the highest power of the polynomial function expression; x prepresents the p-order function term; a p Represents the coefficient value of the p-order function term; p = 1…m; a 0 represents the coefficient value of the constant term;

[0111] Taking the temperature data in the climate data as an example, a polynomial load prediction model between the temperature data and the power load data is established. Since the relationship between the temperature data and the power load data is often not a simple linear relationship, as the temperature changes, the power load may increase first and then decrease, or present a more complex fluctuation. The quadratic polynomial can fit such a U-shaped or inverted U-shaped curve, while the cubic polynomial can better capture curves with more complex shapes, such as S-shaped or wavy curves. Therefore, the present invention can establish a functional relationship between the temperature and the power load through quadratic and cubic polynomials, wherein the expression of the polynomial load prediction model composed of quadratic polynomials is as follows:

[0112] L′ 1 =a 1 T 2 +b 1 T+c 1

[0113] Among them, L′ 1 represents the power load data in the polynomial load forecasting model composed of quadratic polynomials; T represents the temperature data; a 1 b represents the learnable parameter value of the second-order term of the quadratic polynomial; 1 represents the learnable parameter value of the first-order term of the quadratic polynomial; c 1 The constant term representing the quadratic polynomial can learn the parameter value;

[0114] The expression of the polynomial load forecasting model composed of cubic polynomials is as follows:

[0115] L″ 1 =a′ 1 T 3 +b′ 1 T 2 +c 1 'T+d' 1

[0116] Among them, L 1 represents the power load data in the polynomial load forecasting model composed of cubic polynomials; T represents the temperature data; a′ 1 b′ represents the learnable parameter value of the third-order term of the cubic polynomial; 1 represents the learnable parameter value of the quadratic term of the cubic polynomial; c 1 ′ represents the learnable parameter value of the first-order term of the cubic polynomial; d′ 1 The constant term representing the cubic polynomial can learn the parameter value;

[0117] For example, the expression of the exponential load prediction model composed of the exponential function expression between the above-mentioned temperature data and power load data is as follows:

[0118]

[0119] In the formula, L′ 2 represents the power load data in the exponential load forecasting model composed of exponential function expressions; T represents the temperature data; a 2 The first network parameter value representing the exponential function expression; b 2 The value of the second network parameter representing the exponential function expression.

[0120] For example, the expression of the logarithmic load prediction model composed of the logarithmic function expression between the above temperature data and the power load is as follows:

[0121] L′ 3 =a 3 lnb 3 T

[0122] In the formula, L′ 3 represents the power load data in the logarithmic load forecasting model composed of logarithmic function expressions; T represents the temperature data; a 3 b represents the value of the first model parameter of the logarithmic function expression; 3 The value of the second model parameter representing the logarithmic function expression.

[0123] In this implementation, the process of determining the power load forecasting model using the least squares method based on the trained polynomial load forecasting model, the trained exponential load forecasting model, and the trained logarithmic load forecasting model may include:

[0124] Input the input of the training data as an input item to the trained polynomial load forecasting model to obtain the corresponding polynomial load forecasting value; optimize the learnable parameter value in the polynomial load forecasting model with the goal of minimizing the residual sum of squares between the actual power load data corresponding to the input of the training data and the polynomial load forecasting value to obtain the final polynomial load forecasting model;

[0125] Input the input of the training data as an input item to the trained exponential load forecasting model to obtain the corresponding exponential load forecast value; optimize the network parameter value in the exponential load forecasting model with the goal of minimizing the residual sum of squares between the actual power load data corresponding to the input of the training data and the exponential load forecast value to obtain the final exponential load forecasting model;

[0126] Input the input of the training data as an input item to the trained logarithmic load forecasting model to obtain the corresponding logarithmic load forecast value; optimize the model parameter value in the logarithmic load forecasting model with the goal of minimizing the residual sum of squares between the actual power load data corresponding to the input of the training data and the logarithmic load forecast value to obtain the final logarithmic load forecasting model;

[0127] Based on the final polynomial load forecasting model, the final exponential load forecasting model and the final logarithmic load forecasting model, the power load forecasting model is determined.

[0128] The least square method used in the present invention is to find the parameters of the best fitting model by minimizing the residual square sum between the actual value in the observed data and the model predicted value. It only requires that the curve to be fitted can reflect the basic trend of the experimental test data, and does not require that the curve must pass through each data point. This method is regarded as one of the most reliable methods for finding a set of unknown quantities from a set of measured values. The least square method can be used to easily obtain unknown data. The result is usually called the best value or the most reliable value. By selecting and comparing nonlinear functions such as polynomials, exponentials and logarithmic function expressions, the model structure that best describes the complex nonlinear relationship between data can be found, further improving the accuracy of the prediction. Therefore, the power load prediction model constructed by fitting nonlinear functions by the least square method can provide more accurate, flexible and robust prediction results, help optimize the scheduling of generator sets, reduce unnecessary power generation costs and carbon emissions, and is essential for ensuring the stable operation of the power system, improving efficiency and reducing environmental impact.

[0129] For example, suppose there is a set of observation data {T 1 ,…,T n} and {L 1 ,…,L n}, where {T 1 ,…,T n} represents the independent variable consisting of time series data containing n data points, {L 1 ,…,L n} represents the dependent variable consisting of the actual power load data corresponding to the time series data containing n data points. The expression of the residual sum of squares between the actual value and the model prediction value is:

[0130]

[0131] Where s represents the residual sum of squares between the actual value and the model prediction value; n is the number of data points in the time series data; L i represents the actual value of the model, that is, the actual power load data of the time series data corresponding to the i-th data point; i = 1…n; is the model prediction value, that is, the power load prediction value corresponding to the ith data point; the least squares method is used to minimize the residual sum of squares, and the parameters of the function to be fitted are obtained, thereby obtaining the specific expressions of each nonlinear function that best suits the data set as the power load prediction model for the common replacement area.

[0132] In this implementation, the process of determining the power load forecasting model based on the final polynomial load forecasting model, the final exponential load forecasting model and the final logarithmic load forecasting model may include:

[0133] Calculate the fitting error values ​​corresponding to the final polynomial load forecasting model, the final exponential load forecasting model and the final logarithmic load forecasting model respectively;

[0134] The load forecasting model with the smallest fitting error value is selected as the power load forecasting model from the fitting error values ​​of the final polynomial load forecasting model, the fitting error values ​​of the final exponential load forecasting model and the fitting error values ​​of the final logarithmic load forecasting model.

[0135] In this implementation, the present invention uses the root mean square error (RMSE) and the coefficient of determination (R 2 ) is used to calculate the fitting error value of each model and evaluate the performance of the power load forecasting model.

[0136] The root mean square error is the square root of the mean of the squares of the differences between the actual observations and the model predictions. It is used to measure the prediction error of the model. The smaller its value, the better the prediction performance of the model. The calculation formula of the root mean square error is as follows:

[0137]

[0138] In the formula, RMSE represents the fitting error value of the model calculated by the root mean square error; n represents the number of data points of the time series data; L i is the actual value of the model, that is, the actual power load data of the time series data corresponding to the i-th data point; is the model prediction value, that is, the power load prediction value corresponding to the i-th data point; i = 1…n.

[0139] The coefficient of determination is a statistical indicator used to describe the degree to which the model explains the variability of the dependent variable. It is the proportion of the actual data variance that is explained by the model. 2 The value range of R is between 0 and 1. The closer it is to 1, the better the model explains the variability of the observed data, while the closer it is to 0, the worse the model effect. 2When it is equal to 1, it means that the model fits the data perfectly, and when R 2 When it is 0, the model has no ability to explain the observed data, and its calculation formula is as follows:

[0140]

[0141] In the formula, R 2 represents the fitting error value of the model calculated using the coefficient of determination; n represents the number of data points in the time series data; L i is the actual value of the model, that is, the actual power load data of the time series data corresponding to the i-th data point; is the model prediction value, that is, the power load prediction value corresponding to the i-th data point; i = 1…n; is the mean of the actual value of the model, that is, the mean of the actual power load data.

[0142] The present invention uses the root mean square error and the coefficient of determination to evaluate the performance of each prediction model, which can improve the overall view of the prediction model performance: RMSE focuses on the size of the prediction error, while R 2 Focus on the model's ability to explain data changes. Using any one indicator alone may be misleading. For example, a high R 2 A low RMSE value does not necessarily mean a small prediction error, and a low RMSE value does not necessarily mean that the model explains most of the variation in the data. The combination of the two can provide a more accurate assessment of model performance. 2 Selecting the nonlinear function with the smallest fitting error value to establish the climate-economic power load forecasting model of the present invention is beneficial to improving the accuracy of power load forecasting.

[0143] The present invention utilizes the nonlinear least squares method to fit the complex nonlinear correlation between climate-economic data and electric power data, utilizes several common nonlinear functions to perform least squares fitting, and compares the fitting effects of different nonlinear functions by observing the fitting error values ​​of each function, thereby determining the electric power load forecasting model with the best fitting effect. This is beneficial to capturing the complex nonlinear relationship between data. Compared with simple linear models, this sophisticated modeling method can often more accurately reflect the dynamics of the real world, and can select the model that best reflects the inherent laws of the data, deeply understand how climate-economic factors affect changes in electric power, and the intensity and direction of these influences, which not only improves the prediction ability of the model, but also enhances the understanding of the driving mechanism of electric power demand, thereby improving the accuracy of the prediction, and has an important positive impact on the optimization of the power system, the effective allocation of resources, and the response to climate change.

[0144] Based on the above method, a power load forecasting model for common substitution areas based on climate data and economic indicator data is established. The climate data and economic indicator data for a period of time in the future are taken as input, the predicted power load value of the model is taken as output, and the measured value is taken as a reference to test the prediction accuracy of the model. For models with poor prediction accuracy, considering increasing the amount of data to update the model parameters and then correcting the accuracy. The power load predicted by the model can indirectly obtain the power load forecast results of the target area with incomplete information.

[0145] The present invention aims at the fact that in the prior art, when establishing a power load forecasting model for a common alternative region associated with an area with incomplete information, only linear data is considered for modeling, and the influence of nonlinear data on power demand is ignored. Therefore, the complex relationship between power demand and nonlinear change data cannot be captured, which leads to the problem of insufficient accuracy in predicting power load in the prior art. A power load forecasting method based on nonlinear least squares method is proposed. The method can establish a power load forecasting model through historical climate data, historical economic indicator data and historical power load data of the common alternative region, so as to realize accurate prediction of power load. The overall process diagram is shown as follows Figure 2As shown in the figure, firstly, the multidimensional data related to the power load of the target information-imperfect region with similar economy and climate are collected and common substitution processing is performed, including historical power load data, climate data and economic indicator data, and the Spearman rank correlation coefficient method is used to perform correlation analysis on the data, and the climate data and economic indicator data with high correlation with the power load are screened out to form a backup data set; secondly, the nonlinear least squares method is used to construct a power load forecasting model considering climate-economy, and the commonly used nonlinear functions such as polynomial function, exponential function and logarithmic function are considered to establish a power load forecasting model associated with climate-economy, and the fitting error values ​​of the prediction models established by different nonlinear functions are compared, and the model with the best fitting effect is selected as the power load forecasting model for the common substitution region; finally, the selected power load forecasting model is used to predict the power load of the common substitution region in the future, and the predicted value is compared with the measured value of the power load to test the accuracy of the model prediction. For the model with poor prediction accuracy, considering increasing the amount of data to update the model parameters and then correcting the accuracy, the power load predicted by the model can indirectly know the power load forecasting result of the target information-imperfect region. The present invention uses the common substitution method to solve the problem that historical load data is difficult to obtain in areas with incomplete information. The power load forecasting method is solved by considering the common substitution processing of economic, climate data and power load data of areas with similar economic and climate conditions. It can more comprehensively and accurately capture the complex nonlinear relationship between power demand and economic and climate factors. Compared with the traditional linear model, this method is more outstanding in predicting power load when facing power load fluctuations caused by changes in climate conditions and economic situations. In general, the prediction model of the present invention is more accurate and reliable in predicting future power loads.

[0146] Embodiment 2:

[0147] The present invention based on the same inventive concept also provides a power load forecasting system for areas with incomplete information, the structural composition diagram is as follows: Figure 3 As shown, including:

[0148] Data acquisition module: used to obtain nonlinear sequence data in areas with incomplete information;

[0149] Common substitution module: used to obtain common substitution areas associated with the areas with incomplete information based on the nonlinear sequence data of the areas with incomplete information and a preset nonlinear association database;

[0150] Load forecasting module: used to obtain the nonlinear sequence data to be predicted in the common substitution area, and obtain the power load forecasting data of the common substitution area by using the pre-built power load forecasting model according to the nonlinear sequence data to be predicted, and use the power load forecasting data of the common substitution area as the power load forecasting data of the area with incomplete information;

[0151] Among them, the power load forecasting model in the load forecasting module uses the historical nonlinear sequence data of common substitution areas as the input of training data and the actual power load data corresponding to the historical nonlinear sequence data as the output of training data, and is constructed using the least squares method.

[0152] For example, the nonlinear sequence data in the above-mentioned data acquisition module includes one or more of the following: climate data and economic indicator data;

[0153] The above-mentioned climate data includes one or more of the following: monthly average temperature data, monthly average humidity data, monthly average wind speed data and monthly average air pressure data;

[0154] The above economic indicator data include one or more of the following: monthly GDP, monthly per capita GDP, monthly employment population growth rate and monthly total population growth rate.

[0155] In one implementation, the historical nonlinear sequence data of the common substitution area in the above-mentioned load forecasting module includes the following acquisition process:

[0156] Obtain the time series data of common substitution areas and the actual power load data corresponding to the time series data;

[0157] The Spearman rank correlation coefficient method is used to calculate the correlation between time series data and actual power load data;

[0158] Time series data with correlation greater than a preset threshold are selected as historical nonlinear series data of common alternative regions.

[0159] For example, the expression corresponding to the Spearman rank correlation coefficient method in the above load forecasting module is as follows:

[0160]

[0161] Where ρ represents the correlation between the time series data and the corresponding actual power load data; n represents the number of data points in the time series data; d i It represents the rank difference between the time series data corresponding to the i-th data point and the actual power load data; i=1…n; a represents the coefficient value.

[0162] In this implementation, the power load forecasting model in the above-mentioned load forecasting module includes the following construction process:

[0163] Obtain historical nonlinear series data of common substitution areas and actual power load data corresponding to the historical nonlinear series data;

[0164] The historical nonlinear sequence data of the common replacement regions are used as the input of training data;

[0165] The actual power load data corresponding to the historical nonlinear sequence data is used as the output of the training data;

[0166] According to the input and output of training data, the least square method is used to build the power load forecasting model.

[0167] In this implementation, the load forecasting module uses the least square method to construct a power load forecasting model based on the input and output of the training data, including:

[0168] According to the input and output of the training data, a polynomial load forecasting model composed of a polynomial function expression is trained to obtain a trained polynomial load forecasting model; wherein the polynomial function expression includes at least one learnable parameter for fitting the relationship between the input and output of the training data in the polynomial load forecasting model;

[0169] According to the input and output of the training data, an exponential load forecasting model composed of an exponential function expression is trained to obtain a trained exponential load forecasting model; wherein the exponential function expression includes at least one network parameter for fitting the relationship between the input and output of the training data in the exponential load forecasting model;

[0170] According to the input and output of the training data, a logarithmic load forecasting model composed of a logarithmic function expression is trained to obtain a trained logarithmic load forecasting model; wherein the logarithmic function expression includes at least one model parameter for fitting the relationship between the input and output of the training data in the logarithmic load forecasting model;

[0171] Based on the trained polynomial load forecasting model, the trained exponential load forecasting model and the trained logarithmic load forecasting model, the power load forecasting model is determined using the least squares method.

[0172] In this implementation, the load forecasting module uses the least square method to determine the power load forecasting model based on the trained polynomial load forecasting model, the trained exponential load forecasting model and the trained logarithmic load forecasting model, including:

[0173] Input the input of the training data as an input item to the trained polynomial load forecasting model to obtain the corresponding polynomial load forecasting value; optimize the learnable parameter value in the polynomial load forecasting model with the goal of minimizing the residual sum of squares between the actual power load data corresponding to the input of the training data and the polynomial load forecasting value to obtain the final polynomial load forecasting model;

[0174] Input the input of the training data as an input item to the trained exponential load forecasting model to obtain the corresponding exponential load forecast value; optimize the network parameter value in the exponential load forecasting model with the goal of minimizing the residual sum of squares between the actual power load data corresponding to the input of the training data and the exponential load forecast value to obtain the final exponential load forecasting model;

[0175] Input the input of the training data as an input item to the trained logarithmic load forecasting model to obtain the corresponding logarithmic load forecast value; optimize the model parameter value in the logarithmic load forecasting model with the goal of minimizing the residual sum of squares between the actual power load data corresponding to the input of the training data and the logarithmic load forecast value to obtain the final logarithmic load forecasting model;

[0176] Based on the final polynomial load forecasting model, the final exponential load forecasting model and the final logarithmic load forecasting model, the power load forecasting model is determined.

[0177] In this implementation, the load forecasting module determines the power load forecasting model based on the final polynomial load forecasting model, the final exponential load forecasting model and the final logarithmic load forecasting model, including:

[0178] Calculate the fitting error values ​​corresponding to the final polynomial load forecasting model, the final exponential load forecasting model and the final logarithmic load forecasting model respectively;

[0179] The load forecasting model with the smallest fitting error value is selected as the power load forecasting model from the fitting error values ​​of the final polynomial load forecasting model, the fitting error values ​​of the final exponential load forecasting model and the fitting error values ​​of the final logarithmic load forecasting model.

[0180] Embodiment 3:

[0181] like Figure 4 As shown, the present invention also provides an electronic device, which may be a computer device, a single-chip device, an intelligent mobile device, etc. The electronic device in this embodiment may include a processor, a memory, a transceiver component, etc. The memory, the processor, and the transceiver component are connected via a bus; the memory may be used to store an execution program, and an exemplary execution program may include instructions; the processor is used to execute the instructions stored in the memory. The memory may also be used to store data, which may be called and / or modified when the instructions are executed.

[0182] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions in a storage medium to implement the corresponding method flow or corresponding function, so as to implement the steps of a method for predicting power load in an area with incomplete information in the above-mentioned embodiment.

[0183] Embodiment 4:

[0184] Based on the same inventive concept, the present invention also provides a readable storage medium, specifically an electronic device readable storage medium (Memory), which is a memory device in an electronic device for storing programs and data. It can be understood that the storage medium here can include both built-in storage media in electronic devices and, of course, extended storage media supported by electronic devices. The storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by a processor are also stored in the storage space, and these instructions can be one or more execution programs (including program codes). It should be noted that the storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor loads and executes one or more instructions stored in the storage medium, which can implement the steps of a method for predicting power load in an area with incomplete information in the above embodiment.

[0185] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0186] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0187] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0188] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0189] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit its protection scope. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that after reading the present invention, those skilled in the art can still make various changes, modifications or equivalent substitutions to the specific implementation methods of the application, but these changes, modifications or equivalent substitutions are all within the protection scope of the claims to be approved.

Claims

1. A method for predicting power load in areas with incomplete information, characterized in that: include: Acquiring nonlinear series data in areas with incomplete information; According to the nonlinear sequence data of the region with incomplete information and a preset nonlinear association database, a common alternative region associated with the region with incomplete information is obtained; Acquire the nonlinear sequence data to be predicted of the common substitution area, and obtain the power load prediction data of the common substitution area by using a pre-built power load prediction model according to the nonlinear sequence data to be predicted, and use the power load prediction data of the common substitution area as the power load prediction data of the area with incomplete information; The power load forecasting model is constructed using the least squares method, with the historical nonlinear sequence data of the common substitution area as the input of training data and the actual power load data corresponding to the historical nonlinear sequence data as the output of training data.

2. The method according to claim 1, characterized in that The power load forecasting model includes the following construction process: Acquire historical nonlinear sequence data of the common substitution area and actual power load data corresponding to the historical nonlinear sequence data; Using the historical nonlinear sequence data of the common replacement region as input of training data; Using actual power load data corresponding to the historical nonlinear sequence data as output of training data; According to the input and output of the training data, a power load prediction model is constructed using the least squares method.

3. The method according to claim 2, characterized in that The method of constructing a power load prediction model using the least squares method according to the input and output of the training data includes: According to the input and output of the training data, a polynomial load forecasting model composed of a polynomial function expression is trained to obtain a trained polynomial load forecasting model; wherein the polynomial function expression includes at least one learnable parameter for fitting the relationship between the input and output of the training data in the polynomial load forecasting model; According to the input and output of the training data, an exponential load forecasting model composed of an exponential function expression is trained to obtain a trained exponential load forecasting model; wherein the exponential function expression includes at least one network parameter for fitting the relationship between the input and output of the training data in the exponential load forecasting model; According to the input and output of the training data, a logarithmic load forecasting model composed of a logarithmic function expression is trained to obtain a trained logarithmic load forecasting model; wherein the logarithmic function expression includes at least one model parameter for fitting the relationship between the input and output of the training data in the logarithmic load forecasting model; Based on the trained polynomial load forecasting model, the trained exponential load forecasting model and the trained logarithmic load forecasting model, a power load forecasting model is determined using a least squares method.

4. The method according to claim 3, characterized in that The method of determining the power load forecasting model by using the least square method based on the trained polynomial load forecasting model, the trained exponential load forecasting model and the trained logarithmic load forecasting model includes: Input the input of the training data as an input item into the trained polynomial load forecasting model to obtain a corresponding polynomial load forecasting value; optimize the learnable parameter value in the polynomial load forecasting model with the goal of minimizing the residual sum of squares between the actual power load data corresponding to the input of the training data and the polynomial load forecasting value to obtain a final polynomial load forecasting model; Input the input of the training data as an input item into the trained exponential load prediction model to obtain a corresponding exponential load prediction value; optimize the network parameter value in the exponential load prediction model with the goal of minimizing the residual sum of squares between the actual power load data corresponding to the input of the training data and the exponential load prediction value to obtain a final exponential load prediction model; Input the input of the training data as an input item into the trained logarithmic load prediction model to obtain a corresponding logarithmic load prediction value; optimize the model parameter values ​​in the logarithmic load prediction model with the goal of minimizing the residual sum of squares between the actual power load data corresponding to the input of the training data and the logarithmic load prediction value to obtain a final logarithmic load prediction model; An electric power load forecasting model is determined based on the final polynomial load forecasting model, the final exponential load forecasting model and the final logarithmic load forecasting model.

5. The method according to claim 4, characterized in that The determining of the power load forecasting model based on the final polynomial load forecasting model, the final exponential load forecasting model and the final logarithmic load forecasting model comprises: Respectively calculating the fitting error values ​​corresponding to the final polynomial load forecasting model, the final exponential load forecasting model and the final logarithmic load forecasting model; A load forecasting model with the smallest fitting error value is selected as the power load forecasting model from the fitting error value of the final polynomial load forecasting model, the fitting error value of the final exponential load forecasting model and the fitting error value of the final logarithmic load forecasting model.

6. The method according to claim 2, characterized in that The acquisition process of the historical nonlinear series data of the common substitution region includes the following steps: Acquire time series data of the common substitution area and actual power load data corresponding to the time series data; Calculating the correlation between the time series data and the actual power load data using the Spearman rank correlation coefficient method; Time series data with correlation greater than a preset threshold are selected as the historical nonlinear series data of the common replacement area.

7. The method according to claim 6, characterized in that The expression corresponding to the Spearman rank correlation coefficient method is as follows: Where ρ represents the correlation between the time series data and the corresponding actual power load data; n represents the number of data points in the time series data; d i It represents the rank difference between the time series data corresponding to the i-th data point and the actual power load data; i=1…n; a represents the coefficient value.

8. The method according to any one of claims 1 to 7, characterized in that: The non-linear series data includes one or more of the following: climate data and economic indicator data; The climate data includes one or more of the following: monthly average temperature data, monthly average humidity data, monthly average wind speed data and monthly average air pressure data; The economic indicator data include one or more of the following: monthly GDP, monthly per capita GDP, monthly employment population growth rate and monthly total population growth rate.

9. A power load forecasting system for areas with incomplete information, characterized in that: include: Data acquisition module: used to obtain nonlinear sequence data in areas with incomplete information; Common substitution module: used for obtaining common substitution regions associated with the region with incomplete information according to the nonlinear sequence data of the region with incomplete information and a preset nonlinear association database; Load forecasting module: used to obtain the nonlinear sequence data to be predicted of the common substitution area, and obtain the power load forecasting data of the common substitution area by using the pre-built power load forecasting model according to the nonlinear sequence data to be predicted, and use the power load forecasting data of the common substitution area as the power load forecasting data of the area with incomplete information; Among them, the power load forecasting model in the load forecasting module is constructed using the least squares method, with the historical nonlinear sequence data of the common substitution area as the input of training data and the actual power load data corresponding to the historical nonlinear sequence data as the output of training data.

10. The system according to claim 9, characterized in that The nonlinear sequence data in the data acquisition module includes one or more of the following: climate data and economic indicator data; The climate data includes one or more of the following: monthly average temperature data, monthly average humidity data, monthly average wind speed data and monthly average air pressure data; The economic indicator data include one or more of the following: monthly GDP, monthly per capita GDP, monthly employment population growth rate and monthly total population growth rate.