Power consumption prediction method and system based on multi-dimensional data and industry characteristics

Abstract

The invention discloses an electricity consumption prediction method and system based on multi-dimensional data and industry characteristics, and belongs to the field of data processing and analysis, and the method comprises the steps: collecting historical electricity consumption data of a green low-carbon industry in a certain region and multi-dimensional influence factor data of the historical electricity consumption data, the green low-carbon industry comprises a clean energy industry, an electric automobile industry and an energy storage industry; preprocessing the collected multi-dimensional influence factor data, and constructing an influence factor index library and an industry power consumption database; based on the influence factor index database and the industry power consumption database, performing power consumption characteristic analysis on the clean energy industry, the electric vehicle industry and the energy storage industry to obtain power consumption characteristics of each industry; according to the electricity consumption characteristics and historical electricity consumption data of each industry, green low-carbon industry electricity consumption contribution rate prediction and electricity consumption demand prediction are carried out on each industry; and carrying out weighted summation on the power consumption contribution rate prediction and the power consumption demand prediction of each industry, and calculating to obtain a power consumption prediction value of the whole green low-carbon industry.

Description

Technical Field

[0001] The present invention not only relates to the field of data processing and analysis, but specifically relates to a method and system for predicting electricity consumption based on multidimensional data and industry characteristics. Background Art

[0002] At present, the prediction of electricity consumption in green and low-carbon industries mainly relies on single factor or limited factor prediction method, traditional time series prediction method and machine learning prediction method. These methods have their own characteristics, but they all have obvious limitations. Although the single factor or limited factor prediction method has a simple model, it ignores other important factors affecting electricity consumption, such as policy changes and industry characteristics, resulting in inaccurate prediction results. Traditional time series prediction methods mainly focus on the time series characteristics of historical electricity consumption data and lack comprehensive consideration of other influencing factors. Although the machine learning prediction method can handle more complex data and consider more influencing factors, its prediction effect is also challenged due to the complexity of electricity consumption in green and low-carbon industries, as well as issues such as data availability and quality.

[0003] The core deficiencies of existing technologies in the prediction of electricity consumption in green and low-carbon industries are mainly reflected in insufficient data dimensions, insufficient analysis of industry characteristics, insufficient data fusion technology, and insufficient model construction and verification. First, existing prediction methods often only consider a limited number of factors, ignoring other important factors that may affect electricity consumption, resulting in incomplete and inaccurate prediction results. Secondly, various industries under the green and low-carbon industry have different electricity consumption characteristics and patterns, but existing prediction methods often do not conduct in-depth characteristic analysis of these industries, resulting in prediction results that cannot accurately reflect the actual situation of each industry. In addition, when processing multi-dimensional data, existing prediction methods lack effective data fusion technology, resulting in the correlation and complementarity between data not being fully utilized. Finally, existing prediction methods also have deficiencies in model construction and verification. There is often a trade-off between the complexity and accuracy of the model, and insufficient model verification may also lead to reduced reliability of prediction results.

[0004] Prior art document CN113962364A discloses a multi-factor electricity load forecasting method based on deep learning, which first completes data acquisition and storage, including electricity load data and environmental impact data; and preprocesses and standardizes the data based on abnormal data detection, autoregressive interpolation and sequence data normalization of the k-nearest neighbor algorithm and the improved DBSCAN algorithm; then proposes an improved CNN-LSTM electricity load forecasting model, which first uses a CNN feature extraction module to learn the local features of the input data; then inputs it into the LSTM sequence learning model to extract the sequence feature information of the input data; at the same time, the self-attention mechanism is introduced into the LSTM to learn the features of the LSTM hidden layer, and the extraction of key features is achieved by assigning different attention weights to improve the final prediction accuracy; finally, electricity load forecasting is performed. Its shortcomings lie in the limitations of data dimensions. It mainly focuses on electricity load forecasting. Although it introduces some environmental factors (such as temperature, humidity, wind power, etc.), it fails to comprehensively consider multi-dimensional data such as economy, policy and industry characteristics, making it difficult to comprehensively characterize the electricity demand of green and low-carbon industries. At the same time, it lacks quantitative analysis of the electricity contribution rate of various industries under the green and low-carbon industry, and the exploration of industry-specific electricity consumption patterns and the construction of targeted models are relatively insufficient, making it difficult to adapt to the complex needs of different green industries. Summary of the invention

[0005] The present invention aims to construct a green and low-carbon industry electricity consumption prediction system and method based on multi-dimensional data fusion and industry characteristic analysis. By integrating economic, meteorological, policy, industry characteristics and other data, advanced data fusion technology is used to explore the correlation and complementarity between data, and in-depth analysis of the electricity consumption characteristics and patterns of each green and low-carbon industry is conducted to construct a high-precision prediction model, so as to achieve accurate prediction of electricity consumption of various industries under the green and low-carbon industry, thereby facilitating efficient energy utilization, energy conservation and emission reduction, and sustainable development of the industry, and providing a scientific basis for government decision-making, enterprise production plan optimization, etc.

[0006] In order to solve the deficiencies in the prior art, the present invention provides a method and system for predicting electricity consumption based on multi-dimensional data and industry characteristics.

[0007] The present invention adopts the following technical solution.

[0008] A first aspect of the present invention provides a method for predicting power consumption based on multidimensional data and industry characteristics, comprising the following steps:

[0009] Collect historical electricity consumption data of a region's green and low-carbon industries and their multi-dimensional influencing factor data. Green and low-carbon industries include clean energy, electric vehicles, and energy storage. Multi-dimensional influencing factor data include economic factors, meteorological factors, and policy factors. Each factor also contains multiple influencing factors.

[0010] Pre-process the collected multi-dimensional influencing factor data to build an influencing factor index database and an industry electricity consumption database;

[0011] Based on the influencing factor index library and industry electricity consumption database, the electricity consumption characteristics of the clean energy industry, electric vehicle industry and energy storage industry are analyzed to obtain the electricity consumption characteristics of each industry;

[0012] According to the electricity consumption characteristics and historical electricity consumption data of each industry, the green and low-carbon industry electricity consumption contribution rate and electricity demand forecast are carried out for each industry;

[0013] The electricity consumption contribution rate forecast and electricity demand forecast of each industry are weighted and summed to calculate the electricity consumption forecast value of the overall green and low-carbon industry.

[0014] Optionally, each factor further includes multiple influencing factors, specifically:

[0015] Economic factors include return on investment, cost-benefit ratio, construction cost, operating cost, driving range, energy consumption per 100 kilometers, battery cost, charging facility construction cost, energy storage cost and energy storage efficiency;

[0016] Meteorological factors include temperature, humidity, precipitation, wind speed, and sunshine duration;

[0017] Policy factors include electricity price policies, policy subsidies, industry development policy documents, industry planning white papers and industry statistical yearbooks.

[0018] Optionally, the preprocessing of the collected multi-dimensional influencing factor data includes detecting outliers and filling missing values, specifically:

[0019] Use Z-score to detect outliers. Suppose the original data set is X={X1,X2,...,X n}, where n is the total number of indicators, where n i Represents X i The data volume of the indicator, the mean μ and standard deviation σ of each indicator are:

[0020]

[0021] For each data point x i , calculate its Z-score value:

[0022]

[0023] When | Z i |>3, it is considered that x i is an outlier;

[0024] For the detected outliers, appropriate interpolation methods are used according to the data characteristics; for time series data, the autoregressive moving average model is used for interpolation, and for non-time series data, cubic spline interpolation is used.

[0025] Optionally, the analysis of electricity consumption characteristics of the clean energy industry, the electric vehicle industry, and the energy storage industry specifically includes:

[0026] Extract periodicity, trend and volatility characteristics from historical electricity consumption data.

[0027] Optionally, the prediction of the electricity consumption contribution rate and electricity demand of the clean energy industry specifically includes:

[0028] The MLP model is used to predict the contribution rate of electricity consumption in the clean energy industry. The input characteristics of the MLP model include electricity consumption characteristics and influencing factor characteristics. Based on the historical electricity consumption data of the clean energy industry, the SARIMA model is used to predict the electricity demand of the clean energy industry in the future period.

[0029] Optionally, the electricity consumption contribution rate forecast and electricity demand forecast for the electric vehicle industry specifically include:

[0030] The contribution rate of electricity consumption of the electric vehicle industry is calculated through the spatial regression model; the electricity demand of the electric vehicle industry in future periods is predicted based on the Prophet algorithm.

[0031] Optionally, the prediction of power consumption contribution rate and power demand for the energy storage industry specifically includes:

[0032] The electricity consumption contribution rate of the energy storage industry is calculated through reinforcement learning; the electricity demand of the energy storage industry in the future period is predicted based on the LSTM algorithm.

[0033] A second aspect of the present invention provides a power consumption prediction system based on multidimensional data and industry characteristics. Based on the power consumption prediction method based on multidimensional data and industry characteristics described in the first aspect of the present invention, the system comprises:

[0034] Data collection module, data preprocessing module, power consumption characteristics analysis module, power consumption contribution rate and power demand forecasting module and overall power consumption forecasting module;

[0035] The data collection module is used to collect historical electricity consumption data and multi-dimensional influencing factor data of green and low-carbon industries in a certain region, and the green and low-carbon industries include clean energy industry, electric vehicle industry and energy storage industry;

[0036] The data preprocessing module is used to preprocess the collected multi-dimensional influencing factor data and build an influencing factor index library and an industry electricity consumption database;

[0037] The power consumption characteristics analysis module is used to analyze the power consumption characteristics of the clean energy industry, electric vehicle industry and energy storage industry, and obtain the power consumption characteristics of each industry;

[0038] The electricity consumption contribution rate and electricity demand forecasting module is used to forecast the electricity consumption contribution rate and electricity demand of green and low-carbon industries for each industry based on the electricity consumption characteristics and historical electricity consumption data of each industry;

[0039] The overall electricity consumption forecast module is used to perform weighted summation on the electricity consumption contribution rate forecast and electricity demand forecast of each industry, and calculate the electricity consumption forecast value of the overall green and low-carbon industry.

[0040] The third aspect of the present invention provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is loaded into the processor, the method for predicting electricity consumption based on multidimensional data and industry characteristics according to the first aspect of the present invention is implemented.

[0041] A fourth aspect of the present invention provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements a method for predicting electricity consumption based on multidimensional data and industry characteristics according to the first aspect of the present invention.

[0042] Compared with the prior art, the beneficial effects of the present invention include at least: by integrating multi-dimensional data, including economic, meteorological, policy, industry characteristics, etc., and deeply exploring the specific electricity consumption patterns of green and low-carbon industries, such as the seasonality of clean energy, the spatial distribution of electric vehicles, and the charging and discharging patterns of the energy storage industry, the accuracy and pertinence of electricity demand forecasting have been greatly improved; by introducing quantitative analysis of contribution rates, the impact of each green industry on the overall electricity demand is clarified, providing a scientific basis for electricity optimization and resource allocation; customized forecasting methods, including SARIMA, Prophet, LSTM, etc., are used to adapt to the needs of different industries, and ultra-short-term, short-term and medium- to long-term multi-level electricity demand forecasts are achieved; at the same time, energy conservation and emission reduction and efficient energy utilization goals are supported, helping the government to formulate scientific policies, enterprises to optimize production plans and industries to achieve sustainable development. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a schematic flow chart of the method of the present invention. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical scheme and advantages of the present invention clearer, the technical scheme of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. The described embodiments are only embodiments of a part of the present invention, not all embodiments. Based on the spirit of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work belong to the protection scope of the present invention. With the increasing attention paid to environmental protection and sustainable development around the world, green and low-carbon industries have gradually become an important direction for economic development. However, the power consumption forecasting of these industries is crucial for energy management and optimization. The existing power consumption forecasting methods often have the following problems: First, only a single factor or a limited number of factors are considered, which fails to fully reflect the complexity and diversity of power consumption in green and low-carbon industries; second, no separate industry analysis and power consumption characteristics analysis are conducted on each industry under the green and low-carbon industry, resulting in inaccurate prediction results; third, various types of data involving electricity, economy, meteorology, policy, etc. are missing and abnormal values ​​are improperly handled, which further affects the prediction accuracy.

[0045] Therefore, a method for predicting electricity consumption in green and low-carbon industries based on multi-dimensional data fusion and industry characteristics analysis is constructed. This method collects and analyzes multi-dimensional data in economics, meteorology, policies, industry characteristics, etc., and uses advanced data fusion and model building technology to achieve high-precision prediction of electricity consumption in various industries under the green and low-carbon industry, thereby facilitating the prediction of electricity consumption in the entire green and low-carbon industry.

[0046] The present invention aims to construct a method and system for predicting electricity consumption in green and low-carbon industries based on multi-dimensional data fusion and industry characteristics analysis. By integrating economic, meteorological, policy, industry characteristics and other data, advanced data fusion technology is used to explore the correlation and complementarity between data, and in-depth analysis of the electricity consumption characteristics and patterns of various green and low-carbon industries is conducted to construct a high-precision prediction model, so as to achieve accurate prediction of electricity consumption in various industries under the green and low-carbon industry, thereby facilitating efficient energy utilization, energy conservation and emission reduction, and sustainable development of the industry, and providing a scientific basis for government decision-making, enterprise production plan optimization, etc.

[0047] To achieve the above objectives, the technical method provided by the present invention is: a green and low-carbon industry electricity consumption prediction and analysis model based on multi-dimensional data fusion and industry characteristics analysis, so as to achieve a more accurate prediction of green and low-carbon industry electricity consumption.

[0048] Embodiment 1 of the present invention provides a method for predicting power consumption based on multi-dimensional data and industry characteristics, comprising the following steps:

[0049] Step S1: Since the green industries included in the green and low-carbon industry are too complicated, in order to simplify the research route, the present invention takes into account the easy accessibility of electricity consumption data and has practical operability. It is assumed that the green and low-carbon industry in a certain area only includes the clean energy industry, the electric vehicle industry, and the energy storage industry, among which the clean energy industry includes photovoltaic power generation and wind power generation industries.

[0050] Correspondingly, historical electricity consumption data and influencing factors data of the specific area are collected, including economic factors, meteorological factors, and policy factors. Each type of factor also contains multiple influencing factors.

[0051] In a further preferred but non-limiting embodiment, the impact factor data collected in step S1 specifically includes:

[0052]

[0053]

[0054] In a further preferred but non-limiting embodiment, in step S1, the clean energy industry includes photovoltaic power generation and wind power generation industries.

[0055] Step S2: pre-process each item of data of various influencing factors collected in S1 and construct an influencing factor index library and an industry electricity consumption database, wherein the pre-processing includes detecting abnormal values ​​and filling missing values.

[0056] In a further preferred but non-limiting embodiment, in step S2, preprocessing the collected various influencing factors and each data specifically includes:

[0057] Outlier detection uses the Z-score method to determine whether the data is abnormal through statistical standardization. Suppose the original data set is X = {X1, X2, ..., X n}, where n is the total number of indicators, where n i Represents X i The amount of data for the indicator. The mean μ and standard deviation σ of each indicator are:

[0058]

[0059] For each data point x i , calculate its Z-score value:

[0060]

[0061] When | Z i |>3, it is considered that x i is an outlier;

[0062] For the detected outliers, appropriate interpolation methods are used according to the data characteristics.

[0063] Specifically, for time series data, the autoregressive moving average (ARMA) model is used for interpolation, and for non-time series data, the cubic spline interpolation method is used;

[0064] Specifically, the ARMA model consists of two parts: autoregression (AR) and moving average (MA). The time series is represented by {x i}, satisfying the following model:

[0065] x t =φ1x t-1 +φ2x t-2 +...+φ p x t-p +θ1∈ t-1 +θ2∈ t-2 +...+θ q ∈ t-q +∈ t

[0066] Among them, φ1, φ2, ..., φ p are the parameters of the autoregressive (AR) part, θ1,θ2,...,θ q is the parameter of the moving average (MA) part, ∈ t It's white noise.

[0067] The cubic spline interpolation method constructs a smooth cubic polynomial curve through a series of known discrete data points, and uses this curve to estimate the value of the unknown point. Assume that the data points are {(x1,y1),(x2,y2),...,(x n ,y n )}, at the adjacent data point [x i ,x i+1 ] define a cubic polynomial:

[0068] S i (x) = a i +b i (xx i )+c i (xx i ) 2 +d i (xx i ) 3 ,x∈[x i ,x i+1 ]

[0069] The polynomial coefficient a is determined by the following conditions i , b i , c i and di .

[0070] Data point interpolation conditions:

[0071] S i (x i )=y i ,S i (x i+1 )=y i+1 ;

[0072] First-order derivative continuity:

[0073] S′ i (x i+1 ) = S′ i+1 (x i+1 );

[0074] Second-order derivative continuity:

[0075] S″ i (x i+1 )=S″ i+1 (x i+1 ).

[0076] By solving the tridiagonal matrix, the interpolation function S(x) is obtained, which is finally used to fill in the outliers.

[0077] Step S3: Based on the robust influencing factor index library and industry electricity consumption database constructed in step S2, industry characteristics analysis and typical electricity consumption curve extraction are performed for each industry, and electricity consumption characteristics analysis is performed by extracting periodicity, trend and volatility characteristics from historical electricity consumption data.

[0078] In a further preferred but non-limiting embodiment, in step S3, extracting the periodicity, trend and volatility characteristics in the historical electricity consumption data specifically includes:

[0079] Fourier transform is used to analyze the periodic components of the electricity consumption time series and extract the main periodic characteristics. The formula is:

[0080]

[0081] Where F(x) represents the amplitude of the kth frequency component in the frequency domain, Y t represents the tth data point in the time series, t represents the index in the time series, and N represents the total number of data points in the time series.

[0082] Use the moving average method to extract long-term trend T t , the formula is:

[0083]

[0084] Where m is the window size, T t+i Represents the data point at time t+i, where i ranges from -m to m.

[0085] The volatility characteristic is measured by the coefficient of variation CV, which is used to describe the degree of fluctuation of electricity consumption data:

[0086]

[0087] In the formula, σ Y Represents the standard deviation of the data set Y, μ Y Represents the mean of the data set Y.

[0088] Step S4: Based on the electricity consumption characteristics obtained by analyzing step S3, the electricity consumption contribution rate and electricity demand of the clean energy industry, electric vehicle industry and energy storage industry are predicted respectively;

[0089] In a further preferred but non-limiting embodiment, in step S4, predicting the power consumption contribution rate and power demand of the clean energy industry specifically includes:

[0090] In predicting the contribution rate of the clean energy industry to the electricity consumption of the green and low-carbon industries, it is necessary to combine the electricity consumption characteristics with economic, meteorological, policy and other influencing factors, use the MLP model to predict the contribution rate, and finally predict the industry's electricity demand based on the SARIMA model.

[0091] First, based on the regression model combined with normalization, the contribution rate of the clean energy industry is predicted. 1,t The model inputs two types of features: electricity consumption characteristics and influencing factors. Influencing factors include economic factors (such as GDP, industrial growth rate), meteorological factors (such as wind speed, temperature, light radiation intensity) and policy factors (such as subsidy policy score). These features are defined as vector F 1,t ={f 1,t ,f 2, t,...,f n,t}, and input the model together with the electricity consumption characteristics, and indirectly predict the contribution rate CR of the clean energy industry by fitting the dynamic relationship between historical electricity consumption data and green and low-carbon industries. 1,t The regression model is defined as

[0092]

[0093] Where w0 is the bias term, w i is the regression coefficient of the influencing factor. The MLP model is used to train the model parameters through the historical electricity consumption data and the characteristics of the influencing factors, so that the model can fit the relative contribution trend of the industry.

[0094] After the contribution rate forecast is completed, the SARIMA model is used to predict the electricity demand of the clean energy industry. First, the historical electricity consumption time series {Y 1,t} to make it stable. t ′=Y t -Y t-s Remove the trend by seasonally differencing Y t ″=Y t ′-Y′ t-s Remove periodic fluctuations. Stationary series {Y t ″} is then used to construct the SARIMA model, which is mathematically expressed as:

[0095] φ p (B)Φ P (B s )Y t ″=θ p (B)Θ Q (B s )∈ t

[0096] Where B is the lag operator, φ p (B), θ p (B) is a non-seasonal autoregressive and moving average polynomial, Φ P (B s ), Θ Q (B s ) is a seasonal autoregressive and moving average polynomial. The model parameters p, d, q, P, D, Q, s are determined by the maximum likelihood estimation method.

[0097] After the training is completed, the SARIMA model is used to predict the electricity demand of the clean energy industry in the future period. The formula is:

[0098] Y 1,t+h =SARIMA({Y 1,t})

[0099] In a further preferred but non-limiting embodiment, in step S4, predicting the power consumption contribution rate and power demand of the electric vehicle industry specifically includes:

[0100] The contribution rate and electricity demand forecast of the electric vehicle industry requires an analysis of influencing factors and electricity consumption characteristics, which can be completed in two steps:

[0101] First, the contribution rate of the electric vehicle industry to the electricity consumption of the green and low-carbon industry is estimated through the spatial regression model; then, the industry electricity consumption is predicted in time series based on the Prophet algorithm. This method combines the spatial correlation of influencing factors and the long-term trend of time series, providing a scientific basis for accurately predicting industry electricity demand.

[0102] In the contribution rate forecast, it is considered that the electricity consumption of the electric vehicle industry is affected by multiple factors such as economy, meteorology and policy. At the same time, these factors often have significant spatial correlations, such as the promotion effect of the new energy vehicle industry policy in a certain region on the adjacent regions. Therefore, the contribution rate forecast uses a spatial regression model (SpatialRegression). Assuming that the contribution rate of the electric vehicle industry electricity consumption in the total industry electricity consumption is CR 2,i , where i represents a spatial unit (such as a city or region), and its influencing factors include its own characteristics F 2,i ={f 1,i ,f 2,i ,...,f n,i The transmission effect of the characteristics of} and adjacent units through the spatial weight matrix W. The basic form of the spatial lag model (Spatial Lag Model, SLM) is:

[0103]

[0104] Among them, ρ represents the spatial hysteresis coefficient, which reflects the contribution rate of adjacent units to CR 2,i The impact of 2,i β is the linear combination of influencing factors, β is the regression coefficient; ∈ i is the random error term; W ij is an element of the spatial weight matrix, which represents the spatial proximity relationship (e.g., distance or boundary contact) between regions i and j. The model parameters ρ and β are estimated by the maximum likelihood estimation (MLE) method, and the contribution rate CR of the electric vehicle industry can be predicted. 2,i .

[0105] After completing the contribution rate forecast, the Prophet algorithm is used to perform time series forecasting on the electricity demand of the electric vehicle industry. The Prophet algorithm is an additive model that can decompose the time series into components such as trend, seasonality, and holiday effect. It is suitable for electricity consumption data with obvious trends and seasonality. Assume that the historical electricity consumption time series of the electric vehicle industry is {Y 2,t}, the Prophet algorithm expresses it as:

[0106] Y 2,t=g(t)+s(t)+h(t)+∈ t

[0107] Among them, g(t) is a non-periodic trend function, usually expressed as a piecewise linear or logarithmic growth model; s(t) is a periodic seasonal function, modeled using Fourier series; h(t) is a holiday effect function, capturing the impact of a specific time point (such as the policy release day) on electricity consumption; ∈ t is a random error term. Specifically, the trend function g(t) is in the form of:

[0108]

[0109] Where k is the initial level, δ is the growth rate, and II(t>t i ) is the indicator function of the change point, allowing the trend to be j changes have occurred.

[0110] The Prophet algorithm estimates the model parameters k, δ, and a by fitting historical data. j Etc., and generate the predicted value Y of future electricity consumption 2,t+h In model training, periodic features (such as seasonal electricity consumption fluctuations) are modeled through Fourier series expansion as:

[0111]

[0112] Where T is the cycle length, a n and b n are the Fourier coefficients.

[0113] In a further preferred but non-limiting embodiment, in step S4, predicting the power consumption contribution rate and power demand of the energy storage industry specifically includes:

[0114] The contribution rate of the energy storage industry in the green and low-carbon industry is estimated through the reinforcement learning method, and then the industry's electricity consumption is predicted in time series based on the LSTM algorithm. This method constructs a dynamic decision-making process through reinforcement learning, captures the complex interactive relationship between the contribution rate of the energy storage industry's electricity consumption and multi-dimensional influencing factors, and combines the strong time dependence of the LSTM model to capture the law of electricity consumption changes over time, and finally achieves accurate prediction of the electricity demand of the energy storage industry.

[0115] In the contribution rate prediction, reinforcement learning models the problem as a dynamic decision-making process, and optimizes the estimation of the energy storage industry's electricity consumption contribution rate through the agent learning in the environment of influencing factor data. Assuming that the energy storage industry's electricity consumption contribution rate is CR t , the state faced by the agent at each time step t is determined by the influencing factor feature vector F3,t ={f 1,t ,f 2,t ,...,f n,t} indicates that the state space Including economic, meteorological and policy variables. The agent takes action a t Update contribution rate estimate, action space represents the adjustment of contribution rate. The strategy π(a t ,F 3,t ) is a mapping from states to actions, determining the action to take in the current state.

[0116] The goal of reinforcement learning is to maximize the cumulative reward:

[0117]

[0118] where r t is the immediate reward, reflecting the improvement effect of the action on the contribution rate estimation; γ∈[0,1] is the discount factor used to weigh short-term and long-term rewards. Reward function r t The current predicted contribution rate is measured to the actual energy storage industry electricity consumption Y t The fitting effect of:

[0119] r t =-(Y t -g(F t ,CR 3,t )) 2

[0120] Where g(F 3,t ,CR 3,t ) function is used to predict the relative change trend of industry electricity consumption through influencing factors and contribution rates.

[0121] Reinforcement learning uses the Deep Q Network (DQN) algorithm, which uses a deep neural network to approximate the Q function:

[0122]

[0123] Among them, the agent is based on the state S at each time step t Choose the best action a t , and iteratively adjust the predicted contribution rate CR 3,t .

[0124] After the contribution rate forecast is completed, LSTM is used to perform time series forecasting on the electricity demand of the energy storage industry. LSTM is a recurrent neural network that can capture the long-term dependence and complex dynamic changes of electricity consumption in the energy storage industry.

[0125] The historical electricity consumption time series of the energy storage industry is {Y 3,t}, the input features include the periodicity, trend and volatility characteristics of electricity consumption, as well as the contribution rate CR of reinforcement learning prediction 3,t At each time step t, the input feature combination is X t The structure of LSTM is defined by the following formula:

[0126] f t =σ(W f [h t-1 ,X t ]+b j )

[0127] i t =σ(W i [h t-1 ,X t ]+b i )

[0128] o t =σ(W o [h t-1 ,X t ]+b o )

[0129]

[0130] h t =o t ⊙ t anh(C t )

[0131] where h t is the hidden state at time step t, C t is the cell state, f t 、i t and t are the activation values ​​of the forget gate, input gate, and output gate respectively. The LSTM model learns the dynamic change law of electricity consumption in the energy storage industry through these gating mechanisms and generates the predicted value Y of electricity demand at the future time step t+h 3,t+h .

[0132] Step S5: performing weighted summation of the power consumption and contribution rate of each industry obtained in step S4 to calculate the power consumption forecast value of the overall green and low-carbon industry;

[0133] In a further preferred but non-limiting embodiment, step S5 specifically includes:

[0134] The electricity consumption forecast of the green and low-carbon industry can be calculated by weighted summation of the electricity consumption forecast of each industry and the industry's comprehensive contribution rate. This method is directly based on the electricity consumption forecast results of the clean energy, electric vehicles, and energy storage industries, and their industry contribution rates are used as weights for linear combination to obtain the overall green and low-carbon industry electricity consumption forecast.

[0135] The predicted electricity consumption values ​​of the three industries are Y 1,t ,Y 2,t ,Y 3,t The corresponding industry contribution rate forecast values ​​are CR 1,t ,CR 2,t ,CR 3,t . Forecast value of total electricity consumption of green and low-carbon industries T t The calculation formula is:

[0136]

[0137] in, is the normalized contribution rate of the ith industry, which is used to ensure that the weighted sum of the contribution rates of all industries is 1. The normalization process is as follows:

[0138]

[0139] The purpose of normalization is to make the weight ratio between industries reasonable and within the physical meaning when there is a certain deviation in the contribution rate. The weighted summation method combines the electricity demand forecast results of the three industries with the contribution rate information to obtain the overall electricity consumption of the green and low-carbon industry, while retaining the relative contribution size of each industry.

[0140] Embodiment 2 of the present invention provides a power consumption prediction system based on multidimensional data and industry characteristics. Based on the power consumption prediction method based on multidimensional data and industry characteristics described in Embodiment 1 of the present invention, the system includes:

[0141] Data collection module, data preprocessing module, power consumption characteristics analysis module, power consumption contribution rate and power demand forecasting module and overall power consumption forecasting module;

[0142] The data collection module is used to collect historical electricity consumption data and multi-dimensional influencing factor data of green and low-carbon industries in a certain region, and the green and low-carbon industries include clean energy industry, electric vehicle industry and energy storage industry;

[0143] The data preprocessing module is used to preprocess the collected multi-dimensional influencing factor data and build an influencing factor index library and an industry electricity consumption database;

[0144] The power consumption characteristics analysis module is used to analyze the power consumption characteristics of the clean energy industry, electric vehicle industry and energy storage industry, and obtain the power consumption characteristics of each industry;

[0145] The electricity consumption contribution rate and electricity demand forecasting module is used to forecast the electricity consumption contribution rate and electricity demand of green and low-carbon industries for each industry based on the electricity consumption characteristics and historical electricity consumption data of each industry;

[0146] The overall electricity consumption forecast module is used to perform weighted summation on the electricity consumption contribution rate forecast and electricity demand forecast of each industry, and calculate the electricity consumption forecast value of the overall green and low-carbon industry.

[0147] Embodiment 3 of the present invention provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is loaded into the processor, the method for predicting electricity consumption based on multidimensional data and industry characteristics according to Embodiment 1 is implemented.

[0148] Embodiment 4 of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for predicting electricity consumption based on multidimensional data and industry characteristics according to Embodiment 1 is implemented.

[0149] The present invention not only relates to the field of data processing and analysis, but is also widely used in multiple fields such as energy management, smart grid, environmental science, and sustainable development. In particular, in the context of the rapid development of green and low-carbon industries, the present invention provides key technical support for energy planning and scheduling in the clean energy industry, electric vehicle industry, and energy storage industry. At the same time, by improving the accuracy and reliability of electricity consumption forecasts in green and low-carbon industries, it helps promote the healthy development of green and low-carbon industries and promote sustainable economic growth. By optimizing energy utilization and reducing environmental pollution, the present invention also makes a positive contribution to achieving sustainable development goals.

[0150] The present invention proposes a method based on multi-dimensional data fusion technology and green and low-carbon industry characteristic analysis, which not only combines multiple factors such as economy, meteorology, and policy, but also deeply analyzes the specific electricity consumption patterns of clean energy, electric vehicles, energy storage and other industries, and adopts differentiated prediction models and contribution rate quantitative analysis according to the characteristics of different industries, realizing comprehensive prediction and optimization from ultra-short-term to medium- and long-term. Based on this significant difference, the technical effects actually achieved by the present invention include at least: significantly improving prediction accuracy and industry adaptability, supporting energy conservation, emission reduction and energy efficient utilization goals, and providing a scientific basis for government policy formulation and enterprise production plan optimization, and providing all-round support for the sustainable development of green and low-carbon industries.

[0151] The present disclosure may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0152] 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 it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for predicting electricity consumption based on multidimensional data and industry characteristics, characterized in that: The steps include: Collect historical electricity consumption data of a region's green and low-carbon industries and their multi-dimensional influencing factor data. Green and low-carbon industries include clean energy, electric vehicles, and energy storage. Multi-dimensional influencing factor data include economic factors, meteorological factors, and policy factors. Each factor also contains multiple influencing factors. Pre-process the collected multi-dimensional influencing factor data to build an influencing factor index library and industry electricity consumption database; Based on the influencing factor index library and industry electricity consumption database, the electricity consumption characteristics of the clean energy industry, electric vehicle industry and energy storage industry are analyzed to obtain the electricity consumption characteristics of each industry; According to the electricity consumption characteristics and historical electricity consumption data of each industry, the green and low-carbon industry electricity consumption contribution rate and electricity demand forecast are carried out for each industry; The electricity consumption contribution rate forecast and electricity demand forecast of each industry are weighted and summed to calculate the electricity consumption forecast of the overall green and low-carbon industry.

2. The method for predicting power consumption based on multidimensional data and industry characteristics according to claim 1, characterized in that: Each factor also includes multiple influencing factors, specifically: Economic factors include return on investment, cost-benefit ratio, construction cost, operating cost, driving range, energy consumption per 100 kilometers, battery cost, charging facility construction cost, energy storage cost and energy storage efficiency; Meteorological factors include temperature, humidity, precipitation, wind speed, and sunshine duration; Policy factors include electricity price policies, policy subsidies, industry development policy documents, industry planning white papers and industry statistical yearbooks.

3. The method for predicting power consumption based on multidimensional data and industry characteristics according to claim 1, characterized in that: The preprocessing of the collected multi-dimensional influencing factor data includes detecting outliers and filling missing values, specifically: Use Z-score to detect outliers. Suppose the original data set is X={X1,X2,...,X n }, where n is the total number of indicators, where n i Represents X i The data volume of the indicator, the mean μ and standard deviation σ of each indicator are: For each data point x i , calculate its Z-score value: When | Z i |>3, it is considered that x i is an outlier; For the detected outliers, appropriate interpolation methods are used according to the data characteristics; for time series data, the autoregressive moving average model is used for interpolation, and for non-time series data, cubic spline interpolation is used.

4. The method for predicting power consumption based on multidimensional data and industry characteristics according to claim 1, characterized in that: The analysis of electricity consumption characteristics of the clean energy industry, electric vehicle industry and energy storage industry specifically includes: Extract periodicity, trend and volatility characteristics from historical electricity consumption data.

5. The method for predicting power consumption based on multidimensional data and industry characteristics according to claim 1, characterized in that: The prediction of electricity consumption contribution rate and electricity demand of the clean energy industry specifically includes: The MLP model is used to predict the contribution rate of electricity consumption in the clean energy industry. The input characteristics of the MLP model include electricity consumption characteristics and influencing factor characteristics. Based on the historical electricity consumption data of the clean energy industry, the SARIMA model is used to predict the electricity demand of the clean energy industry in the future period.

6. The method for predicting power consumption based on multidimensional data and industry characteristics according to claim 1, characterized in that: The electricity consumption contribution rate forecast and electricity demand forecast for the electric vehicle industry specifically include: The contribution rate of electricity consumption of the electric vehicle industry is calculated through the spatial regression model; the electricity demand of the electric vehicle industry in future periods is predicted based on the Prophet algorithm.

7. The method for predicting power consumption based on multidimensional data and industry characteristics according to claim 1, characterized in that: The electricity consumption contribution rate forecast and electricity demand forecast for the energy storage industry specifically include: The electricity consumption contribution rate of the energy storage industry is calculated through reinforcement learning; the electricity demand of the energy storage industry in the future period is predicted based on the LSTM algorithm.

8. A power consumption prediction system based on multidimensional data and industry characteristics, based on a power consumption prediction method based on multidimensional data and industry characteristics as claimed in any one of claims 1 to 7, characterized in that: The system includes: Data collection module, data preprocessing module, power consumption characteristics analysis module, power consumption contribution rate and power demand forecasting module and overall power consumption forecasting module; The data collection module is used to collect historical electricity consumption data and multi-dimensional influencing factor data of green and low-carbon industries in a certain region, and the green and low-carbon industries include clean energy industry, electric vehicle industry and energy storage industry; The data preprocessing module is used to preprocess the collected multi-dimensional influencing factor data and build an influencing factor index library and an industry electricity consumption database; The power consumption characteristics analysis module is used to analyze the power consumption characteristics of the clean energy industry, electric vehicle industry and energy storage industry, and obtain the power consumption characteristics of each industry; The electricity consumption contribution rate and electricity demand forecasting module is used to forecast the electricity consumption contribution rate and electricity demand of green and low-carbon industries in various industries based on the electricity consumption characteristics and historical electricity consumption data of various industries; The overall electricity consumption forecast module is used to perform weighted summation on the electricity consumption contribution rate forecast and electricity demand forecast of each industry, and calculate the electricity consumption forecast value of the overall green and low-carbon industry.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the computer program is loaded into the processor, it implements a method for predicting electricity consumption based on multi-dimensional data and industry characteristics according to any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, a method for predicting electricity consumption based on multidimensional data and industry characteristics according to any one of claims 1 to 7 is implemented.

Images