Double decomposition method and device for short-term daily power consumption of user and storage medium

Abstract

The invention relates to a user short-term daily power consumption double decomposition method and device and a storage medium, which are applied to the technical field of power system load prediction, and comprise the following steps: extracting key feature data from historical data through a feature identification technology, and obtaining a trend component and a periodic component by using the feature data and adopting a regression periodic decomposition method, and obtaining random components by adopting empirical mode decomposition, predicting each component by adopting different prediction models, and finally, effectively fusing prediction results of each component by combining a deep learning technology and an integrated learning technology so as to improve prediction precision and robustness. According to the method provided by the invention, the key factors influencing the short-term daily electric quantity of the industrial user are accurately identified, and the subtle characteristics and rules of the electric quantity change are captured, so that a more accurate prediction model is constructed, the requirements of a power grid marketing department on high precision and timeliness of electric quantity prediction are met, and powerful support is provided for operation and decision-making of an electric power market.

Description

Technical Field

[0001] The present invention relates to the technical field of power system load forecasting, and in particular to a method, a device and a storage medium for double decomposing a user's short-term daily electricity consumption. Background Art

[0002] In recent years, with the continuous development of the power market and the improvement of power supply service level, the demand for accurate prediction of short-term electricity consumption of large users has been increasing. The power market has the demand for multi-day electricity transactions, and short-term electricity forecast can provide support for this. In addition, with the development of enterprises themselves and the continuous adjustment of industrial structure, the energy consumption characteristics of large power users are becoming increasingly different. Therefore, targeted and differentiated treatment is needed in the prediction.

[0003] For power grid marketing departments, they are currently facing challenges in electricity forecasting, such as inconsistent data quality, limited prediction model accuracy, and diversified prediction needs. Although various methods such as time series models and machine learning models have been used for forecasting, the prediction accuracy is low when dealing with complex and changeable electricity demand patterns, which does not meet the high accuracy and timeliness requirements of power grid marketing departments for electricity forecasting. Summary of the invention

[0004] In view of this, the purpose of the present invention is to provide a method, device and storage medium for double decomposition of user's short-term daily electricity consumption, so as to solve the problem of low prediction accuracy of the time series model and machine learning model currently used in the prior art when dealing with complex and changeable electricity demand patterns.

[0005] According to a first aspect of an embodiment of the present invention, a method for double decomposition of short-term daily electricity consumption of a user is provided, the method comprising:

[0006] Obtain the user's daily electricity consumption data within a preset time period;

[0007] Preprocess the acquired daily electricity consumption data within a preset historical time period to obtain standard daily electricity consumption data;

[0008] Extracting characteristic data from the standard daily electricity consumption data;

[0009] According to the extracted characteristic data, the regression periodic decomposition method is used to obtain the trend component and periodic component;

[0010] Obtaining an initial random component according to the trend component and the period component;

[0011] Adding different random noises to the initial random component to generate multiple perturbation versions;

[0012] Each disturbance version generated is subjected to empirical mode decomposition, and each empirical mode decomposition obtains high-frequency data, medium-frequency data, and low-frequency data. The high-frequency data obtained by multiple decompositions are accumulated to obtain the final random component;

[0013] The pre-built polynomial regression model framework is iteratively trained through the trend component to obtain a trend component prediction model; the pre-built autoregressive moving average model framework is iteratively trained through the periodic component to obtain a periodic component prediction model; the pre-built random forest model framework is iteratively trained through the final random component to obtain a random component prediction model;

[0014] Output the trend component prediction value, the period component prediction value and the random component prediction value of the user to be predicted respectively through the trend component prediction model, the period component prediction model and the random component prediction model;

[0015] The trend component prediction value, the period component prediction value and the random component prediction value are integrated to obtain the daily power consumption prediction value of the user to be predicted.

[0016] Preferably,

[0017] The preprocessing of the acquired daily electricity consumption data within the preset historical time period to obtain the standard daily electricity consumption data includes:

[0018] Using regression interpolation method to fill missing values ​​in the daily electricity consumption data;

[0019] The filled daily electricity consumption data is standardized using the Min-Max standardization method to obtain standard daily electricity consumption data.

[0020] Preferably,

[0021] The extracting of characteristic data from the standard daily power consumption data comprises:

[0022] Obtaining the maximum load value, the minimum load value and the average load value in the standard daily electricity consumption data;

[0023] Obtaining a peak-to-valley difference according to the maximum load value and the minimum load value;

[0024] Obtaining a peak-to-valley difference rate according to the peak-to-valley difference and the maximum load value;

[0025] The average daily power consumption on working days and the average daily power consumption on rest days are obtained according to the standard daily power consumption data; and the power consumption ratio of working days to rest days is obtained according to the average daily power consumption on working days and the average daily power consumption on rest days.

[0026] Preferably,

[0027] The method of obtaining the trend component by using the regression period decomposition method according to the extracted characteristic data includes:

[0028] Setting regression parameters of regression period decomposition, and obtaining initial trend components according to the regression parameters, maximum load value, minimum load value and average load value;

[0029] The initial trend component is smoothed by using a moving average method to eliminate short-term noise in the initial trend component to obtain the trend component.

[0030] Preferably,

[0031] The method of obtaining the periodic component by using the regression periodic decomposition method according to the extracted characteristic data comprises:

[0032] The regression coefficient of the dummy variable is set, and the periodic dummy variable is obtained according to the regression coefficient of the dummy variable, the peak-to-valley difference rate, and the working day-rest day power consumption ratio;

[0033] Obtaining a designated component according to the standard daily electricity consumption data, the trend component and the periodic dummy variable;

[0034] Decomposing the designated component by using the empirical mode decomposition method to extract the intermediate frequency IMF component therein, wherein the intermediate frequency IMF component is a periodic residual variable;

[0035] The periodic residual variable and the periodic dummy variable are combined to obtain the periodic component.

[0036] Preferably,

[0037] The step of fusing the trend component prediction value, the period component prediction value and the random component prediction value to obtain the daily power consumption prediction value of the user to be predicted includes:

[0038] The trend component weight, the period component weight and the random component weight are set respectively, and the trend component weight, the period component weight and the random component weight are used to combine the trend component prediction value, the period component prediction value and the random component prediction value to obtain the daily electricity consumption prediction value of the user to be predicted.

[0039] According to a second aspect of an embodiment of the present invention, a device for dual decomposition of short-term daily electricity consumption of a user is provided, the device comprising:

[0040] Data acquisition module: used to obtain the user's daily electricity consumption data within a preset historical time period;

[0041] Preprocessing module: used to preprocess the daily electricity consumption data acquired within a preset historical time period to obtain standard daily electricity consumption data;

[0042] Feature extraction module: used to extract feature data from the standard daily power consumption data;

[0043] Regression period decomposition module: used to obtain trend component and period component by regression period decomposition method according to the extracted characteristic data;

[0044] An initial random component acquisition module is used to acquire an initial random component according to the trend component and the period component;

[0045] Noise disturbance module: used for adding different random noises to the initial random component to generate multiple disturbance versions;

[0046] Empirical mode decomposition module: used to perform empirical mode decomposition on each disturbance version generated. Each empirical mode decomposition obtains high-frequency data, medium-frequency data and low-frequency data. The high-frequency data obtained by multiple decompositions are accumulated to obtain the final random component.

[0047] Model training module: used to iteratively train the pre-built polynomial regression model framework through the trend component to obtain a trend component prediction model; iteratively train the pre-built autoregressive moving average model framework through the periodic component to obtain a periodic component prediction model; iteratively train the pre-built random forest model framework through the final random component to obtain a random component prediction model;

[0048] Prediction module: used to output the trend component prediction value, the period component prediction value and the random component prediction value of the user to be predicted respectively through the trend component prediction model, the period component prediction model and the random component prediction model;

[0049] Prediction result fusion module: used to fuse the trend component prediction value, the period component prediction value and the random component prediction value to obtain the daily power consumption prediction value of the user to be predicted.

[0050] According to a third aspect of an embodiment of the present invention, there is provided a storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by a main controller, each step in the above method is implemented.

[0051] The technical solution provided by the embodiments of the present invention may have the following beneficial effects:

[0052] This application extracts key feature data from historical data through feature identification technology, uses the feature data to adopt the regression periodic decomposition method to obtain the trend component and the periodic component, and adopts the empirical mode decomposition to obtain the random component, and adopts different prediction models for each component for prediction. Finally, it combines deep learning technology and integrated learning technology to effectively integrate the prediction results of each component to improve the prediction accuracy and robustness. The method of this application accurately identifies the key factors affecting the short-term daily electricity consumption of industrial users and captures the subtle characteristics and laws of electricity consumption changes, thereby constructing a more accurate prediction model, reducing the workload of manual intervention and data processing, improving prediction efficiency, meeting the high-precision and timeliness requirements of power grid marketing departments for electricity consumption forecasts, and providing strong support for the operation and decision-making of the power market.

[0053] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0055] Figure 1 It is a flow chart showing a method for double decomposition of short-term daily electricity consumption of a user according to an exemplary embodiment;

[0056] Figure 2 is a system schematic diagram of a device for double decomposing short-term daily electricity consumption of a user according to another exemplary embodiment;

[0057] In the attached figure: 1-data acquisition module, 2-preprocessing module, 3-feature extraction module, 4-regression period decomposition module, 5-initial random component acquisition module, 6-noise disturbance module, 7-empirical mode decomposition module, 8-model training module, 9-prediction module, 10-prediction result fusion module. DETAILED DESCRIPTION

[0058] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0059] Embodiment 1

[0060] Figure 1is a flow chart showing a method for double decomposition of short-term daily electricity consumption of a user according to an exemplary embodiment. Figure 1 As shown, the method includes:

[0061] S1, obtaining the user's daily electricity consumption data within a preset historical time period;

[0062] S2, preprocessing the acquired daily electricity consumption data within a preset historical time period to obtain standard daily electricity consumption data;

[0063] S3, extracting characteristic data from the standard daily electricity consumption data;

[0064] S4, according to the extracted characteristic data, the regression periodic decomposition method is used to obtain the trend component and the periodic component;

[0065] S5, obtaining an initial random component according to the trend component and the period component;

[0066] S6, adding different random noises to the initial random component to generate multiple perturbation versions;

[0067] S7, performing empirical mode decomposition on each disturbance version generated, obtaining high-frequency data, medium-frequency data and low-frequency data each time through empirical mode decomposition, accumulating the high-frequency data obtained through multiple decompositions to obtain the final random component;

[0068] S8, iteratively training the pre-built polynomial regression model framework through the trend component to obtain a trend component prediction model; iteratively training the pre-built autoregressive moving average model framework through the periodic component to obtain a periodic component prediction model; iteratively training the pre-built random forest model framework through the final random component to obtain a random component prediction model;

[0069] S9, outputting the trend component prediction value, the period component prediction value and the random component prediction value of the user to be predicted respectively through the trend component prediction model, the period component prediction model and the random component prediction model;

[0070] S10, fusing the trend component prediction value, the period component prediction value and the random component prediction value to obtain a daily power consumption prediction value of the user to be predicted;

[0071] It can be understood that the specific contents of this embodiment include:

[0072] Data collection and preprocessing:

[0073] Collect daily electricity consumption data of electricity users within a preset time period (usually one week);

[0074] Fill missing values ​​and normalize the collected data, including:

[0075] The collected data are interpolated using the regression interpolation method. The formula is as follows:

[0076] Z=β 0 +β 1 X 1 +β 2 X 2 +···+β p X p +ε#(1)

[0077] Among them, Z is the interpolation data to be repaired, X 1 , X 2 …X p is time, β 0 , β 1 ,…β p is the regression parameter, ε is the bias parameter;

[0078] The data is standardized based on the Min-Max standardization method to ensure that data sets of different dimensions are suitable for unified analysis. The specific formula is as follows:

[0079]

[0080] Where x is the original data value, min(x) is the smallest value in the data set, max(x) is the largest value in the data set, and x′ is the standardized data, with the range of x′ being [0,1].

[0081] Identification of user daily electricity consumption characteristics:

[0082] The load characteristic index is used to quantitatively describe the fluctuation characteristics and change rules of the load, and to identify the characteristics of the user's daily power consumption, including holiday characteristics and weekly (working days, weekends) periodic characteristics, etc. Specifically:

[0083] Maximum load value: the maximum load value within a specified time range. Applicable objects: all loads, including regional loads, industrial loads, etc.

[0084] Minimum load value: the minimum load value within a specified time range. Applicable objects: all loads, including regional loads, industrial loads, etc.

[0085] Average load value:

[0086]

[0087] Where N is the number of load collection points within the specified time range, P nIt is the load value of the nth collection point. Applicable objects: all loads, including regional loads, industrial loads, etc.

[0088] Peak to valley difference:

[0089] P d =P max -P min

[0090] Where P d Refers to the peak-to-valley difference, P max Refers to the maximum load value, P min Refers to the minimum load value; it can be used to describe the fluctuation of load, and also directly reflects the peak load regulation capacity required by the power grid;

[0091] Peak-to-valley rate:

[0092]

[0093] Where P d Refers to the peak-to-valley difference, P max Refers to the maximum load value, applicable to: all loads, including regional loads, industrial loads, etc.;

[0094] Power consumption ratio between working day and rest day:

[0095]

[0096] In the formula, τ WR is the ratio of power consumption on working days to rest days, P wav is the average daily power consumption on working days, P rav The average daily electricity consumption on rest days reflects the regularity and difference of electricity consumption between working days and rest days;

[0097] Assume that the user's daily electricity consumption sequence is Y(t), which can be expressed as:

[0098] Y(t)=T(t)+S(t)+R(t)#(3)

[0099] Where Y(t) is the actual electricity consumption of the user at time t, that is, the standard daily electricity consumption data mentioned above, T(t) is the trend component, which reflects the long-term trend of electricity consumption over time, S(t) is the periodic component, which reflects the fluctuation of electricity consumption in a specific period (such as season, week), and R(t) is the random component, which represents the random fluctuation or noise in the data;

[0100] The trend component polynomial regression based on feature identification is as follows:

[0101] T 0 (t) = β 0 +β 1 t+β 2 t2 +β 3 P max +β 4 P min +β 5 P av #(4)

[0102] Among them, T 0 (t) is the initial trend component, t is time, P max is the maximum load value, P min is the minimum load value, P av is the average load value, β 0 , β 1 ,…β 5 is the regression parameter;

[0103] The extracted initial trend component is smoothed using the moving average method to further eliminate short-term noise. The formula is as follows:

[0104]

[0105] Where k is the window size, T 0 (i) is the initial trend component;

[0106] The periodic component polynomial regression based on feature identification is as follows:

[0107]

[0108] In the formula, S RCD (t) is a periodic dummy variable, D i (t) represents the explicit periodic component (such as month, week), γ i is the regression coefficient of the corresponding dummy variable, τ PV is the peak-to-valley difference rate, τ WR The ratio of power consumption on working days to rest days;

[0109] Compute the residual sequence after showing the periodic component:

[0110] R rem (t) = Y(t) - T(t) - S RCD (t)#(7)

[0111] The empirical mode decomposition (EEMD) method was used to analyze the R rem (t) decomposition, extracting the intermediate frequency IMF components, which represent complex periodic signals and cannot be represented by explicit dummy variables.

[0112]

[0113] In the formula, S EEMD (t) is the periodic residual variable;

[0114] Combining the dummy variable with the residual variable, the period component result is:

[0115] S(t)=S RCD (t)+S EEMD (t)#(9)

[0116] EEMD method for random components:

[0117] First, according to the definition of random components, the remaining sequence in the electricity data that is not explained by trend and periodic changes is used as the initial random component sequence:

[0118] R init (t) = Y(t) - T(t) - S(t) # (10)

[0119] In order to extract the true characteristics of the random components, the empirical mode decomposition (EEMD) method is used. EEMD is an adaptive decomposition method that can handle nonlinear and non-stationary data. Specifically, EEMD first init (t) Add different random noises to generate multiple perturbation versions, and perform empirical mode decomposition (EMD) on them each time to decompose the signal into several intrinsic mode function (IMF) components. These IMF components can be divided into high-frequency, medium-frequency and low-frequency parts according to their frequency characteristics. In random component extraction, the goal is to extract the high-frequency part from these IMF components, because the high-frequency component usually represents the randomness in the signal. The sum of the high-frequency IMF components can be expressed as:

[0120] R EEMD (t) = ∑ i∈高频iMF IMF i (t)#(11)

[0121] All high-frequency IMF components are accumulated to obtain the complete random component:

[0122] R(t)=R EEMD (t)#(12)

[0123] Model building and training:

[0124] For the trend component, a polynomial regression model is used for modeling, and the regression model is fitted using the training set data to predict the future trend component;

[0125] For the periodic component, the autoregressive moving average model ARMA is used for trend forecasting:

[0126]

[0127] Among them, T t is the current value, is the autoregressive coefficient, θ is the moving average coefficient, ∈ t is a white noise term, and the autocorrelation function (ACF) and partial autocorrelation function (PACF) are used to determine the order of AR(p) and MA(q) to complete trend prediction;

[0128] Based on the cycle decomposition results, seasonal ARIMA model is used for cycle forecasting;

[0129] For random components, there are large fluctuations and complex nonlinear relationships. The random forest model has strong nonlinear fitting capabilities and is suitable for processing random components.

[0130]

[0131] Where N is the number of decision trees, f i (X t ) is the prediction result of the i-th decision tree, X t Input features (such as high-frequency IMF components, holiday labels, etc.);

[0132] Forecast of user's daily electricity consumption:

[0133] Weights are added to the trend component prediction value, the period component prediction value, and the random component prediction value. The weights can be obtained by verifying the training set data:

[0134] y(t)=ω T T(t)+ω S S(t)+ω R R(t)#(15)

[0135] In the formula, ω T is the trend component weight, ω S is the periodic component weight, ω R is the random component weight, which is integrated to complete the prediction of the final daily electricity consumption.

[0136] Embodiment 2

[0137] Figure 2 : is a system schematic diagram of a device for double decomposition of short-term daily electricity consumption of a user according to another exemplary embodiment, the device comprising:

[0138] Data acquisition module 1: used to obtain the user's daily electricity consumption data within a preset historical time period;

[0139] Preprocessing module 2: used to preprocess the acquired daily electricity consumption data within a preset historical time period to obtain standard daily electricity consumption data;

[0140] Feature extraction module 3: used to extract feature data from the standard daily power consumption data;

[0141] Regression period decomposition module 4: used to obtain trend component and period component by using regression period decomposition method according to the extracted characteristic data;

[0142] Initial random component acquisition module 5: used to acquire the initial random component according to the trend component and the period component;

[0143] Noise disturbance module 6: used for adding different random noises to the initial random component to generate multiple disturbance versions;

[0144] Empirical mode decomposition module 7: used to perform empirical mode decomposition on each disturbance version generated, and each empirical mode decomposition obtains high-frequency data, medium-frequency data and low-frequency data. The high-frequency data obtained by multiple decompositions are accumulated to obtain the final random component;

[0145] Model training module 8: used to iteratively train the pre-built polynomial regression model framework through the trend component to obtain a trend component prediction model; iteratively train the pre-built autoregressive moving average model framework through the periodic component to obtain a periodic component prediction model; iteratively train the pre-built random forest model framework through the final random component to obtain a random component prediction model;

[0146] Prediction module 9: used to output the trend component prediction value, the period component prediction value and the random component prediction value of the user to be predicted respectively through the trend component prediction model, the period component prediction model and the random component prediction model;

[0147] The prediction result fusion module 10 is used to fuse the trend component prediction value, the period component prediction value and the random component prediction value to obtain the daily power consumption prediction value of the user to be predicted.

[0148] Embodiment three:

[0149] This embodiment provides a storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by a main controller, each step in the above method is implemented;

[0150] It is understandable that the storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0151] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

[0152] It should be noted that, in the description of the present invention, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of the present invention, unless otherwise specified, the meaning of "plurality" refers to at least two.

[0153] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention belong.

[0154] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0155] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0156] In addition, each functional unit in each embodiment of the present invention may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0157] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0158] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0159] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A method for double decomposition of short-term daily electricity consumption of users, characterized in that: The method comprises: Obtain the user's daily electricity consumption data within a preset time period; Preprocess the acquired daily electricity consumption data within a preset historical time period to obtain standard daily electricity consumption data; Extracting characteristic data from the standard daily electricity consumption data; According to the extracted characteristic data, the regression periodic decomposition method is used to obtain the trend component and periodic component; Obtaining an initial random component according to the trend component and the period component; Adding different random noises to the initial random component to generate multiple perturbation versions; Each disturbance version generated is subjected to empirical mode decomposition, and each empirical mode decomposition obtains high-frequency data, medium-frequency data, and low-frequency data. The high-frequency data obtained by multiple decompositions are accumulated to obtain the final random component; The pre-built polynomial regression model framework is iteratively trained through the trend component to obtain a trend component prediction model; the pre-built autoregressive moving average model framework is iteratively trained through the periodic component to obtain a periodic component prediction model; the pre-built random forest model framework is iteratively trained through the final random component to obtain a random component prediction model; Output the trend component prediction value, the period component prediction value and the random component prediction value of the user to be predicted respectively through the trend component prediction model, the period component prediction model and the random component prediction model; The trend component prediction value, the period component prediction value and the random component prediction value are integrated to obtain the daily power consumption prediction value of the user to be predicted.

2. The method according to claim 1, characterized in that The preprocessing of the acquired daily electricity consumption data within the preset historical time period to obtain the standard daily electricity consumption data includes: Using regression interpolation method to fill missing values ​​in the daily electricity consumption data; The filled daily electricity consumption data is standardized using the Min-Max standardization method to obtain standard daily electricity consumption data.

3. The method according to claim 2, characterized in that The extracting of characteristic data from the standard daily power consumption data comprises: Obtaining the maximum load value, the minimum load value and the average load value in the standard daily electricity consumption data; Obtaining a peak-to-valley difference according to the maximum load value and the minimum load value; Obtaining a peak-to-valley difference rate according to the peak-to-valley difference and the maximum load value; The average daily power consumption on working days and the average daily power consumption on rest days are obtained according to the standard daily power consumption data; and the power consumption ratio of working days to rest days is obtained according to the average daily power consumption on working days and the average daily power consumption on rest days.

4. The method according to claim 3, characterized in that The method of obtaining the trend component by using the regression period decomposition method according to the extracted characteristic data includes: Setting regression parameters of regression period decomposition, and obtaining initial trend components according to the regression parameters, maximum load value, minimum load value and average load value; The initial trend component is smoothed by using a moving average method to eliminate short-term noise in the initial trend component to obtain the trend component.

5. The method according to claim 4, characterized in that The method of obtaining the periodic component by using the regression periodic decomposition method according to the extracted characteristic data comprises: The regression coefficient of the dummy variable is set, and the periodic dummy variable is obtained according to the regression coefficient of the dummy variable, the peak-to-valley difference rate, and the working day-rest day power consumption ratio; Obtaining a designated component according to the standard daily electricity consumption data, the trend component and the periodic dummy variable; Decomposing the designated component by using the empirical mode decomposition method to extract the intermediate frequency IMF component therein, wherein the intermediate frequency IMF component is a periodic residual variable; The periodic residual variable and the periodic dummy variable are combined to obtain the periodic component.

6. The method according to claim 5, characterized in that The step of fusing the trend component prediction value, the period component prediction value and the random component prediction value to obtain the daily power consumption prediction value of the user to be predicted includes: The trend component weight, the period component weight and the random component weight are set respectively, and the trend component weight, the period component weight and the random component weight are used to combine the trend component prediction value, the period component prediction value and the random component prediction value to obtain the daily electricity consumption prediction value of the user to be predicted.

7. A device for double decomposing short-term daily electricity consumption of users, characterized in that: The device comprises: Data acquisition module: used to obtain the user's daily electricity consumption data within a preset historical time period; Preprocessing module: used to preprocess the daily electricity consumption data acquired within a preset historical time period to obtain standard daily electricity consumption data; Feature extraction module: used to extract feature data from the standard daily power consumption data; Regression period decomposition module: used to obtain trend component and period component by regression period decomposition method according to the extracted characteristic data; An initial random component acquisition module is used to acquire an initial random component according to the trend component and the period component; Noise disturbance module: used for adding different random noises to the initial random component to generate multiple disturbance versions; Empirical mode decomposition module: used to perform empirical mode decomposition on each disturbance version generated. Each empirical mode decomposition obtains high-frequency data, medium-frequency data and low-frequency data. The high-frequency data obtained by multiple decompositions are accumulated to obtain the final random component. Model training module: used to iteratively train the pre-built polynomial regression model framework through the trend component to obtain a trend component prediction model; iteratively train the pre-built autoregressive moving average model framework through the periodic component to obtain a periodic component prediction model; iteratively train the pre-built random forest model framework through the final random component to obtain a random component prediction model; Prediction module: used to output the trend component prediction value, the period component prediction value and the random component prediction value of the user to be predicted respectively through the trend component prediction model, the period component prediction model and the random component prediction model; Prediction result fusion module: used to fuse the trend component prediction value, the period component prediction value and the random component prediction value to obtain the daily power consumption prediction value of the user to be predicted.

8. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by the main controller, each step of the method for double decomposition of short-term daily electricity consumption of a user as described in any one of claims 1 to 6 is implemented.

Images