Power consumption data processing method and system based on time sequence correlation and trend fitting

Abstract

The invention belongs to the field of power consumption data processing, and provides a power consumption data processing method and system based on time sequence correlation and trend fitting. The method comprises the following steps: acquiring historical power consumption data and extracting power consumption characteristics from the historical power consumption data; performing time sequence association importance evaluation on each power consumption feature associated feature, and screening out an optimal time sequence power consumption feature set associated with the user power consumption features; capturing user behavior trend characteristics in the historical power consumption data by using a long-short time neural network algorithm, and encoding the user behavior trend characteristics and an optimal time sequence characteristic set associated with the user power consumption characteristics by using an encoder to obtain trend encoding data at a missing moment; and in combination with the environment state data, decoding the trend coding data at the missing moment by using a decoder to obtain a predicted value of the power consumption characteristic missing value of the user so as to complement missing data at the missing value moment. The power consumption data missing value complementation accuracy can be improved, so that the operation and maintenance efficiency of an intelligent power system is improved.

Description

Technical Field

[0001] The present invention belongs to the field of power consumption data processing, and in particular relates to a power consumption data processing method and system based on time series correlation and trend fitting. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] With the advancement of the construction of new power systems, facing the complex energy supply and demand optimization and balance requirements, business applications based on power data play an increasingly important role in the development of the power market. Therefore, the power grid operation has higher and higher requirements for real-time data quality, and it is necessary to obtain accurate measurement data from each node of the power system to the maximum extent. However, due to abnormal power system sensors and transmission line failures, there is a certain proportion of missing data in the massive power data, which will directly affect the analysis and decision-making performance of the intelligent power system. Missing value completion is an important method to improve the quality of power data.

[0004] At present, traditional methods mainly rely on data experts, business experts and other manual labor to complete missing values. However, as the scale of power data gradually expands, relying on traditional methods to complete missing values ​​for real-time heterogeneous data is gradually insufficient. At the same time, manual management will also cause errors and omissions, making it impossible to quickly and accurately meet the needs of business data use. In addition, relying on manpower to complete missing values ​​is costly and time-consuming, which seriously affects the operation and maintenance efficiency of intelligent power systems. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides a method and system for processing electricity consumption data based on time series correlation and trend fitting, which can improve the efficiency of electricity consumption data processing to enhance the operation and maintenance efficiency of the intelligent power system.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] A first aspect of the present invention provides a method for processing electricity consumption data based on time series correlation and trend fitting.

[0008] In one or more embodiments, a method for processing electricity consumption data based on time series correlation and trend fitting is provided, comprising:

[0009] Obtain historical electricity consumption data and extract electricity consumption features from it; evaluate the importance of time series correlation for each electricity consumption feature, and select the optimal time series electricity consumption feature set associated with the user's electricity consumption features;

[0010] A long-short time neural network is used to capture the user behavior trend characteristics in the historical electricity consumption data, and an encoder is used to encode the optimal time series feature set associated with the user behavior trend characteristics and the user electricity consumption characteristics to obtain the trend encoding data at the missing moment;

[0011] Combined with the environmental status data, the decoder is used to decode the trend coded data at the missing time to obtain the predicted value of the missing value of the user's electricity consumption characteristics to complete the missing data at the time of the missing value.

[0012] As an implementation method, the Forest God algorithm is used to evaluate the time-series correlation importance of each power consumption feature-related feature.

[0013] As an implementation method, based on the real features contained in each electricity consumption data subsample set, the importance index value of each electricity consumption feature is calculated using the information gain method, and the process is as follows:

[0014] According to the number of abnormal sample data and normal sample data in the historical electricity consumption data sample set, the information amount of the abnormal sample is identified;

[0015] In the initial electricity consumption feature set of electricity consumption data samples, missing data of each initial electricity consumption feature is completed, and the information amount of other features and abnormal samples in the initial electricity consumption feature set of electricity consumption data samples is calculated. The information gain rate of each initial electricity consumption feature is calculated and used as the importance index value of the corresponding initial electricity consumption feature. All electricity consumption features are traversed to obtain the importance index values ​​of all electricity consumption features.

[0016] As an implementation method, in the process of screening out the optimal time series power consumption feature set associated with the user's power consumption feature:

[0017] The collection time sequence of the real features contained in each electricity consumption data sub-sample set is disrupted to form shadow features;

[0018] For each power consumption feature, the information gain method is used to calculate the importance index value of the real feature and its corresponding shadow feature. If the importance index value of a power consumption feature is significantly higher than that of its corresponding shadow feature, then this power consumption feature is a time-series important feature; if the importance index value of a feature is significantly lower than that of its corresponding shadow feature, then this power consumption feature is marked as a time-series irrelevant feature and deleted from the feature set; until all feature screening is completed;

[0019] Delete all shadow attributes, obtain the determined important feature set, and compare it with the importance index values ​​of other power consumption features; until all features are screened, obtain the updated important feature set.

[0020] As an implementation method, the mean of the importance index values ​​of all electricity consumption features in the important feature set is calculated, and the importance index threshold is set as a set times the mean. Features exceeding the importance index threshold are retained as important features, and features below the importance index threshold are deleted.

[0021] As an implementation method, the loss function during the training of the long-short time neural network is:

[0022]

[0023] Among them, Loss is the loss function; ∑ represents the predicted value z after a batch of sample data is input. t With real data Cumulative sum of cross entropy loss values.

[0024] A second aspect of the present invention provides an electricity consumption data processing system based on time series correlation and trend fitting.

[0025] In one or more embodiments, a power consumption data processing system based on time series correlation and trend fitting includes:

[0026] The optimal time-series power consumption feature set screening module is used to obtain historical power consumption data and extract power consumption features from it; perform time-series correlation importance evaluation on each power consumption feature-related feature, and screen out the optimal time-series power consumption feature set associated with the user's power consumption features;

[0027] A missing moment trend coding data encoding module is used to capture the user behavior trend characteristics in the historical power consumption data using a long-short time neural network algorithm, and use an encoder to encode the user behavior trend characteristics and the optimal time series feature set associated with the user power consumption characteristics to obtain the trend coding data of the missing moment;

[0028] The power consumption feature missing value prediction value decoding module is used to combine the environmental status data and use the decoder to decode the trend encoding data at the missing time to obtain the prediction value of the user's power consumption feature missing value to complete the missing data at the missing value time.

[0029] A third aspect of the present invention provides a computer-readable storage medium.

[0030] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps in the above-mentioned method for processing electricity consumption data based on time series association and trend fitting.

[0031] A fourth aspect of the present invention provides an electronic device.

[0032] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the method for processing electricity consumption data based on time series correlation and trend fitting as described above are implemented.

[0033] A fifth aspect of the present invention provides a computer program product.

[0034] A computer program product includes a computer program / instruction, which, when executed by a processor, implements the steps in the above-mentioned method for processing electricity consumption data based on time series correlation and trend fitting.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] (1) In order to improve the data quality of collected electricity consumption data, the present invention proposes an abnormal data management method based on time series correlation and trend fitting to manage abnormal electricity consumption data. The method performs correlation analysis based on the collected electricity data from two aspects: user behavior and time series correlation data. Then, the mutual correlation relationship and behavior representation learning between the data are captured, and the missing value completion of the electricity data is completed through the encoder and decoder, thereby improving the quality of electricity consumption collection data.

[0037] (2) In the process of building complex models, there are too many irrelevant or weakly correlated features, which often lead to difficulties in model training and parameter adjustment and reduced performance. The present invention uses the Forest God algorithm to evaluate the time-series correlation importance of the associated features of each feature and select the optimal feature set, thereby improving the accuracy of modeling and completing various types of missing feature data.

[0038] (3) Since electricity consumption data will change with the changes in environmental conditions and human behavior, and the environmental conditions and user electricity consumption behavior characteristics are hidden in the currently acquired environmental data and part of the electricity consumption data, the present invention adopts a long-short time neural network algorithm to capture the user behavior trend characteristics, and combines the current collected data and environmental status data of the important time series related features of the missing features to predict and complete the missing data. That is, the historical trend characteristics of the features to be completed are taken into account, and the changing characteristics of the environment and related features at the time of missing are taken into account horizontally, thereby improving the effectiveness of data completion management. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0040] Figure 1 is a flow chart of a method for processing electricity consumption data based on time series association and trend fitting according to an embodiment of the present invention;

[0041] Figure 2 is a schematic diagram of a power consumption data processing process based on time series association and trend fitting according to an embodiment of the present invention;

[0042] Figure 3 is a schematic diagram of the structure of a power consumption data processing system based on time series association and trend fitting according to an embodiment of the present invention;

[0043] Figure 4 is a schematic diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0044] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0045] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0046] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0047] Figure 1 is a flow chart of a method for processing power consumption data based on time series association and trend fitting in an embodiment of the present invention, such as Figure 1 The power consumption data processing method based on time series correlation and trend fitting in the embodiment shown may include:

[0048] S101, obtaining historical electricity consumption data and extracting electricity consumption features from it; performing a time series correlation importance assessment on each electricity consumption feature-related feature, and screening out an optimal time series electricity consumption feature set associated with the user's electricity consumption features.

[0049] In the process of building a complex model, there are too many irrelevant or weakly correlated features, which often makes model training and parameter adjustment difficult and performance degraded. In this embodiment, the Forest God algorithm is used to evaluate the importance of time series correlation of each power consumption feature. This improves the accuracy of modeling and completion of various missing feature data.

[0050] In this embodiment, multiple random resamplings are performed based on the historical electricity consumption data sample set, and different electricity consumption data sub-sample sets are obtained each time. Based on the real features contained in each electricity consumption data sub-sample set, the importance index value of each feature is calculated using the information gain method.

[0051] Among them, based on the real features contained in each electricity consumption data sub-sample set, the information gain method is used to calculate the importance index value of each electricity consumption feature, and the process is as follows:

[0052] According to the number of abnormal sample data and normal sample data in the historical electricity consumption data sample set, the information amount of the abnormal sample is identified;

[0053] Assuming that the number of abnormal sample data and normal sample data in the historical electricity consumption data sample set is p and n respectively, the amount of information to accurately identify abnormal samples is I(p,n):

[0054]

[0055] In the initial electricity consumption feature set of electricity consumption data samples, missing data of each initial electricity consumption feature is completed, and the information amount of other features and abnormal samples in the initial electricity consumption feature set of electricity consumption data samples is calculated. The information gain rate of each initial electricity consumption feature is calculated and used as the importance index value of the corresponding initial electricity consumption feature. All electricity consumption features are traversed to obtain the importance index values ​​of all electricity consumption features.

[0056] In the initial feature set of the electricity consumption data sample, the current feature data z (i.e., a type of feature in the initial feature set of the electricity consumption data sample) is supplemented with missing data. Other features in the initial feature set of the electricity consumption data sample include time, temperature, sunshine intensity, power generation, power consumption, voltage, and communication bandwidth. z has v values ​​(z1, z2, L, z v ), the historical electricity consumption data sample set is divided into v subsets (H1, H2, L, H v ), assuming that the subset H i Contains p i abnormal sample data and n i normal sample data, where i∈[1,2,3,…,v], subset H i The information entropy E(H i ):

[0057]

[0058] The information entropy of feature z is E(A)

[0059]

[0060] The information gain of feature z is Gain(A):

[0061] Gain(z)=I(p,n)-E(z) (4)

[0062] The information gain rate is:

[0063]

[0064] Among them, the split information rate Split(z) is:

[0065]

[0066] Gain-Ratio(z) is used as the importance index value of feature z, and all features are traversed as described above to obtain the importance index values ​​of all features.

[0067] In the process of screening out the optimal time series power consumption feature set associated with the user's power consumption features:

[0068] The collection time sequence of the real features contained in each electricity consumption data sub-sample set is disrupted to form shadow features;

[0069] For each power consumption feature, the information gain method is used to calculate the importance index value of the real feature and its corresponding shadow feature. If the importance index value of a power consumption feature is significantly higher than that of its corresponding shadow feature, then this power consumption feature is a time-series important feature; if the importance index value of a feature is significantly lower than that of its corresponding shadow feature, then this power consumption feature is marked as a time-series irrelevant feature and deleted from the feature set; until all feature screening is completed;

[0070] Delete all shadow attributes, obtain the determined important feature set, and compare it with the importance index values ​​of other power consumption features; until all features are screened, obtain the updated important feature set.

[0071] Specifically, the mean of the importance index values ​​of all power consumption features in the important feature set is calculated, and the importance index threshold is set by setting a multiple (such as γ) of the mean value. Features exceeding the importance index threshold are retained as important features, and features below the importance index threshold are deleted. Among them, γ is set to 0.2 according to the statistics of the sample set data during the experiment.

[0072] Multiple rounds of iterations are performed, and different sub-sample sets of electricity consumption data are re-acquired in each round. The important features updated in the previous round and their corresponding shadow features are used to calculate the importance index value in this round, and compared with the importance index values ​​of other features until importance index values ​​are assigned to all features, or the number of iterations has reached the set number; through iteration and comparison, the important time series features associated with the z feature are screened out, and their output is used as the selection set of time series features of electricity consumption behavior.

[0073] S102, using a long short-term memory network (LSTM) to capture user behavior trend characteristics in historical electricity consumption data, using an encoder to encode the user behavior trend characteristics and the optimal time series feature set associated with the user electricity consumption characteristics, and obtaining trend encoding data at the missing moment.

[0074] like Figure 2 As shown in the figure, electricity consumption data will change with the changes in environmental conditions and human behavior, so traditional fixed data completion methods (such as mean or data collected at the previous moment) often limit the effectiveness of data governance. At the same time, the environmental conditions and user electricity consumption behavior characteristics are hidden in the currently acquired environmental data and part of the electricity consumption data, so LSTM is used to capture the missing values ​​of user electricity consumption features t and the behavioral trend feature encoding h in the previous period. t-1 , and use the encoder to h t-1 The time series feature matrix X of the associated power consumption behavior collected at the time t of the user's missing value t ′ is encoded to obtain e t , e t And the environmental state data τ at time t with missing values t The decoder is input together to obtain the predicted value z of the missing value of the user's electricity consumption feature t , using the predicted value to complete the missing data of the user's features at time t.

[0075] Among them, LSTM is used to capture the behavior trend feature encoding of the period before the missing value of the user's electricity consumption feature t, that is, the historical data of the user's feature z (z1, z2, ..., z t-1 ) and historical environmental status data (τ1, τ2, …, τ t-1 ) is composed of (x1,x2,…,x t-1 ), input LSTM model, where x = [τ, z], and obtain the latent vector encoding h representing the user behavior representation t-1 , the specific process is:

[0076] According to the three control gates (input gate, forget gate and output gate) of the gate control device in the LSTM model, the input sample set data is recognized and the state of the metering device is judged. The data processing process of the three control gates, memory information and output judgment is as follows:

[0077] Update the forget gate output:

[0078] f t-1 =σ(W f h t-2 +U f x t-1 +bf ) (7)

[0079] Update the two-part output of the input gate:

[0080] I t-1 =σ(W I h t-2 +U I x t-1 +b I ) (8)

[0081] a t-1 =tanh(W a h t-1 +U a x t-1 +b a ) (9)

[0082] Memory information status:

[0083] C t-1 =C t-2 ·f t-1 +I t-1 ·a t-1 (10)

[0084] Update the output gate output:

[0085] o t-1 =σ(W o h t-2 +U o x t-1 +b o ) (11)

[0086] h t-1 =ot -1 tanh(C t-1 ) (12)

[0087] Among them, W f , W I , W a , W o Represents the weight parameter matrices of the forget gate, input gate, input gate, and output gate of the hidden layer, U f , U I , U a , U o are the weight parameter matrices of the sample data input forget gate, input gate, input gate and output gate, respectively. t-2 is the previous moment of t-1, t-1 is the previous moment, t is the current moment, and b f 、b I 、b a 、b oThey are the weight parameter offset parameter matrices of the forget gate, input gate, input gate, and output gate respectively.

[0088] Then, the latent vector encoding h representing the user behavior representation is t-1 , the important time series feature matrix X′ associated with the z feature t Enter Encoder to get the encoding t , the code e t And the environmental state data τ at time t with missing values t Input the decoder together to obtain the predicted value z of the missing value of the user's electricity consumption feature t , according to the predicted value z t With real data Construct the loss function Loss, and iterate the feedback to continuously adjust the LSTM model, Encoder, and Decoder parameters to improve the accuracy of the completed data.

[0089] Among them, the loss function in the training process of long-short time neural network is:

[0090]

[0091] Among them, Loss is the loss function; ∑ represents the predicted value z after a batch of sample data is input. t With real data Cumulative sum of cross entropy loss values.

[0092] During the model training process, the true values ​​of sample features are masked to increase the amount of abnormal sample data and ensure the accuracy of loss function construction.

[0093] S103, combining the environmental status data, using a decoder to decode the trend coded data at the missing time, and obtain the predicted value of the missing value of the user's power consumption characteristics to complete the missing data at the missing value time.

[0094] In order to verify the effectiveness of the model proposed in this paper, more than 22,000 industrial and commercial users in a certain area from November 2023 to October 2024 were obtained from the power consumption information collection system of a provincial company of the State Grid. The data were collected once an hour and at 24 time points every day. In addition, the present invention selected 4656 power consumption data of industrial and commercial users in the area in November 2024 as the verification data of the model in this paper. In this paper, the data taken at each time point is regarded as a piece of power consumption data.

[0095] In order to verify the performance of the model proposed in the present invention in filling missing values ​​in data, the present invention selects 5 baseline models: Linear Interpolation (LI), Historical Average (HA), LSTM, GAN and ST-LRAGAN, and the present invention selects the root mean square error (RMSE) and the mean percentage absolute error (MAE) to evaluate the performance of the missing value filling method proposed in the present invention.

[0096] Table 1 Performance comparison of different missing value filling methods on RMSE

[0097]

[0098] Table 1 describes the comparison of different missing data value completion methods in terms of RMSE and MAE indicators when the data missing rate of feature z is 10%, 20%, 30%, and 40% respectively. The lower the error value, the closer the completed data is to the real data. It can be concluded from Table 1 that compared with the baseline model, the performance of the model proposed in this paper has achieved the highest results in both RMSE and MAE indicators. Because the model in this paper not only considers the complex correlation between user behavior and environmental data and user electricity consumption data, but also uses the Boruta algorithm to evaluate the time series correlation importance of the associated features of each feature, thereby improving the modeling efficiency and accuracy of the data completion model.

[0099] Figure 3 is a schematic diagram of a power consumption data processing system based on time series correlation and trend fitting in an embodiment of the present invention. Figure 1 The power consumption data processing method based on time series correlation and trend fitting corresponds to Figure 3 As shown, the power consumption data processing system based on time series association and trend fitting in this embodiment may include:

[0100] The optimal time-series power consumption feature set screening module 301 is used to obtain historical power consumption data and extract power consumption features from it; perform time-series correlation importance evaluation on each power consumption feature-related feature, and screen out the optimal time-series power consumption feature set associated with the user's power consumption features;

[0101] The missing moment trend coding data encoding module 302 is used to capture the user behavior trend characteristics in the historical power consumption data by using the long-short time neural network algorithm, and use the encoder to encode the user behavior trend characteristics and the optimal time series feature set associated with the user power consumption characteristics to obtain the trend coding data of the missing moment;

[0102] The power consumption feature missing value prediction value decoding module 303 is used to combine the environmental status data and use a decoder to decode the trend encoding data at the missing time to obtain the prediction value of the user's power consumption feature missing value to complete the missing data at the missing value time.

[0103] It should be noted here that Figure 3 The various modules in the power consumption data processing system based on time series correlation and trend fitting are Figure 1 The various steps in the electricity consumption data processing method based on time series association and trend fitting correspond to each other one by one, and the specific implementation process is the same, which will not be repeated here.

[0104] Reference Figure 4 , a schematic diagram of an electronic device is given. It should be noted that, Figure 4 The electronic device 400 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0105] like Figure 4 As shown, electronic device 400 includes a central processing unit (CPU) 401, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 402 or a program loaded from a storage portion 408 into a random access memory (RAM) 403. In RAM 403, various programs and data required for system operation are also stored. Central processing unit 401, ROM 402 and RAM 403 are connected to each other via bus 404. Input / output (I / O) interface 405 is also connected to bus 404.

[0106] The following components are connected to the I / O interface 405: an input section 406 including a keyboard, a mouse, etc.; an output section 407 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 408 including a hard disk, etc.; and a communication section 409 including a network interface card such as a local area network (LAN) card, a modem, etc. The communication section 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the I / O interface 405 as needed. A removable medium 411, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 410 as needed, so that a computer program read therefrom is installed into the storage section 408 as needed.

[0107] When the central processing unit 401 in the electronic device of this embodiment executes the program, the following is achieved: Figure 1 The steps in the electricity consumption data processing method based on time series correlation and trend fitting are shown.

[0108] In particular, according to an embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a computer readable medium, the computer program including a computer program for executing Figure 1 In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 409, and / or installed from the removable medium 411. When the computer program is executed by the central processing unit 401, various functions defined in the apparatus of the present application are executed.

[0109] in, Figure 1 The computer program instructions corresponding to the method shown may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0110] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0111] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for processing electricity consumption data based on time series correlation and trend fitting, characterized in that: include: Obtain historical electricity consumption data and extract electricity consumption characteristics from it; Perform a time-series correlation importance assessment on each power consumption feature-related feature, and select the optimal time-series power consumption feature set associated with the user's power consumption features; A long-short time neural network is used to capture the user behavior trend characteristics in the historical electricity consumption data, and an encoder is used to encode the optimal time series feature set associated with the user behavior trend characteristics and the user electricity consumption characteristics to obtain the trend encoding data at the missing moment; Combined with the environmental status data, the decoder is used to decode the trend coded data at the missing time to obtain the predicted value of the missing value of the user's electricity consumption characteristics to complete the missing data at the time of the missing value.

2. The method for processing electricity consumption data based on time series correlation and trend fitting according to claim 1, characterized in that: The Forest God algorithm is used to evaluate the importance of time series correlation of each power consumption feature.

3. The method for processing electricity consumption data based on time series correlation and trend fitting according to claim 2, characterized in that: Based on the real features contained in each electricity consumption data sub-sample set, the information gain method is used to calculate the importance index value of each electricity consumption feature. The process is as follows: According to the number of abnormal sample data and normal sample data in the historical electricity consumption data sample set, the information amount of the abnormal sample is identified; In the initial electricity consumption feature set of electricity consumption data samples, missing data of each initial electricity consumption feature is completed, and the information amount of other features and abnormal samples in the initial electricity consumption feature set of electricity consumption data samples is calculated. The information gain rate of each initial electricity consumption feature is calculated and used as the importance index value of the corresponding initial electricity consumption feature. All electricity consumption features are traversed to obtain the importance index values ​​of all electricity consumption features.

4. The method for processing electricity consumption data based on time series correlation and trend fitting according to claim 1, characterized in that: In the process of screening out the optimal time series power consumption feature set associated with the user's power consumption features: The collection time sequence of the real features contained in each electricity consumption data sub-sample set is disrupted to form shadow features; For each power consumption feature, the information gain method is used to calculate the importance index value of the real feature and its corresponding shadow feature. If the importance index value of a power consumption feature is significantly higher than that of its corresponding shadow feature, then this power consumption feature is an important time series feature. If the importance index value of a feature is significantly lower than that of its corresponding shadow feature, the power consumption feature is marked as a time-independent feature and deleted from the feature set until all features are screened; Delete all shadow attributes, obtain the determined important feature set, and compare it with the importance index values ​​of other power consumption features; until all features are screened, obtain the updated important feature set.

5. The method for processing electricity consumption data based on time series correlation and trend fitting according to claim 4, characterized in that: Calculate the mean of the importance index values ​​of all electricity consumption features in the important feature set, set the importance index threshold by multiple of the mean, and retain the important features if they exceed the importance index threshold, and delete the features if they are lower than the importance index threshold.

6. The method for processing electricity consumption data based on time series correlation and trend fitting according to claim 1, characterized in that: The loss function during the training of the long-short time neural network is: Among them, Loss is the loss function; ∑ represents the predicted value z after a batch of sample data is input. t With real data Cumulative sum of cross entropy loss values.

7. A power consumption data processing system based on time series correlation and trend fitting, characterized in that: include: An optimal time-series electricity consumption feature set screening module is used to obtain historical electricity consumption data and extract electricity consumption features from it; Perform a time-series correlation importance assessment on each power consumption feature-related feature, and select the optimal time-series power consumption feature set associated with the user's power consumption features; A missing moment trend coding data encoding module is used to capture the user behavior trend characteristics in the historical power consumption data using a long-short time neural network algorithm, and use an encoder to encode the user behavior trend characteristics and the optimal time series feature set associated with the user power consumption characteristics to obtain the trend coding data of the missing moment; The power consumption feature missing value prediction value decoding module is used to combine the environmental status data and use the decoder to decode the trend encoding data at the missing time to obtain the prediction value of the user's power consumption feature missing value to complete the missing data at the missing value time.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps in the method for processing electricity consumption data based on time series correlation and trend fitting as described in any one of claims 1 to 6 are implemented.

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 processor executes the program, the steps in the method for processing electricity consumption data based on time series correlation and trend fitting are implemented as described in any one of claims 1 to 6.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, the steps in the method for processing electricity consumption data based on time series correlation and trend fitting as described in any one of claims 1 to 6 are implemented.

Images