Energy consumption data correction method and system based on machine learning algorithm
Through the energy consumption data correction method based on machine learning algorithm, the problem of missing or abnormal data collection of energy consumption data is solved, and the stability recovery of energy consumption system and the accuracy of energy management decisions are achieved.
Patent Information
- Application Number
- CN202411188810.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-28
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-08-28
AI Technical Summary
Existing energy consumption data acquisition systems are prone to missing or abnormal data in failure or network fluctuations, affecting the accuracy and stability of energy consumption management, and thus affecting decision-making and economic benefits.
Using the energy consumption data correction method based on machine learning algorithm, by obtaining the original data, anchoring the energy consumption data to be corrected, data segmentation and local timing feature extraction are performed, and forward and backward LSTM models are used to combine the eigenvalue granularity correlation mask. The attention fusion module can be distinguished and generates a two-way energy consumption time-sequence propagation fusion representation vector, which is used to determine the decoded value of the energy consumption data correction.
Effectively correct abnormal energy consumption data, restore the stability of the energy consumption system, provide a more accurate data basis for energy management, and improve the reliability of energy management decisions.
Smart Images

Figure CN118710475B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of intelligent data correction, and specifically to an energy consumption data correction method and system based on a machine learning algorithm. Background Art
[0002] As the world's attention to energy efficiency and sustainable development continues to increase, the efforts of enterprises and institutions in energy conservation and emission reduction are particularly important. As the core tool to achieve this goal, the stability and accuracy of the energy consumption system directly affect the effect of energy management. Effective energy consumption management can not only reduce operating costs, but also improve resource utilization efficiency and promote environmental protection.
[0003] In practical applications, the collection of energy consumption data usually relies on various sensors and monitoring devices. However, these devices may encounter failures during operation, or data collection may be unstable due to network fluctuations, resulting in missing or abnormal data. These problems not only affect the normal operation of the energy consumption system, but may also lead to wrong decisions, which in turn affects the economic benefits and social responsibility of the enterprise.
[0004] Therefore, an energy consumption data correction solution is desired. Summary of the invention
[0005] This application is made in consideration of the above problems. One purpose of this application is to provide an energy consumption data correction method and system based on a machine learning algorithm.
[0006] The embodiment of the present application provides an energy consumption data correction method based on a machine learning algorithm, which includes:
[0007] Get the original data;
[0008] Anchoring the energy consumption data to be corrected in the original data;
[0009] Based on the energy consumption data to be corrected, data segmentation is performed on the original data to obtain previous energy consumption history data and subsequent energy consumption extension data;
[0010] Performing local time series feature extraction on the previous energy consumption history data and the subsequent energy consumption extension data respectively to obtain a sequence of previous energy consumption local time series feature vectors and a sequence of subsequent energy consumption local time series feature vectors;
[0011] Inputting the sequence of the previous energy consumption local time series feature vectors into the forward LSTM model to obtain the previous energy consumption time series propagation representation vector;
[0012] Inputting the sequence of the local time series feature vectors of the subsequent energy consumption into a backward LSTM model to obtain a backward energy consumption time series propagation representation vector;
[0013] Inputting the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector into a feature value granularity association mask distinguishable attention fusion module to obtain a bidirectional energy consumption time series propagation fusion representation vector;
[0014] Based on the bidirectional energy consumption time series propagation fusion representation vector, a corrected decoding value of the energy consumption data is determined.
[0015] For example, according to the energy consumption data correction method based on the machine learning algorithm of the embodiment of the present application, the local time series feature extraction is performed on the previous energy consumption history data and the subsequent energy consumption extension data to obtain a sequence of the previous energy consumption local time series feature vector and a sequence of the subsequent energy consumption local time series feature vector, including:
[0016] The previous energy consumption history data and the subsequent energy consumption extension data are respectively input into the energy consumption local time series feature extractor based on the 1D-CNN model to obtain a sequence of the previous energy consumption local time series feature vectors and a sequence of the subsequent energy consumption local time series feature vectors.
[0017] For example, according to the energy consumption data correction method based on the machine learning algorithm of the embodiment of the present application, the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector are input into the feature value granularity association mask distinguishable attention fusion module to obtain a bidirectional energy consumption time series propagation fusion representation vector, including:
[0018] Calculating a global eigenvalue association matrix between the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector to obtain a previous-backward energy consumption global eigenvalue association matrix;
[0019] Inputting the global eigenvalue association matrix of the previous-backward energy consumption into a learnable gating function to obtain a distinguishable weight matrix of the previous-backward energy consumption correlation;
[0020] Using the previous-backward energy consumption correlation distinguishable weight matrix as a fusion mask matrix, respectively calculating the matrix products between the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector and the previous-backward energy consumption correlation distinguishable weight matrix to obtain a distinguished and enhanced previous energy consumption time series propagation representation vector and a distinguished and enhanced backward energy consumption time series propagation representation vector;
[0021] The differentiated enhanced previous energy consumption time series propagation representation vector and the differentiated enhanced backward energy consumption time series propagation representation vector are fused to obtain the bidirectional energy consumption time series propagation fusion representation vector.
[0022] For example, according to the energy consumption data correction method based on the machine learning algorithm of the embodiment of the present application, the global eigenvalue association matrix between the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector is calculated to obtain the previous-backward energy consumption global eigenvalue association matrix, including:
[0023] The vector multiplication between the transposed vector of the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector is calculated to obtain the previous-backward energy consumption global eigenvalue association matrix.
[0024] For example, according to the energy consumption data correction method based on the machine learning algorithm of the embodiment of the present application, the previous-backward energy consumption global eigenvalue association matrix is input into a learnable gating function to obtain a previous-backward energy consumption correlation distinguishable weight matrix, including:
[0025] Taking the negative number of each position eigenvalue in the global eigenvalue association matrix of the previous-backward energy consumption as the exponential power, calculating the value of the natural exponential function with the natural constant e as the base to obtain the global eigenvalue association class support matrix of the previous-backward energy consumption;
[0026] Calculate the reciprocal of the sum of the eigenvalues of each position in the previous-backward energy consumption global eigenvalue association class support matrix and the constant 1 to obtain a previous-backward energy consumption global eigenvalue association weight matrix;
[0027] The previous-backward energy consumption correlation distinguishable weight matrix is obtained by calculating the position point multiplication between the previous-backward energy consumption global eigenvalue association weight matrix and the previous-backward energy consumption global eigenvalue association matrix.
[0028] For example, according to the energy consumption data correction method based on the machine learning algorithm of the embodiment of the present application, the distinguishing and strengthening previous energy consumption time series propagation representation vector and the distinguishing and strengthening backward energy consumption time series propagation representation vector are fused to obtain the bidirectional energy consumption time series propagation fusion representation vector, including:
[0029] The differentiated enhanced previous energy consumption time series propagation representation vector and the differentiated enhanced backward energy consumption time series propagation representation vector are cascaded to obtain the bidirectional energy consumption time series propagation fusion representation vector.
[0030] For example, according to the energy consumption data correction method based on the machine learning algorithm of the embodiment of the present application, the energy consumption data correction decoding value is determined based on the bidirectional energy consumption time series propagation fusion representation vector, including:
[0031] The bidirectional energy consumption time series propagation fusion representation vector is input into a decoder-based data corrector to obtain the energy consumption data correction decoding value.
[0032] The embodiment of the present application also provides an energy consumption data correction system based on a machine learning algorithm, which includes:
[0033] A data acquisition module, used to acquire original data;
[0034] An anchoring module, used for anchoring the energy consumption data to be corrected in the original data;
[0035] A data segmentation module, configured to segment the original data based on the energy consumption data to be corrected to obtain previous energy consumption history data and subsequent energy consumption extension data;
[0036] A local time series feature extraction module, used to extract local time series features from the previous energy consumption history data and the subsequent energy consumption extension data to obtain a sequence of local time series feature vectors of previous energy consumption and a sequence of local time series feature vectors of subsequent energy consumption;
[0037] A forward time series propagation module, used for inputting the sequence of the previous energy consumption local time series feature vectors into a forward LSTM model to obtain a previous energy consumption time series propagation representation vector;
[0038] A backward time series propagation module, used for inputting the sequence of the backward energy consumption local time series feature vectors into a backward LSTM model to obtain a backward energy consumption time series propagation representation vector;
[0039] A fusion module, configured to input the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector into a feature value granularity association mask distinguishable attention fusion module to obtain a bidirectional energy consumption time series propagation fusion representation vector;
[0040] The energy consumption data correction module is used to determine the energy consumption data correction decoding value based on the bidirectional energy consumption time series propagation fusion representation vector.
[0041] For example, according to the energy consumption data correction system based on machine learning algorithm according to an embodiment of the present application, the local time series feature extraction module is used to:
[0042] The previous energy consumption history data and the subsequent energy consumption extension data are respectively input into the energy consumption local time series feature extractor based on the 1D-CNN model to obtain a sequence of the previous energy consumption local time series feature vectors and a sequence of the subsequent energy consumption local time series feature vectors.
[0043] For example, according to the energy consumption data correction system based on machine learning algorithm in an embodiment of the present application, the fusion module is used to:
[0044] Calculating a global eigenvalue association matrix between the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector to obtain a previous-backward energy consumption global eigenvalue association matrix;
[0045] Inputting the global eigenvalue association matrix of the previous-backward energy consumption into a learnable gating function to obtain a distinguishable weight matrix of the previous-backward energy consumption correlation;
[0046] Using the previous-backward energy consumption correlation distinguishable weight matrix as a fusion mask matrix, respectively calculating the matrix products between the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector and the previous-backward energy consumption correlation distinguishable weight matrix to obtain a distinguished and enhanced previous energy consumption time series propagation representation vector and a distinguished and enhanced backward energy consumption time series propagation representation vector;
[0047] The differentiated enhanced previous energy consumption time series propagation representation vector and the differentiated enhanced backward energy consumption time series propagation representation vector are fused to obtain the bidirectional energy consumption time series propagation fusion representation vector.
[0048] According to the energy consumption data correction method and system based on machine learning algorithm of the embodiment of the present application, it can effectively correct abnormal energy consumption data based on the forward and backward contextual temporal semantics of the energy consumption data, restore the stability of the energy consumption system, and provide a more accurate data basis for subsequent energy management decisions. BRIEF DESCRIPTION OF THE DRAWINGS
[0049] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings of the embodiments of the present application are briefly introduced below. Obviously, the drawings described below only relate to some embodiments of the present application, and are not intended to limit the present application.
[0050] Figure 1 A schematic diagram of the application architecture of the energy consumption data correction method based on a machine learning algorithm in an embodiment of the present application is shown;
[0051] Figure 2 A flow chart of an energy consumption data correction method based on a machine learning algorithm in an embodiment of the present application is shown;
[0052] Figure 3 A flowchart of sub-step S570 of the energy consumption data correction method based on a machine learning algorithm in an embodiment of the present application is shown;
[0053] Figure 4 A schematic diagram showing the structure of an energy consumption data correction system based on a machine learning algorithm in an embodiment of the present application is shown; and
[0054] Figure 5 An application scenario diagram of the energy consumption data correction method based on a machine learning algorithm in an embodiment of the present application is shown. DETAILED DESCRIPTION
[0055] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work also fall within the scope of protection of the present application.
[0056] The terms used in this specification are those common terms currently widely used in the art in consideration of the functions of the present application, but these terms may vary according to the intentions of those skilled in the art, precedents, or new technologies in the art. In addition, specific terms may be selected, and in this case, their detailed meanings will be described in the detailed description of the present application. Therefore, the terms used in the specification should not be understood as simple names, but rather as a general description based on the meaning of the terms and the present application.
[0057] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.
[0058] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. At the same time, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.
[0059] Figure 1 A schematic diagram of the application architecture of the energy consumption data correction method based on a machine learning algorithm in an embodiment of the present application is shown, including a server 100 and a terminal device 200.
[0060] The terminal device 200 and the server 100 can be connected via the Internet to achieve mutual communication. Optionally, the above-mentioned Internet uses standard communication technology and / or protocol. The Internet is usually the Internet, but it can also be any network, including but not limited to any combination of a local area network (LAN), a metropolitan area network (MAN), a wide area network (WAN), a mobile, wired or wireless network, a private network or a virtual private network. In some embodiments, the data exchanged through the network can be represented by technologies and / or formats including Hyper Text Markup Language (HTML), Extensible Markup Language (XML), etc. In addition, conventional encryption technologies such as Secure Socket Layer (SSL), Transport Layer Security (TLS), Virtual Private Network (VPN), Internet Protocol Security (IPsec), etc. can be used to encrypt all or some links. In other embodiments, customized and / or dedicated data communication technologies can also be used to replace or supplement the above-mentioned data communication technologies.
[0061] The server 100 can provide various network services for the terminal device 200, wherein the server 100 can be a single server, a server cluster consisting of several servers, or a cloud computing center. Specifically, the server 100 may include a processor 110 (Center Processing Unit, CPU), a memory 120, an input device 130, and an output device 140, etc. The input device 130 may include a keyboard, a mouse, a touch screen, etc., and the output device 140 may include a display device, such as a liquid crystal display (Liquid Crystal Display, LCD), a cathode ray tube (Cathode Ray Tube, CRT), etc.
[0062] The memory 120 may include a read-only memory (ROM) and a random access memory (RAM), and provides the processor 110 with program instructions and data stored in the memory 120. In the embodiment of the present application, the memory 120 may be used to store the program of the energy consumption data correction method based on the machine learning algorithm in the embodiment of the present application.
[0063] The processor 110 calls the program instructions stored in the memory 120, and the processor 110 is used to execute the steps of any one of the energy consumption data correction methods based on the machine learning algorithm in the embodiments of the present application according to the obtained program instructions.
[0064] In addition, the application architecture diagram in the embodiment of the present application is intended to more clearly illustrate the technical solution in the embodiment of the present application, and does not constitute a limitation on the technical solution provided in the embodiment of the present application. Of course, for other application architectures and business applications, the technical solution provided in the embodiment of the present application is also applicable to similar problems.
[0065] The following is a non-restrictive explanation of the energy consumption data correction method based on a machine learning algorithm provided according to at least one embodiment of the present application through several examples or embodiments. As described below, different features in these specific examples or embodiments can be combined with each other without conflicting with each other to obtain new examples or embodiments, and these new examples or embodiments also fall within the scope of protection of the present application.
[0066] In response to the above technical problems, in the technical solution of the present application, a method for correcting energy consumption data based on a machine learning algorithm is proposed, which can group the collected original energy consumption data based on the energy consumption data to be corrected, so as to determine the previous energy consumption history data before the energy consumption data to be corrected and the subsequent energy consumption extension data. Then, a data processing and analysis algorithm based on artificial intelligence and machine learning is introduced at the back end to perform forward and backward time series analysis and information propagation fusion on the previous energy consumption history data and the subsequent energy consumption extension data, so as to obtain the energy consumption data correction decoding value based on the contextual corrected energy consumption data time series semantics. In this way, the energy consumption abnormal data can be effectively corrected based on the forward and backward contextual time series semantics of the energy consumption data, and the stability of the energy consumption system can be restored, providing a more accurate data basis for subsequent energy management decisions.
[0067] Figure 2 The flowchart of the energy consumption data correction method based on the machine learning algorithm in the embodiment of the present application is shown. For example, the energy consumption data correction method based on the machine learning algorithm can be executed by a server, which can be Figure 1 The server 100 shown in FIG. Figure 2As shown, according to the energy consumption data correction method based on the machine learning algorithm of the embodiment of the present application, the method includes the following steps: S510, acquiring original data; S520, anchoring the energy consumption data to be corrected in the original data; S530, based on the energy consumption data to be corrected, performing data segmentation on the original data to obtain previous energy consumption history data and subsequent energy consumption extension data; S540, respectively performing local time series feature extraction on the previous energy consumption history data and the subsequent energy consumption extension data to obtain a sequence of local time series feature vectors of previous energy consumption and a sequence of local time series feature vectors of subsequent energy consumption; S550, The sequence of previous energy consumption local time series feature vectors is input into the forward LSTM model to obtain the previous energy consumption time series propagation representation vector; S560, the sequence of subsequent energy consumption local time series feature vectors is input into the backward LSTM model to obtain the backward energy consumption time series propagation representation vector; S570, the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector are input into the feature value granularity association mask distinguishable attention fusion module to obtain the bidirectional energy consumption time series propagation fusion representation vector; S580, based on the bidirectional energy consumption time series propagation fusion representation vector, the energy consumption data correction decoding value is determined.
[0068] Specifically, in the technical solution of the present application, first, the original data is acquired, then, the energy consumption data to be corrected is anchored in the original data, and based on the energy consumption data to be corrected, the original data is segmented to obtain previous energy consumption history data and subsequent energy consumption extension data.
[0069] Then, the previous energy consumption history data and the subsequent energy consumption extension data are respectively input into the energy consumption local time series feature extractor based on the 1D-CNN model for feature mining, so as to respectively extract the local time series feature information of the previous energy consumption history data and the subsequent energy consumption extension data in the time dimension, thereby obtaining a sequence of the previous energy consumption local time series feature vector and a sequence of the subsequent energy consumption local time series feature vector. It should be understood that by performing independent time series coding processing on the previous and subsequent energy consumption data of the energy consumption data to be corrected, the local time series pattern and change trend of the previous energy consumption history data, as well as the local time series pattern and change law of the subsequent energy consumption extension data can be captured, which helps to improve the model's ability to understand the dynamic changes of the local time series of energy consumption data, thereby providing a basis for subsequent energy consumption data correction and decision support.
[0070] Correspondingly, in step S540, local time series feature extraction is performed on the previous energy consumption history data and the subsequent energy consumption extension data respectively to obtain a sequence of previous energy consumption local time series feature vectors and a sequence of subsequent energy consumption local time series feature vectors, including: inputting the previous energy consumption history data and the subsequent energy consumption extension data into an energy consumption local time series feature extractor based on a 1D-CNN model respectively to obtain a sequence of previous energy consumption local time series feature vectors and a sequence of subsequent energy consumption local time series feature vectors.
[0071] Furthermore, considering that the 1D-CNN model can capture the characteristic information of energy consumption data in the time dimension, due to the inherent limitations of the one-dimensional convolutional neural network, it can only perceive and understand the local temporal dynamic characteristics and change patterns of the previous energy consumption history data and the subsequent energy consumption extension data in the time dimension. Based on this, in order to better understand the temporal change laws and dynamic characteristics of the previous energy consumption history data and the subsequent energy consumption extension data, in the technical solution of the present application, the sequence of the local temporal feature vectors of the previous energy consumption is further input into the forward LSTM model to obtain the previous energy consumption temporal propagation representation vector, and the sequence of the local temporal feature vectors of the subsequent energy consumption is input into the backward LSTM model to obtain the backward energy consumption temporal propagation representation vector. It should be understood that the LSTM model is particularly suitable for processing long-term dependencies in time series data. By performing forward LSTM encoding and backward LSTM encoding processing on previous energy consumption history data and subsequent energy consumption extension data respectively, it can capture the time series long-distance dependency association feature information from the past to the time point of the energy consumption data to be corrected and the time series long-distance dependency association feature information from the future to the time point of the energy consumption data to be corrected respectively. This can effectively learn the temporal evolution of energy consumption data and the dynamic forward and backward time series characteristics, so as to better understand the dynamic changes of energy consumption data, so that the model can more accurately process and correct energy consumption data.
[0072] It should be understood that the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector respectively contain the forward and backward time series feature information about the energy consumption data before and after the energy consumption data to be corrected, and the correction of energy consumption data needs to consider not only the pattern of historical data, but also the dynamic change pattern of future data. Based on this, in the technical solution of the present application, the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector are further input into the feature value granularity association mask distinguishable attention fusion module to obtain a bidirectional energy consumption time series propagation fusion representation vector. Through the processing of the feature value granularity association mask distinguishable attention fusion module, the previous and backward energy consumption time series propagation representation vectors can be fused to integrate the long-distance dependent correlation feature information of the energy consumption data in these two directions, enhance the expression ability of the energy consumption data time series features, and form a more comprehensive energy consumption data time series feature representation. In addition, the feature value granularity association mask distinguishable attention mechanism can dynamically adjust the focus according to the fine-grained importance of the input energy consumption forward time series features and backward time series features. Through this mechanism, the model can automatically identify and focus on the features that are most helpful for correcting energy consumption data, improve the model's overall understanding of the time series pattern of energy consumption data, and help subsequent energy consumption data correction tasks.
[0073] Specifically, first, the global eigenvalue association matrix between the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector is calculated to quantify the interaction and correlation between the previous energy consumption time series propagation time series features and the backward energy consumption time series propagation features, providing a basis for the subsequent fine-grained feature importance identification and adaptive weight allocation. This step establishes a global relationship model between the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector, which is the basis of the entire fusion process. Subsequently, the global eigenvalue association matrix is input into a learnable gating function, which can learn the eigenvalue granularity correlation relationship and interaction between the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector, and output a weight matrix with distinguishable correlation. The introduction of the gating mechanism enables the model to dynamically adjust the weights between the eigenvalues in the two eigenvectors, realize feature selection and enhancement, so as to highlight the important features in the previous energy consumption time series propagation features and the backward energy consumption time series propagation features and suppress unimportant features. This step is the key to achieving the differentiation and enhancement of feature vectors. Through the self-learning ability of the gating function, the model can adaptively adjust the weights between the forward and backward time series features of energy consumption. Next, the obtained weight matrix is used as a fusion mask matrix to perform a matrix product operation with the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector to obtain the differentiated and enhanced feature vector of the two. This step achieves the differentiated enhancement of the feature vectors of the previous energy consumption time series propagation information and the backward energy consumption time series propagation information by weighting the fusion mask matrix, further highlighting the key information in the feature vector. The application of the fusion mask matrix not only considers the relative importance between the forward and backward time series features of energy consumption, but also considers the contribution of each internal element of the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector, so that the differentiated and enhanced feature vector can more comprehensively reflect the previous energy consumption time series features and the backward energy consumption time series features of the energy consumption data to be corrected. Finally, the differentiated enhanced previous energy consumption time series propagation representation vector and the differentiated enhanced backward energy consumption time series propagation representation vector after differentiation and enhancement are fused to obtain the final bidirectional energy consumption time series propagation fusion representation vector. This step integrates the useful information in the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector to form a richer and more discriminative feature representation. Through the processing of the feature value granularity association mask distinguishable attention fusion module, the fine-grained key features in the previous energy consumption time series propagation feature information and the backward energy consumption time series propagation feature information can be effectively integrated to help the model ignore unimportant or interfering information, thereby improving the accuracy of the correction and the generalization ability of the model.
[0074] Accordingly, in step S570, if Figure 3As shown, the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector are input into the eigenvalue granularity association mask distinguishable attention fusion module to obtain a bidirectional energy consumption time series propagation fusion representation vector, including: S571, calculating the global eigenvalue association matrix between the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector to obtain a previous-backward energy consumption global eigenvalue association matrix; S572, inputting the previous-backward energy consumption global eigenvalue association matrix into a learnable gating function to obtain a previous-backward energy consumption correlation distinguishable weight matrix ; S573, using the previous-backward energy consumption correlation distinguishable weight matrix as the fusion mask matrix, respectively calculate the matrix products between the previous energy consumption timing propagation representation vector and the backward energy consumption timing propagation representation vector and the previous-backward energy consumption correlation distinguishable weight matrix to obtain the distinguished and enhanced previous energy consumption timing propagation representation vector and the distinguished and enhanced backward energy consumption timing propagation representation vector; S574, fuse the distinguished and enhanced previous energy consumption timing propagation representation vector and the distinguished and enhanced backward energy consumption timing propagation representation vector to obtain the bidirectional energy consumption timing propagation fusion representation vector.
[0075] Among them, in step S571, the global eigenvalue association matrix between the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector is calculated to obtain the previous-backward energy consumption global eigenvalue association matrix, including: calculating the vector multiplication between the transposed vector of the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector to obtain the previous-backward energy consumption global eigenvalue association matrix.
[0076] Among them, in step S572, the previous-backward energy consumption global eigenvalue association matrix is input into a learnable gating function to obtain a previous-backward energy consumption correlation distinguishable weight matrix, including: using the negative number of the eigenvalues at each position in the previous-backward energy consumption global eigenvalue association matrix as the exponential power, calculating the natural exponential function value with the natural constant e as the base to obtain the previous-backward energy consumption global eigenvalue association class support matrix; calculating the reciprocal of the sum of the eigenvalues at each position in the previous-backward energy consumption global eigenvalue association class support matrix and the constant 1 to obtain the previous-backward energy consumption global eigenvalue association weight matrix; calculating the position point multiplication between the previous-backward energy consumption global eigenvalue association weight matrix and the previous-backward energy consumption global eigenvalue association matrix to obtain the previous-backward energy consumption correlation distinguishable weight matrix.
[0077] Among them, in step S574, the differentiated and enhanced previous energy consumption time series propagation representation vector and the differentiated and enhanced backward energy consumption time series propagation representation vector are fused to obtain the bidirectional energy consumption time series propagation fusion representation vector, including: cascading the differentiated and enhanced previous energy consumption time series propagation representation vector and the differentiated and enhanced backward energy consumption time series propagation representation vector to obtain the bidirectional energy consumption time series propagation fusion representation vector.
[0078] In a specific example, the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector are input into the eigenvalue granularity associated mask distinguishable attention fusion module to obtain the bidirectional energy consumption time series propagation fusion representation vector, including: inputting the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector into the eigenvalue granularity associated mask distinguishable attention fusion module to process them with the following distinguishable attention fusion formula to obtain the bidirectional energy consumption time series propagation fusion representation vector; wherein the distinguishable attention fusion formula is:
[0079]
[0080] in, and are the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector, respectively, is the global eigenvalue association matrix of the previous-backward energy consumption, represents the transpose of a vector, For The exponential operation with base , is the global eigenvalue association weight matrix of the previous-backward energy consumption, is the distinguishable weight matrix of the previous-backward energy consumption correlation, and are respectively distinguishing and strengthening the previous energy consumption time series propagation representation vector and distinguishing and strengthening the backward energy consumption time series propagation representation vector, is the cascade operation of vectors, is a representation vector for the bidirectional energy consumption time series propagation fusion.
[0081] Furthermore, the bidirectional energy consumption time series propagation fusion representation vector is input into the decoder-based data corrector to obtain the energy consumption data correction decoding value. In other words, the forward and backward bidirectional energy consumption time series propagation fusion feature information of the energy consumption data to be corrected is used to perform decoding regression, so as to obtain the energy consumption data correction decoding value based on the context-based corrected energy consumption data time series semantics. In this way, the energy consumption abnormal data can be effectively corrected based on the forward and backward context time series semantics of the energy consumption data, the stability of the energy consumption system can be restored, and a more accurate data basis can be provided for subsequent energy management decisions.
[0082] Accordingly, in step S580, based on the bidirectional energy consumption time series propagation fusion representation vector, the energy consumption data correction decoding value is determined, including: inputting the bidirectional energy consumption time series propagation fusion representation vector into a decoder-based data corrector to obtain the energy consumption data correction decoding value.
[0083] Furthermore, considering that the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector respectively represent the local time series implicit association features of the previous energy consumption historical data and the subsequent energy consumption extension data in the global time domain, the time series context association features based on the forward LSTM model and the backward LSTM model. In this way, after the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector are input into the feature value granularity association mask distinguishable attention fusion module, the obtained bidirectional energy consumption time series propagation fusion representation vector will also be caused by the feature vector association fusion offset caused by the difference in the association distinguishable attention weight caused by the fine-grained difference in the feature distribution time series between the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector. Therefore, it is expected to further improve the feature vector association fusion regression comprehensibility of the bidirectional energy consumption time series propagation fusion representation vector, thereby improving the accuracy of the energy consumption data correction decoding value obtained by the decoder-based data corrector.
[0084] In a preferred example, inputting the bidirectional energy consumption time series propagation fusion representation vector into a decoder-based data corrector to obtain an energy consumption data correction decoded value includes:
[0085] Subtract the zero norm of the bidirectional energy consumption time series propagation fusion representation vector from the length of the bidirectional energy consumption time series propagation fusion representation vector to obtain a bidirectional energy consumption time series propagation fusion isolated representation value;
[0086] Calculate the logarithm of the sum of the square of the two-way energy consumption time series propagation fusion isolated representation value and the two-way energy consumption time series propagation fusion isolated representation value with base 2 to obtain the two-way energy consumption time series propagation fusion information order value;
[0087] Calculate a power function with each eigenvalue of the two-way energy consumption time series propagation fusion representation vector as the base, the difference of the two-way energy consumption time series propagation fusion isolated representation value minus one as the exponential value, and perform point multiplication with the two-way energy consumption time series propagation fusion information order value to obtain the two-way energy consumption time series propagation fusion leading vector;
[0088] After performing a dot multiplication on the bidirectional energy consumption time series propagation fusion representation vector and the difference between the bidirectional energy consumption time series propagation fusion isolated representation value minus one, the dot multiplication is performed on the inverse of the bidirectional energy consumption time series propagation fusion isolated representation value to obtain a bidirectional energy consumption time series propagation fusion field constraint vector;
[0089] Calculate an exponential function with a natural constant as the base and each eigenvalue of the bidirectional energy consumption time series propagation fusion field constraint vector as the exponent to obtain a bidirectional energy consumption time series propagation fusion field bias vector;
[0090] Performing dot addition of the bidirectional energy consumption time series propagation fusion leading vector and the bidirectional energy consumption time series propagation fusion field bias vector to obtain an optimized bidirectional energy consumption time series propagation fusion representation vector; and
[0091] The optimized bidirectional energy consumption time series propagation fusion representation vector is input into a decoder-based data corrector to obtain an energy consumption data correction decoding value.
[0092] Specifically, the bidirectional energy consumption time series propagation fusion representation vector is, for example, recorded as The optimization process is expressed as:
[0093]
[0094] in, represents the bidirectional energy consumption time series propagation fusion representation vector, represents the zero norm of a vector, represents the length of the bidirectional energy consumption time series propagation fusion representation vector, represents the bidirectional energy consumption time series propagation fusion isolated representation value, It is the point multiplication by position. is added by position, A power function whose exponent is the difference between the bidirectional energy consumption time series propagation fusion isolated representation value minus one represents the calculation of the feature value of each position in the feature vector, Represents the optimized bidirectional energy consumption time series propagation fusion representation vector.
[0095] Here, for the high-dimensional feature manifold of the bidirectional energy consumption time series propagation fusion representation vector, the eigenvalue of the feature set is used as the vector field representation of the aggregation dimension, and the superposition value of the vector field of the bidirectional energy consumption time series propagation fusion representation vector at the isolated zero position is used as the order information to fix the local position of the eigenvalue of its feature set, and add a bias for the reversibility of the feature regression distribution field of the bidirectional energy consumption time series propagation fusion representation vector as a reward to achieve the mapping target tracking of the regression distribution of the bidirectional energy consumption time series propagation fusion representation vector for the eigenvalue position, so that the feature set of the bidirectional energy consumption time series propagation fusion representation vector perceives the mapping migration to the aggregate distribution, thereby improving the accuracy of the energy consumption data correction decoding value obtained by the data corrector based on the decoder of the bidirectional energy consumption time series propagation fusion representation vector by improving the comprehensibility of the feature vector associated fusion regression of the bidirectional energy consumption time series propagation fusion representation vector. In this way, the energy consumption abnormal data can be effectively corrected based on the forward and backward contextual time series semantics of the energy consumption data, the stability of the energy consumption system can be restored, and a more accurate data basis can be provided for subsequent energy management decisions.
[0096] Based on the above embodiments, see Figure 4 As shown, it is a structural schematic diagram of an energy consumption data correction system 800 based on a machine learning algorithm in an embodiment of the present application. The energy consumption data correction system 800 based on a machine learning algorithm includes: a data acquisition module 810, used to acquire original data; an anchoring module 820, used to anchor the energy consumption data to be corrected in the original data; a data segmentation module 830, used to perform data segmentation on the original data based on the energy consumption data to be corrected to obtain previous energy consumption history data and subsequent energy consumption extension data; a local time series feature extraction module 840, used to perform local time series feature extraction on the previous energy consumption history data and the subsequent energy consumption extension data respectively to obtain a sequence of local time series feature vectors of previous energy consumption and a sequence of local time series feature vectors of subsequent energy consumption; a forward time series propagation module 850, Used to input the sequence of previous energy consumption local time series feature vectors into the forward LSTM model to obtain the previous energy consumption time series propagation representation vector; the backward time series propagation module 860 is used to input the sequence of subsequent energy consumption local time series feature vectors into the backward LSTM model to obtain the backward energy consumption time series propagation representation vector; the fusion module 870 is used to input the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector into the feature value granularity association mask distinguishable attention fusion module to obtain the bidirectional energy consumption time series propagation fusion representation vector; the energy consumption data correction module 880 is used to determine the energy consumption data correction decoding value based on the bidirectional energy consumption time series propagation fusion representation vector.
[0097] In one example, in the above-mentioned energy consumption data correction system 800 based on machine learning algorithm, the local time series feature extraction module 840 is used to: input the previous energy consumption history data and the subsequent energy consumption extension data into the energy consumption local time series feature extractor based on the 1D-CNN model respectively to obtain the sequence of the previous energy consumption local time series feature vectors and the sequence of the subsequent energy consumption local time series feature vectors.
[0098] In one example, in the above-mentioned energy consumption data correction system 800 based on machine learning algorithm, the fusion module 870 is used to: calculate the global eigenvalue association matrix between the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector to obtain the previous-backward energy consumption global eigenvalue association matrix; input the previous-backward energy consumption global eigenvalue association matrix into a learnable gating function to obtain the previous-backward energy consumption correlation distinguishable weight matrix; use the previous-backward energy consumption correlation distinguishable weight matrix as a fusion mask matrix, and respectively calculate the matrix product between the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector and the previous-backward energy consumption correlation distinguishable weight matrix to obtain the differentiated enhanced previous energy consumption time series propagation representation vector and the differentiated enhanced backward energy consumption time series propagation representation vector; fuse the differentiated enhanced previous energy consumption time series propagation representation vector and the differentiated enhanced backward energy consumption time series propagation representation vector to obtain the bidirectional energy consumption time series propagation fusion representation vector.
[0099] Here, those skilled in the art can understand that the specific functions and operations of each module in the energy consumption data correction system 800 based on the machine learning algorithm have been referred to above. Figure 2 to Figure 3 The energy consumption data correction method based on machine learning algorithm has been introduced in detail, and therefore, its repeated description will be omitted.
[0100] Figure 5 FIG. 1 is an application scenario diagram of the energy consumption data correction method based on the machine learning algorithm according to the embodiment of the present application. Figure 5 As shown, in this application scenario, first, the original data (for example, Figure 5 D), and then, the raw data is input to a server (for example, Figure 5 In S) as shown in , the server can use the energy consumption data correction algorithm based on the machine learning algorithm to process the original data to obtain the energy consumption data correction decoding value.
[0101] Based on the above embodiments, another exemplary embodiment of an electronic device is also provided in the embodiments of the present application. In some possible implementations, the electronic device in the embodiments of the present application may include a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor can implement the steps of the energy consumption data correction method based on the machine learning algorithm in the above embodiments when executing the program.
[0102] For example, in the case of electronic equipment Figure 1 Taking the server 100 in the example as an example, the processor in the electronic device is the processor 110 in the server 100, and the memory in the electronic device is the memory 120 in the server 100.
[0103] The embodiment of the present application also provides a computer-readable storage medium, on which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, the energy consumption data correction method based on a machine learning algorithm according to the embodiment of the present application described with reference to the above figures can be executed. The computer-readable storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. The volatile memory may, for example, include a random access memory (RAM) and / or a cache memory (cache), etc. The non-volatile memory may, for example, include a read-only memory (ROM), a hard disk, a flash memory, etc.
[0104] The embodiment of the present application also provides a computer program product or a computer program, which includes computer executable instructions, and the computer executable instructions are stored in a computer readable storage medium. The processor of the computer device reads the computer executable instructions from the computer readable storage medium, and the processor executes the computer executable instructions, so that the computer device executes the energy consumption data correction method based on the machine learning algorithm according to the embodiment of the present application.
[0105] Those skilled in the art will appreciate that the contents disclosed in this application may be subject to various variations and improvements. For example, the various devices or components described above may be implemented by hardware, or by software, firmware, or a combination of some or all of the three.
[0106] In addition, although the present application makes various references to certain units in the system according to embodiments of the present application, any number of different units can be used and run on the client and / or server. The units are only illustrative, and different aspects of the system and method can use different units.
[0107] Those skilled in the art will appreciate that all or part of the steps in the above method can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, such as a read-only memory, a disk or an optical disk. Optionally, all or part of the steps in the above embodiment can also be implemented using one or more integrated circuits. Accordingly, each module / unit in the above embodiment can be implemented in the form of hardware or in the form of software function modules. The present application is not limited to any particular form of combination of hardware and software.
[0108] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those of ordinary skill in the art to which this application belongs. It should also be understood that terms such as those defined in common dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an idealized or extremely formal sense, unless explicitly defined as such herein.
[0109] The above is an explanation of the present application and should not be considered as limiting thereof. Although several exemplary embodiments of the present application have been described, those skilled in the art will readily appreciate that many modifications may be made to the exemplary embodiments without departing from the novel teachings and advantages of the present application.
Claims
1. A method for correcting energy consumption data based on machine learning algorithm, characterized in that: include: Get the original data; Anchoring the energy consumption data to be corrected in the original data; Based on the energy consumption data to be corrected, data segmentation is performed on the original data to obtain previous energy consumption history data and subsequent energy consumption extension data; Performing local time series feature extraction on the previous energy consumption history data and the subsequent energy consumption extension data respectively to obtain a sequence of previous energy consumption local time series feature vectors and a sequence of subsequent energy consumption local time series feature vectors; Inputting the sequence of the previous energy consumption local time series feature vectors into the forward LSTM model to obtain the previous energy consumption time series propagation representation vector; Inputting the sequence of the local time series feature vectors of the subsequent energy consumption into a backward LSTM model to obtain a backward energy consumption time series propagation representation vector; Inputting the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector into a feature value granularity association mask distinguishable attention fusion module to obtain a bidirectional energy consumption time series propagation fusion representation vector; Determine a corrected decoding value of energy consumption data based on the bidirectional energy consumption time series propagation fusion representation vector; The previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector are input into the feature value granularity association mask distinguishable attention fusion module to obtain a bidirectional energy consumption time series propagation fusion representation vector, including: Calculating a global eigenvalue association matrix between the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector to obtain a previous-backward energy consumption global eigenvalue association matrix; Inputting the global eigenvalue association matrix of the previous-backward energy consumption into a learnable gating function to obtain a distinguishable weight matrix of the previous-backward energy consumption correlation; Using the previous-backward energy consumption correlation distinguishable weight matrix as a fusion mask matrix, respectively calculating the matrix products between the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector and the previous-backward energy consumption correlation distinguishable weight matrix to obtain a distinguished and enhanced previous energy consumption time series propagation representation vector and a distinguished and enhanced backward energy consumption time series propagation representation vector; Fusing the differentiated enhanced previous energy consumption time series propagation representation vector and the differentiated enhanced backward energy consumption time series propagation representation vector to obtain the bidirectional energy consumption time series propagation fusion representation vector; The step of calculating the global eigenvalue association matrix between the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector to obtain the previous-backward energy consumption global eigenvalue association matrix includes: Calculating the vector multiplication between the transposed vector of the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector to obtain the previous-backward energy consumption global eigenvalue association matrix; The preceding-backward energy consumption global eigenvalue association matrix is input into a learnable gating function to obtain a preceding-backward energy consumption correlation distinguishable weight matrix, including: Taking the negative number of each position eigenvalue in the global eigenvalue association matrix of the previous-backward energy consumption as the exponential power, calculating the value of the natural exponential function with the natural constant e as the base to obtain the global eigenvalue association class support matrix of the previous-backward energy consumption; Calculate the reciprocal of the sum of the eigenvalues of each position in the previous-backward energy consumption global eigenvalue association class support matrix and the constant 1 to obtain a previous-backward energy consumption global eigenvalue association weight matrix; Calculate the position point multiplication between the previous-backward energy consumption global eigenvalue association weight matrix and the previous-backward energy consumption global eigenvalue association matrix to obtain the previous-backward energy consumption correlation distinguishable weight matrix; The step of fusing the differentiated enhanced previous energy consumption time series propagation representation vector and the differentiated enhanced backward energy consumption time series propagation representation vector to obtain the bidirectional energy consumption time series propagation fusion representation vector includes: The differentiated enhanced previous energy consumption time series propagation representation vector and the differentiated enhanced backward energy consumption time series propagation representation vector are cascaded to obtain the bidirectional energy consumption time series propagation fusion representation vector.
2. The energy consumption data correction method based on machine learning algorithm according to claim 1 is characterized in that: The local time series feature extraction is performed on the previous energy consumption history data and the subsequent energy consumption extension data respectively to obtain a sequence of the previous energy consumption local time series feature vectors and a sequence of the subsequent energy consumption local time series feature vectors, including: The previous energy consumption history data and the subsequent energy consumption extension data are respectively input into the energy consumption local time series feature extractor based on the 1D-CNN model to obtain a sequence of the previous energy consumption local time series feature vectors and a sequence of the subsequent energy consumption local time series feature vectors.
3. The energy consumption data correction method based on machine learning algorithm according to claim 2 is characterized in that: Based on the bidirectional energy consumption time series propagation fusion representation vector, determining the energy consumption data correction decoding value includes: The bidirectional energy consumption time series propagation fusion representation vector is input into a decoder-based data corrector to obtain the energy consumption data correction decoding value.
4. An energy consumption data correction system based on machine learning algorithm, characterized in that: include: A data acquisition module, used to acquire raw data; An anchoring module, used for anchoring the energy consumption data to be corrected in the original data; A data segmentation module, configured to segment the original data based on the energy consumption data to be corrected to obtain previous energy consumption history data and subsequent energy consumption extension data; A local time series feature extraction module, used to extract local time series features from the previous energy consumption history data and the subsequent energy consumption extension data to obtain a sequence of local time series feature vectors of previous energy consumption and a sequence of local time series feature vectors of subsequent energy consumption; A forward time series propagation module, used for inputting the sequence of the previous energy consumption local time series feature vectors into a forward LSTM model to obtain a previous energy consumption time series propagation representation vector; A backward time series propagation module, used for inputting the sequence of the backward energy consumption local time series feature vectors into a backward LSTM model to obtain a backward energy consumption time series propagation representation vector; A fusion module, configured to input the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector into a feature value granularity association mask distinguishable attention fusion module to obtain a bidirectional energy consumption time series propagation fusion representation vector; An energy consumption data correction module, used to determine the energy consumption data correction decoding value based on the bidirectional energy consumption time series propagation fusion representation vector; Wherein, the fusion module is used to: Calculating a global eigenvalue association matrix between the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector to obtain a previous-backward energy consumption global eigenvalue association matrix; Inputting the global eigenvalue association matrix of the previous-backward energy consumption into a learnable gating function to obtain a distinguishable weight matrix of the previous-backward energy consumption correlation; Using the previous-backward energy consumption correlation distinguishable weight matrix as a fusion mask matrix, respectively calculating the matrix products between the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector and the previous-backward energy consumption correlation distinguishable weight matrix to obtain a distinguished and enhanced previous energy consumption time series propagation representation vector and a distinguished and enhanced backward energy consumption time series propagation representation vector; Fusing the differentiated enhanced previous energy consumption time series propagation representation vector and the differentiated enhanced backward energy consumption time series propagation representation vector to obtain the bidirectional energy consumption time series propagation fusion representation vector; The step of calculating the global eigenvalue association matrix between the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector to obtain the previous-backward energy consumption global eigenvalue association matrix includes: Calculating the vector multiplication between the transposed vector of the previous energy consumption time series propagation representation vector and the backward energy consumption time series propagation representation vector to obtain the previous-backward energy consumption global eigenvalue association matrix; The preceding-backward energy consumption global eigenvalue association matrix is input into a learnable gating function to obtain a preceding-backward energy consumption correlation distinguishable weight matrix, including: Taking the negative number of each position eigenvalue in the global eigenvalue association matrix of the previous-backward energy consumption as the exponential power, calculating the value of the natural exponential function with the natural constant e as the base to obtain the global eigenvalue association class support matrix of the previous-backward energy consumption; Calculate the reciprocal of the sum of the eigenvalues of each position in the previous-backward energy consumption global eigenvalue association class support matrix and the constant 1 to obtain a previous-backward energy consumption global eigenvalue association weight matrix; Calculate the position point multiplication between the previous-backward energy consumption global eigenvalue association weight matrix and the previous-backward energy consumption global eigenvalue association matrix to obtain the previous-backward energy consumption correlation distinguishable weight matrix; The step of fusing the differentiated enhanced previous energy consumption time series propagation representation vector and the differentiated enhanced backward energy consumption time series propagation representation vector to obtain the bidirectional energy consumption time series propagation fusion representation vector includes: The differentiated enhanced previous energy consumption time series propagation representation vector and the differentiated enhanced backward energy consumption time series propagation representation vector are cascaded to obtain the bidirectional energy consumption time series propagation fusion representation vector.
5. The energy consumption data correction system based on machine learning algorithm according to claim 4 is characterized in that: The local temporal feature extraction module is used to: The previous energy consumption history data and the subsequent energy consumption extension data are respectively input into the energy consumption local time series feature extractor based on the 1D-CNN model to obtain a sequence of the previous energy consumption local time series feature vectors and a sequence of the subsequent energy consumption local time series feature vectors.
