CVT error prediction method and device, electronic equipment and storage medium

Abstract

The invention relates to a CVT error prediction method and device, electronic equipment and a storage medium, and belongs to the technical field of data processing.The CVT error prediction method comprises the steps that historical error data and historical environment parameter data of a target CVT are preprocessed; taking the preprocessed historical error data and historical environmental parameter data as the input of an error prediction model, and determining the real-time error prediction result of the target CVT based on the output of the error prediction model; wherein the error prediction model is formed by cyclically connecting a plurality of multivariate learning networks, and each multivariate learning network comprises a feature selection unit used for extracting effective features from the preprocessed historical environmental parameter data, a feature memory unit used for determining a memory state value of the preprocessed historical environmental parameter data, and a feature prediction unit used for predicting the memory state value of the preprocessed historical environmental parameter data. And a feature fusion unit. According to the invention, the accuracy of the error prediction result is improved while the CVT real-time error prediction is realized.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a CVT error prediction method, device, electronic equipment and storage medium. Background Art

[0002] In the field of electric energy metering, the accurate and reliable metering performance of the transformer is directly related to economic benefits and social effects, and has an impact on the healthy and stable operation of the power grid that cannot be ignored. As an electric energy metering device, the capacitor voltage transformer (CVT) can not only complete the voltage conversion by dividing the voltage with a series capacitor and using the voltage reduction and isolation functions of the electromagnetic transformer, but also effectively couple the carrier frequency to the transmission line, thereby realizing long-distance communication, high-frequency line protection selectivity, remote control and other functions. Compared with traditional electromagnetic voltage transformers, CVT has significant advantages in economy and safety with its excellent impact insulation strength, simple manufacturing process, and light size and weight.

[0003] However, in actual operation, due to the limitations of the acquisition principle and the harsh environment, the measurement error of CVT may exceed the established range within the working life, resulting in deviation exceeding the limit. The traditional CVT calibration method cannot detect whether it is abnormal in time during the calibration cycle, which may easily cause potential problems for power trade disputes and power system safety. Therefore, how to predict CVT errors online to warn of possible risks of CVT in advance has become a technical problem that needs to be solved urgently. Summary of the invention

[0004] In view of this, it is necessary to provide a CVT error prediction method, device, electronic device and storage medium to solve the problem that traditional CVT calibration methods are difficult to predict CVT errors in a timely manner.

[0005] In order to solve the above problems, the present invention provides a CVT error prediction method, comprising: Preprocessing the historical error data and historical environmental parameter data of the target CVT; Using the preprocessed historical error data and historical environmental parameter data as inputs of an error prediction model, and determining a real-time error prediction result of the target CVT based on an output of the error prediction model; Among them, the error prediction model is composed of multiple multivariate learning networks connected in a loop, and each multivariate learning network includes: a feature selection unit for extracting effective features from preprocessed historical environmental parameter data, a feature memory unit for determining the memory state value of the preprocessed historical environmental parameter data, and a feature fusion unit.

[0006] In a possible implementation, extracting effective features from the preprocessed historical environmental parameter data includes: Based on the environmental parameter vector at the target moment and the error hidden state value at the previous moment of the target moment, the effective feature vector at the target moment is determined.

[0007] In a possible implementation, the effective feature vector at the target time is determined based on the following formula:

[0008] in, represents the effective feature vector at the target time, represents the normalized exponential function, represents the hyperbolic tangent function, represents the environmental parameter vector at the target time, Represents the error hidden state value at the previous moment of the target moment, It means adding the vector to each column of the matrix for addition calculation. , , is the weight matrix.

[0009] In a possible implementation, determining the memory state value of the preprocessed historical environmental parameter data includes: The memory state value at the target moment is determined based on the effective feature vector at the target moment, and the error hidden state value and the memory state value at the previous moment of the target moment.

[0010] In a possible implementation, the memory state value at the target moment is determined based on the following formula:

[0011]

[0012]

[0013]

[0014] in, Indicates the memory state value at the target moment, represents the retention rate of the effective feature vector at the target moment, represents the effective feature vector at the target time, Indicates the retention rate of the memory state value at the previous moment of the target moment, Indicates the memory state value of the previous moment of the target moment, represents the memory retention function, Indicates the total amount of input information at the target time and the time before the target time, represents the hyperbolic tangent function, represents the Sigmoid function, , , , is the weight matrix, , , , is the bias term, represents vector modulo operation, Represents a vector concatenation operation.

[0015] In a possible implementation, the feature fusion unit is used to determine the error concealment state value at the target moment based on the error vector and the memory state value at the target moment, and the error concealment state value at a moment before the target moment.

[0016] In a possible implementation, the error concealment state value at the target time is determined based on the following formula:

[0017]

[0018]

[0019]

[0020] in, represents the error hidden state value at the target time, Represents the intermediate state of the error hidden state value at the previous moment of the target moment, represents the intermediate state of the error vector at the target time, represents the hyperbolic tangent function, , , is the weight matrix, , , represents the bias term, is the fusion feature allocation coefficient, represents the fusion state value, represents the Sigmoid function, represents the error vector at the target time, Indicates the memory state value at the target moment, Indicates the error hidden state value at the previous moment of the target moment.

[0021] The present invention also provides a CVT error prediction device, comprising: A preprocessing module, used for preprocessing historical error data and historical environmental parameter data of a target CVT; A prediction module, used to use the preprocessed historical error data and historical environmental parameter data as inputs of an error prediction model, and determine a real-time error prediction result of the target CVT based on an output of the error prediction model; Among them, the error prediction model is composed of multiple multivariate learning networks connected in a loop, and each multivariate learning network includes: a feature selection unit for extracting effective features from preprocessed historical environmental parameter data, a feature memory unit for determining the memory state value of the preprocessed historical environmental parameter data, and a feature fusion unit.

[0022] The present invention also provides an electronic device, including a memory and a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the CVT error prediction method as described above is implemented.

[0023] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the CVT error prediction method as described above when executed by a processor.

[0024] The beneficial effects of the present invention are as follows: the CVT error prediction method, device, electronic device and storage medium provided by the present invention extract and fuse effective features in historical error data and historical environmental parameter data of the target CVT through an error prediction model, and the error prediction model is formed by cyclically connecting multiple multivariate learning networks, so that a single model can be used to extract cyclic features of multivariate time series variables, thereby realizing real-time error prediction of CVT and improving the accuracy of error prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 A schematic flow chart of an embodiment of a CVT error prediction method provided by the present invention; Figure 2 A schematic diagram of a flow chart of an embodiment of a CVT error prediction process provided by the present invention; Figure 3 A schematic diagram of the structure of an embodiment of a multi-element cyclic network provided by the present invention; Figure 4 A schematic diagram of the structure of an embodiment of a multivariate learning network provided by the present invention; Figure 5 A schematic structural diagram of an embodiment of a CVT error prediction device provided by the present invention; Figure 6 A schematic structural diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0026] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0027] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0028] In the description of the present invention, reference to an "embodiment" means that a particular feature, structure, or characteristic described in conjunction with the embodiment may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive of other embodiments. It is explicitly and implicitly understood by those skilled in the art that the described embodiments may be combined with other embodiments.

[0029] In the field of electric energy metering, the accurate and reliable metering performance of the transformer is directly related to economic benefits and social effects, and has an impact on the healthy and stable operation of the power grid that cannot be ignored. As an electric energy metering device, CVT can not only complete the voltage conversion by dividing the voltage with a series capacitor and using the voltage reduction and isolation functions of the electromagnetic transformer, but also effectively couple the carrier frequency to the transmission line, thereby realizing long-distance communication, high-frequency line protection selectivity, remote control and other functions. Compared with traditional electromagnetic voltage transformers, CVT has significant advantages in economy and safety with its excellent impact insulation strength, simple manufacturing process, and light size and weight.

[0030] However, in actual operation, due to the limitations of the acquisition principle and the harsh environment, the measurement error of CVT may exceed the established range within the working life, and the deviation may exceed the limit. The traditional CVT calibration method cannot detect whether it is abnormal in time during the calibration cycle, which may easily cause potential problems for power trade disputes and power system safety.

[0031] The specific embodiments are described in detail below: A specific embodiment of the present invention discloses a CVT error prediction method, combining Figure 1 Come and see, Figure 1 The flowchart of an embodiment of the CVT error prediction method provided by the present invention includes step S101 and step S102, wherein: In step S101, historical error data and historical environmental parameter data of the target CVT are preprocessed; In step S102, the preprocessed historical error data and historical environmental parameter data are used as inputs of an error prediction model, and a real-time error prediction result of the target CVT is determined based on an output of the error prediction model; Among them, the error prediction model is composed of multiple multivariate learning networks connected in a loop, and each multivariate learning network includes: a feature selection unit for extracting effective features from preprocessed historical environmental parameter data, a feature memory unit for determining the memory state value of the preprocessed historical environmental parameter data, and a feature fusion unit.

[0032] During implementation, the historical error data and historical environmental parameter data of the target CVT can be obtained first, that is, the error value data and voltage, secondary load, frequency, and temperature data of the target CVT during historical operation. Then, the historical error data and historical environmental parameter data of the target CVT are preprocessed. For example, the historical error data and historical environmental parameter data of the target CVT are normalized to prevent the dimensions and gaps between different data from being too large, and to retain the relationships existing in the original data.

[0033] Then, the preprocessed historical error data and historical environmental parameter data can be used as inputs of the error prediction model, and the real-time error prediction results of the target CVT can be determined according to the output of the error prediction model. For example, the output results of the error prediction model can be input into a fully connected network for weight distribution and integration to obtain the real-time error prediction results of the target CVT.

[0034] The error prediction model can be composed of multiple multivariate learning networks connected in a loop, and each multivariate learning network can include a feature selection unit, a feature memory unit, and a feature fusion unit. The feature selection unit is used to extract effective features from the preprocessed historical environmental parameter data. The feature memory unit is used to determine the memory state value of the preprocessed historical environmental parameter data. The feature fusion unit is used to fuse the effective features carried by the historical error data with the effective features carried by the historical environmental parameter data.

[0035] Compared with the prior art, the CVT error prediction method provided in this embodiment extracts and fuses effective features from historical error data and historical environmental parameter data of the target CVT through an error prediction model. The error prediction model is formed by cyclically connecting multiple multivariate learning networks, so that a single model can be used to extract cyclic features of multivariate time series variables, thereby realizing real-time error prediction of CVT and improving the accuracy of error prediction results.

[0036] Exemplarily, extracting effective features from the preprocessed historical environmental parameter data includes: Based on the environmental parameter vector at the target moment and the error hidden state value at the previous moment of the target moment, the effective feature vector at the target moment is determined.

[0037] Specifically, when extracting effective features from the preprocessed historical environmental parameter data, the effective feature vector at the target moment may be determined based on the environmental parameter vector at the target moment and the error hidden state value at the previous moment of the target moment.

[0038] Exemplarily, the effective feature vector at the target time is determined based on the following formula:

[0039] in, represents the effective feature vector at the target time, represents the normalized exponential function, represents the hyperbolic tangent function, represents the environmental parameter vector at the target time, Represents the error hidden state value at the previous moment of the target moment, It means adding the vector to each column of the matrix for addition calculation. , , is the weight matrix.

[0040] Specifically, the effective feature vector at the target moment can be determined according to the above formula.

[0041] Exemplarily, determining the memory state value of the preprocessed historical environmental parameter data includes: The memory state value at the target moment is determined based on the effective feature vector at the target moment, and the error hidden state value and the memory state value at the previous moment of the target moment.

[0042] Specifically, when determining the memory state value of the preprocessed historical environmental parameter data, the memory state value at the target moment can be determined based on the effective feature vector at the target moment, and the error hidden state value and memory state value at the previous moment of the target moment.

[0043] Exemplarily, the memory state value at the target moment is determined based on the following formula:

[0044]

[0045]

[0046]

[0047] in, Indicates the memory state value at the target moment, represents the retention rate of the effective feature vector at the target moment, represents the effective feature vector at the target time, Indicates the retention rate of the memory state value at the previous moment of the target moment, Indicates the memory state value of the previous moment of the target moment, represents the memory retention function, Indicates the total amount of input information at the target time and the time before the target time, represents the hyperbolic tangent function, represents the Sigmoid function, , , , is the weight matrix, , , , is the bias term, represents vector modulo operation, Represents a vector concatenation operation.

[0048] Specifically, the memory state value at the target moment can be determined according to the above formula.

[0049] Exemplarily, the feature fusion unit is used to determine the error concealment state value at the target moment based on the error vector and the memory state value at the target moment, and the error concealment state value at a moment before the target moment.

[0050] Specifically, the feature fusion unit in the error prediction model can determine the error hidden state value at the target moment according to the error vector and the memory state value at the target moment, and the error hidden state value at the previous moment of the target moment.

[0051] Exemplarily, the error concealment state value at the target time is determined based on the following formula:

[0052]

[0053]

[0054]

[0055] in, represents the error hidden state value at the target time, Represents the intermediate state of the error hidden state value at the previous moment of the target moment, represents the intermediate state of the error vector at the target time, represents the hyperbolic tangent function, , , is the weight matrix, , , represents the bias term, is the fusion feature allocation coefficient, represents the fusion state value, represents the Sigmoid function, represents the error vector at the target time, Indicates the memory state value at the target moment, Indicates the error hidden state value at the previous moment of the target moment.

[0056] Specifically, the error concealment state value at the target moment can be determined according to the above formula.

[0057] The following is a specific application scenario to better illustrate the technical solution of the present invention: Combination Figure 2 Come and see, Figure 2 This is a flow chart of an embodiment of a CVT error prediction process provided by the present invention, and the process specifically includes the following steps: 1. Obtain the error data and environmental parameter data of the CVT during historical operation in the substation, including the error value data of the voltage transformer during historical operation and the voltage, secondary load, frequency, and temperature data to form a time series data set, and use the vector normalization method to standardize the time series data.

[0058] 2. Construct a multivariate recurrent network, which is composed of multiple multivariate learning units connected in a loop. Each multivariate learning unit consists of three parts: a feature selection unit, a feature memory unit, and a feature fusion unit. The feature selection unit is used to select useful feature information in the environmental parameters and generate a significant feature vector. The feature memory unit is used to record and save important information carried by the environmental parameters over time, and discard some useless information. The feature fusion unit is used to fuse the valid feature information carried by the error value data with the valid feature information provided by the environmental parameters. The hidden state results output by the multivariate recurrent network are then input into the fully connected network for weight distribution and integration, and the final error prediction results are output. Combined with Figure 3 Come and see, Figure 3 A schematic diagram of the structure of an embodiment of a multi-element cyclic network provided by the present invention.

[0059] Existing recurrent neural networks and their variant networks can only realize the cyclic extraction of single time series data features. For time series prediction tasks that process multiple variables, they need to be combined with other networks to complete. The multivariate recurrent network provided by the present invention can realize parallel feature extraction of multiple variables. The multivariate learning units composed of feature selection units, feature memory units and feature fusion units constructed in the network are cyclically connected, so that the time series feature extraction results are more effective than those of recurrent neural networks that can only input single variables. Combined with Figure 4 Come and see, Figure 4 A schematic diagram of the structure of an embodiment of a multivariate learning network provided by the present invention.

[0060] In each multivariate learning unit, the data to be input each time include the current environment parameter matrix and error vector. Taking time t as an example, is the matrix composed of environmental parameter data at time t, is the error vector at time t. The multivariate learning unit combines the input environment parameter matrix and the error value hidden state of the previous moment through the feature selection unit , select the useful feature information in the environmental parameters and generate a significant feature vector Used for subsequent prediction process. The calculation formula of this process can be expressed as follows:

[0061] Since the feature information contained in the past information is also very important in the time series prediction task, in order to effectively obtain the relevant information of the past information and the current input information, a feature memory unit is designed. Combination and Together, we calculate the important information contained in the past information and retain it for use. The calculation process is as follows:

[0062] Since the feature information contained in the past information is also very important in the time series prediction task, in order to effectively obtain the relevant information of the past information and the current input information, a feature memory unit is designed. Combination and Together, we calculate the important information contained in the past information and retain it for use. The calculation process is as follows:

[0063] Then, the significant feature vector generated by calculating the environmental parameter data and the memory state value of the environmental parameter information at time t-1 The importance of the information in the overall information determines how much information the two should forget and how much information they should retain.

[0064]

[0065]

[0066] The above calculations obtained and , the output result of the feature memory unit can be calculated , the result represents the memory state value of the environmental parameter information at time t, which will be used by the subsequent feature fusion unit to calculate the memory state value at the next moment.

[0067]

[0068] In the feature fusion unit, the error vector at time t is calculated by fusion And the corresponding hidden state at time t-1 and the memory state value of the environmental parameter information at time t . Get the hidden state value at time t The overall calculation process can be expressed as follows:

[0069]

[0070]

[0071]

[0072] Finally, all hidden state results calculated by the multivariate recurrent network are sent to the fully connected network for feature integration to output the final error prediction value.

[0073] 3. The constructed multivariate cyclic network error prediction model is trained with the historical data obtained above, and then the error prediction result of the voltage transformer to be tested is obtained by using the trained model combined with the data of the voltage transformer to be tested.

[0074] The present invention collects historical operating data of three transformers, extracts three groups of different test data samples, and performs multi-sample algorithm performance tests under the extracted different data samples. The actual implementation effect of the method provided by the present invention is verified by comparing the current mainstream recurrent neural network and its variant forms including: RNN, LSTM, GRU, Bi-LSTM, Bi-GRU, and the following experimental result comparison table is obtained: Table 1: Comparison results of various algorithms under three samples

[0075] It can be seen that the RMSE and MAE of the method of the present invention are lower than those of the same type of algorithms under three different data samples. This indicates that the method proposed in the present invention has the highest prediction accuracy, and the results of error prediction using the method of the present invention are more reliable.

[0076] The present invention proposes a multivariate recurrent network model, which mainly consists of a multivariate learning unit through a feature selection unit, a feature memory unit and a feature fusion unit, and then multiple multivariate learning units are cyclically connected to form a multivariate recurrent network, so that a single model can be used to extract cyclic features of multivariate time series variables, making the prediction results more accurate. The feature memory unit proposed in the present invention can effectively filter and retain environmental parameter feature information, and the feature fusion unit can combine multiple states for efficient fusion.

[0077] The embodiment of the present invention also provides a CVT error prediction device, combined with Figure 5 Come and see, Figure 5 This is a schematic diagram of the structure of an embodiment of a CVT error prediction device provided by the present invention. The CVT error prediction device 500 includes: A preprocessing module 501 is used to preprocess the historical error data and historical environmental parameter data of the target CVT; A prediction module 502, configured to use the preprocessed historical error data and historical environmental parameter data as inputs of an error prediction model, and determine a real-time error prediction result of the target CVT based on an output of the error prediction model; Among them, the error prediction model is composed of multiple multivariate learning networks connected in a loop, and each multivariate learning network includes: a feature selection unit for extracting effective features from preprocessed historical environmental parameter data, a feature memory unit for determining the memory state value of the preprocessed historical environmental parameter data, and a feature fusion unit.

[0078] The specific implementation methods of each module of the CVT error prediction device can refer to the description of the above-mentioned CVT error prediction method, and have similar beneficial effects, which will not be repeated here.

[0079] The embodiment of the present invention further provides an electronic device, Figure 6 Come and see, Figure 6 This is a structural diagram of an embodiment of an electronic device provided by the present invention. The electronic device 600 includes a processor 601, a memory 602, and a computer program stored in the memory 602 and executable on the processor 601. When the processor 601 executes the program, the CVT error prediction method as described above is implemented.

[0080] As a preferred embodiment, the electronic device 600 further includes a display 603 for displaying the CVT error prediction method as executed by the processor 601 .

[0081] Exemplarily, the computer program may be divided into one or more modules / units, one or more modules / units are stored in the memory 602, and executed by the processor 601 to complete the present invention. One or more modules / units may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program in the electronic device 600. For example, the computer program may be divided into the preprocessing module 501 and the prediction module 502 in the above embodiment, and the specific functions of each module are as described above, which will not be repeated here.

[0082] The electronic device 600 may be a desktop computer, a notebook, a PDA or a smart phone with an adjustable camera module.

[0083] Among them, the processor 601 may be an integrated circuit chip with signal processing capabilities. The above-mentioned processor 601 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The disclosed methods, steps and logic block diagrams in the embodiments of the present invention can be implemented or executed. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0084] The memory 602 may be, but is not limited to, a random access memory (RAM), a read only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable read-only memory (EEPROM), etc. The memory 602 is used to store programs, and the processor 601 executes the program after receiving an execution instruction. The process definition method disclosed in any of the above embodiments of the present invention can be applied to the processor 601, or implemented by the processor 601.

[0085] The display 603 may be an LCD display screen or an LED display screen, for example, a display screen on a mobile phone.

[0086] Understandably, Figure 6 The structure shown is only a schematic diagram of a structure of the electronic device 600. The electronic device 600 may also include Figure 6 More or fewer components as shown. Figure 6 Each component shown in the figure can be implemented by hardware, software or a combination thereof.

[0087] The electronic device provided according to the above embodiment of the present invention can be implemented with reference to the specific description of the CVT error prediction method described above according to the present invention, and has similar beneficial effects as the CVT error prediction method described above, which will not be repeated here.

[0088] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the CVT error prediction method as described above is implemented.

[0089] Generally speaking, the computer instructions for implementing the method of the present invention can be carried by any combination of one or more computer-readable storage media. Non-transitory computer-readable storage media can include any computer-readable media except for the signal itself that is temporarily propagating.

[0090] Computer-readable storage media may be, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more conductors, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium may be any tangible medium that contains or stores a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0091] Computer program codes for performing the operations of the present invention may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages, in particular, Python suitable for neural network computing and platform frameworks based on TensorFlow, PyTorch, etc. may be used. The program code may be executed entirely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0092] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0093] The present invention discloses a CVT error prediction method, device, electronic device and storage medium. The error prediction model is used to extract and fuse the effective features in the historical error data and historical environmental parameter data of the target CVT. The error prediction model is formed by cyclically connecting multiple multivariate learning networks, so that a single model can be used to extract the cyclic features of multivariate time series variables. While realizing real-time error prediction of CVT, the accuracy of the error prediction result is improved.

[0094] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A CVT error prediction method, characterized in that: include: Preprocessing the historical error data and historical environmental parameter data of the target CVT; Using the preprocessed historical error data and historical environmental parameter data as inputs of an error prediction model, and determining a real-time error prediction result of the target CVT based on an output of the error prediction model; Among them, the error prediction model is composed of multiple multivariate learning networks connected in a loop, and each multivariate learning network includes: a feature selection unit for extracting effective features from preprocessed historical environmental parameter data, a feature memory unit for determining the memory state value of the preprocessed historical environmental parameter data, and a feature fusion unit.

2. The CVT error prediction method according to claim 1, characterized in that: The extracting effective features from the pre-processed historical environmental parameter data includes: Based on the environmental parameter vector at the target moment and the error hidden state value at the previous moment of the target moment, the effective feature vector at the target moment is determined.

3. The CVT error prediction method according to claim 2, characterized in that: The effective feature vector at the target time is determined based on the following formula: in, represents the effective feature vector at the target time, represents the normalized exponential function, represents the hyperbolic tangent function, represents the environmental parameter vector at the target time, Represents the error hidden state value at the previous moment of the target moment, It means adding the vector to each column of the matrix for addition calculation. , , is the weight matrix.

4. The CVT error prediction method according to claim 1, characterized in that: The step of determining the memory state value of the preprocessed historical environmental parameter data includes: The memory state value at the target moment is determined based on the effective feature vector at the target moment, and the error hidden state value and the memory state value at the previous moment of the target moment.

5. The CVT error prediction method according to claim 4, characterized in that: The memory state value at the target moment is determined based on the following formula: in, Indicates the memory state value at the target moment, represents the retention rate of the effective feature vector at the target moment, represents the effective feature vector at the target time, Indicates the retention rate of the memory state value at the previous moment of the target moment, Indicates the memory state value of the previous moment of the target moment, represents the memory retention function, Indicates the total amount of input information at the target time and the time before the target time, represents the hyperbolic tangent function, represents the Sigmoid function, , , , is the weight matrix, , , , is the bias term, represents vector modulo operation, Represents a vector concatenation operation.

6. The CVT error prediction method according to claim 1, characterized in that: The feature fusion unit is used to determine the error concealment state value at the target moment based on the error vector and the memory state value at the target moment, and the error concealment state value at the previous moment of the target moment.

7. The CVT error prediction method according to claim 6, characterized in that: The error concealment state value at the target time is determined based on the following formula: in, represents the error hidden state value at the target time, Represents the intermediate state of the error hidden state value at the previous moment of the target moment, represents the intermediate state of the error vector at the target time, represents the hyperbolic tangent function, , , is the weight matrix, , , represents the bias term, is the fusion feature allocation coefficient, represents the fusion state value, represents the Sigmoid function, represents the error vector at the target time, Indicates the memory state value at the target moment, Indicates the error hidden state value at the previous moment of the target moment.

8. A CVT error prediction device, characterized in that: include: A preprocessing module, used for preprocessing historical error data and historical environmental parameter data of a target CVT; A prediction module, used to use the preprocessed historical error data and historical environmental parameter data as inputs of an error prediction model, and determine a real-time error prediction result of the target CVT based on an output of the error prediction model; Among them, the error prediction model is composed of multiple multivariate learning networks connected in a loop, and each multivariate learning network includes: a feature selection unit for extracting effective features from preprocessed historical environmental parameter data, a feature memory unit for determining the memory state value of the preprocessed historical environmental parameter data, and a feature fusion unit.

9. An electronic device, characterized in that: The invention comprises a memory and a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the CVT error prediction method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor, the CVT error prediction method according to any one of claims 1 to 7 is implemented.

Images