A method, device, electronic device, and computer-readable medium for predicting the metering error of a charging pile

Abstract

The present invention relates to a method, device, electronic device, and computer-readable medium for predicting the metering error of a charging pile. The method includes: obtaining historical charging data of a target DC charging pile and preprocessing the historical charging data; the historical charging data includes output voltage, output current, state of charge of the battery, ripple coefficient, and energy indication error; training an LSTM model improved based on the coefficient of variation based on the preprocessed historical charging data; inputting the charging data of the target DC charging pile obtained in real time into the trained LSTM model to obtain a predicted value of the energy indication error. The present invention constructs a multi-scale prediction model through the historical data of the charging pile and combines the data fluctuation characteristics, improving the accuracy and adaptability of the prediction of the metering error of the DC charging pile.

Description

Technical Field

[0001] The present invention belongs to the technical field of deep learning and charging pile metering, and particularly relates to a method, device, electronic device, and computer-readable medium for predicting the metering error of a charging pile. Background Art

[0002] With the rapid development of the electric vehicle industry, charging stations, as key facilities for their energy supply, the accuracy and stability of their metering are attracting increasing attention. A main task of a charging station is to provide AC or DC power to electric vehicles. The metering system is a core component of the charging station, which directly affects the charging fees paid by users and the energy replenishment efficiency of electric vehicles.

[0003] In the actual operation process, the metering of a charging station may have errors. These errors may stem from various factors, such as equipment aging, sensor failures, interference from environmental factors, and errors in software algorithms. These factors may all lead to a mismatch between the metering result and the actual charging amount, thereby harming the interests of users and potentially causing disputes.

[0004] To ensure the accuracy of metering, it is very necessary to regularly calibrate the charging station. Traditional calibration methods require the charging station to be in an offline state, which not only hinders its normal operation but may also cause economic losses. Traditional offline calibration techniques no longer meet modern requirements. Using artificial intelligence technology for online error prediction can accurately predict the possible errors in the metering process of a charging station, realize early warning and prompt for abnormal charging stations, thereby improving the accuracy and efficiency of metering and protecting the rights and interests of users. However, due to the metering error of the charging pile being affected by various factors, the error sequence usually changes due to non-stationary fluctuations, making the prediction complex. Therefore, researching how to use intelligent algorithms to more accurately capture this fluctuation for error prediction, thereby improving the accuracy of prediction, has become a current technical problem. Summary of the Invention

[0005] To improve the prediction accuracy and adaptability of the metering error of a charging station, in the first aspect of the present invention, a method for predicting the metering error of a charging pile is provided, including: obtaining the historical charging data of a target DC charging pile and preprocessing the historical charging data; the historical charging data includes output voltage, output current, state of charge of the battery, ripple coefficient, and energy indication error; training an LSTM model improved based on the coefficient of variation based on the preprocessed historical charging data; inputting the charging data of the target DC charging pile obtained in real time into the trained LSTM model to obtain a predicted value of the energy indication error.

[0006] In some embodiments of the present invention, the LSTM model includes: a plurality of LSTM networks for extracting time-series features of multiple scales from the preprocessed historical charging data; a coefficient of variation calculation module for calculating the coefficient of variation of each LSTM network according to the preprocessed historical charging data, and dynamically adjusting the output gate of each LSTM network through the coefficient of variation; a fully connected layer for fusing time-series features of multiple scales and predicting the electricity indication error according to the fused time-series features.

[0007] Further, calculating the coefficient of variation of each LSTM network according to the preprocessed historical charging data includes: extracting multiple electricity indication error sequences of each LSTM network from the preprocessed historical charging data; calculating the coefficient of variation of each LSTM network based on the multiple electricity indication error sequences.

[0008] Furthermore, calculating the coefficient of variation of each LSTM network based on the multiple electricity indication error sequences: calculating the mean and standard deviation based on the multiple electricity indication error sequences; calculating the coefficient of variation based on the mean and standard value.

[0009] Further, dynamically adjusting the output gate of each LSTM network through the coefficient of variation includes: dividing the activation value of the output gate into at least two parts; respectively determining the gating weight matrix of each part through the coefficient of variation; calculating the activation value of the output gate based on the gating weight matrix of each part and the Logistic activation function.

[0010] Furthermore, calculating the activation value of the output gate based on the gating weight matrix of each part and the Logistic activation function is expressed as:

[0011] 。

[0012] Where represents the activation value of the output gate, and respectively represent the dynamic gating weight matrices converted by the coefficient of variation corresponding to when the input data is ; represents the hidden state value at the previous moment; and and are respectively the weight matrices corresponding to generating the first part and value through and the weight matrices corresponding to generating the second part value; and are respectively the weight matrices corresponding to generating the first part and value through The bias term corresponding to the value and generate the second part The bias term corresponding to the value.

[0013] In a second aspect of the present invention, there is provided a device for predicting the metering error of a charging pile, including: an acquisition module, configured to acquire historical charging data of a target DC charging pile and preprocess the historical charging data; the historical charging data includes output voltage, output current, state of charge of the battery, ripple coefficient, and energy indication error; a training module, configured to train an LSTM model improved based on the coefficient of variation based on the preprocessed historical charging data; a prediction module, configured to input the charging data of the target DC charging pile acquired in real time into the trained LSTM model to obtain a predicted value of the energy indication error.

[0014] In a third aspect of the present invention, there is provided an electronic device, including: one or more processors; a storage device, configured to store one or more programs, when the one or more programs are executed by the one or more processors, enabling the one or more processors to implement the method for predicting the metering error of a charging pile provided in the first aspect of the present invention.

[0015] In a fourth aspect of the present invention, there is provided a computer-readable medium, on which a computer program is stored, wherein when the computer program is executed by a processor, the method for predicting the metering error of a charging pile provided in the first aspect of the present invention is implemented.

[0016] The beneficial effects of the present invention are:

[0017] The present invention relates to a method, device, electronic device, and computer-readable medium for predicting the metering error of a charging pile. The method includes: acquiring historical charging data of a target DC charging pile and preprocessing the historical charging data; the historical charging data includes output voltage, output current, state of charge of the battery, ripple coefficient, and energy indication error; training an LSTM model improved based on the coefficient of variation based on the preprocessed historical charging data; inputting the charging data of the target DC charging pile acquired in real time into the trained LSTM model to obtain a predicted value of the energy indication error.

[0018] It can be seen that the present invention constructs a multi-scale LSTM model for historical charging data, and improves the internal part of the LSTM network through the coefficient of variation reflecting the fluctuation degree of the error sequence, enabling the model to utilize the non-stationary feature of the sequence and obtain a more accurate error prediction result. Description of the Drawings

[0019] Figure 1 It is a schematic diagram of the basic process of the method for predicting the metering error of a charging pile in some embodiments of the present invention;

[0020] Figure 2Schematic diagram of the LSTM model in some embodiments of the present invention;

[0021] Figure 3 Structural diagram of the improved LSTM network with coefficient of variation in some embodiments of the present invention;

[0022] Figure 4 Schematic diagram of the basic process of the charging pile metering error prediction method in some embodiments of the present invention;

[0023] Figure 5 Schematic diagram of the structure of an electronic device in some embodiments of the present invention. Detailed implementation manners

[0024] The principles and features of the present invention will be described below with reference to the accompanying drawings. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0025] Refer to Figure 1 And Figure 2 In the first aspect of the present invention, a charging pile metering error prediction method is provided, including: S100. Obtain historical charging data of a target DC charging pile and preprocess the historical charging data; the historical charging data includes output voltage, output current, state of charge of the battery, ripple coefficient, and energy indication error; S200. Based on the preprocessed historical charging data, train an LSTM model improved based on the coefficient of variation; S300. Input the charging data of the target DC charging pile obtained in real time into the trained LSTM model to obtain a predicted value of the energy indication error.

[0026] In step S100 of some embodiments of the present invention, obtain historical charging data of a target DC charging pile and preprocess the historical charging data; the historical charging data includes output voltage, output current, state of charge of the battery, ripple coefficient, and energy indication error;

[0027] Specifically, collect historical charging data of a sample DC charging pile, where the historical charging data includes the output voltage U, current I, state of charge S of the battery, ripple coefficient K, and energy indication error E of the charging pile, and preprocess the obtained data.

[0028] Collect secondary voltage data U, current data I, state of charge S of the battery, ripple coefficient K, and energy indication error E. Since the above collected data are original collected data, there may be problems such as data missing and large differences in order of magnitude, which affect the training quality of the model. Therefore, before inputting into the model, various types of data obtained are subjected to outlier cleaning and standardization processing.

[0029] Refer to Figure 2, in step S200 or S300 of some embodiments of the present invention, the LSTM model includes: a plurality of LSTM networks for extracting time series features of multiple scales from the preprocessed historical charging data; a coefficient of variation calculation module for calculating the coefficient of variation of each LSTM network according to the preprocessed historical charging data, and dynamically adjusting the output gate of each LSTM network through the coefficient of variation; a fully connected layer for fusing time series features of multiple scales and predicting the electricity indication error according to the fused time series features.

[0030] Specifically, after the input data enters the multi-scale convolutional layer, multi-scale feature extraction is performed through convolutions of 1×1, 3×3, and 5×5 respectively. At this time, the convolution result of 1×1 is denoted as the time series feature of the first scale, the convolution result of 3×3 is denoted as the time series feature of the second scale, and the convolution result of 5×5 is denoted as the time series feature of the third scale; the time series feature of the first scale obtained through the 1×1 convolution is directly sent into the LSTM as an input feature on the one hand, and on the other hand, it will be upsampled and then superimposed and fused with the time series feature of the second scale obtained through the 3×3 convolution, and then the fused result is sent into the LSTM as an input feature. Similarly, the time series feature of the second scale obtained through the 3×3 convolution is input into the LSTM after being fused with the upsampled time series feature of the first scale on the one hand, and on the other hand, it will be upsampled and then superimposed and fused with the time series feature of the third scale obtained through the 5×5 convolution, and then the fused result is also sent into the LSTM.

[0031] It can be understood that the above describes the process of constructing an error prediction model combining the coefficient of variation and multi-scale LSTM. The model provided by the present invention designs a multi-scale LSTM network for extracting time series features of different granularities of input data, and at the same time combines the coefficient of variation that can reflect the fluctuation degree of the error sequence to improve the inside of the LSTM network, so that the model can utilize the non-stationary feature of the sequence to obtain a more accurate error prediction result.

[0032] Further, calculating the coefficient of variation of each LSTM network according to the preprocessed historical charging data includes: extracting multiple electricity indication error sequences of each LSTM network from the preprocessed historical charging data; calculating the coefficient of variation of each LSTM network based on the multiple electricity indication error sequences.

[0033] Specifically, if there is an electricity indication error sequence in the current input data , is the electricity indication error value at time t. The coefficient of variation is calculated through this data. The smaller the coefficient of variation, the lower the volatility of the data. On the contrary, the larger the coefficient of variation, the higher the volatility of the data. The specific calculation process is as follows:

[0034] First, calculate the mean and standard deviation of the sequence data:

[0035] Mean: .

[0036] Standard deviation: .

[0037] In the formula, is the sequence mean, T is the length corresponding to the sequence at time t, is the power indication error value at time t. is the sequence standard deviation.

[0038] Calculate the corresponding coefficient of variation through the mean and standard deviation of the sequence at this time .

[0039] .

[0040] Refer to Figure 3 , furthermore, based on multiple power indication error sequences, calculate the coefficient of variation of each LSTM network: calculate the mean and standard deviation based on multiple power indication error sequences; calculate the coefficient of variation based on the mean and standard values.

[0041] Further, dynamically adjusting the output gate of each LSTM network through the coefficient of variation includes: dividing the activation value of the output gate into at least two parts; respectively determining the gating weight matrix of each part through the coefficient of variation; calculating the activation value of the output gate based on the gating weight matrix of each part and the Logistic activation function.

[0042] Specifically, Figure 3 shows the improved LSTM network structure diagram provided by the present invention. The input gate, forget gate and memory unit therein are not different from the traditional LSTM network. The difference is that the coefficient of variation is introduced to set the dynamic output gate. If the sequence value of the feature sequence X extracted by multi-scale convolution at time t is input into the network, the forget gate, input gate, output gate and memory unit will be updated and calculated respectively. The specific process is as follows:

[0043] .

[0044] .

[0045] .

[0046] .

[0047] The calculations for updating the forget gate, input gate and memory unit are not different from those of the traditional LSTM, and are determined by the current input and the hidden state value at the previous moment are calculated. As shown in the above formula, where is the calculation result of the forget gate. is the activation value calculated by the input gate, is the candidate memory cell. is the Logistic activation function, is the hyperbolic tangent activation function. , , are respectively the weight matrices for calculating , , when. , , are respectively the bias terms for calculating , , when. is the updated memory cell at the current moment, is the memory cell at the previous moment. is element-wise multiplication.

[0048] The calculation of the output gate dynamically adjusts the activation value by combining the coefficient of variation. Specifically, it can be explained as follows: The activation value of the output gate is split into two parts, and are respectively dynamically regulated by the weight matrix and of the coefficient of variation processed by the Logistic activation function. It can be expressed by the formula:

[0049] .

[0050] In the above formula, is the calculation result of the activation value of the improved output gate. and are the dynamic gate control weight matrices converted from the coefficient of variation when the input data is . is the hidden state value at the previous moment. and are respectively the weight matrices corresponding to generating the first part and value through and the weight matrix corresponding to generating the second part value. and are respectively the weight matrices corresponding to generating the first part and through The bias term corresponding to the value and generate the second part The bias term corresponding to the value.

[0051] Then the hidden state output of the improved LSTM can be obtained by and calculated and expressed by the formula as:

[0052] .

[0053] Since the time series features of three different scales will respectively pass through three improved LSTMs, and three hidden state results are output. These three hidden state results will then be fed into the fully connected layer, and the three results are integrated through the softmax function to output the final error prediction result.

[0054] In step S300 of some embodiments of the present invention, the charging data of the target DC charging pile obtained in real time is input into the trained LSTM model to obtain the predicted value of the electric energy indication error.

[0055] Specifically, the charging pile electric energy indication error prediction model is trained through the collected historical data. After the training is completed, the trained model is used to accurately predict the electric energy indication error of the charging pile at a future moment, and the predicted result of the charging electric energy indication error can be obtained.

[0056] Embodiment 2

[0057] Referring to Figure 4 , in the second aspect of the present invention, there is provided a charging pile metering error prediction device 1 including: an acquisition module 11, configured to acquire the historical charging data of the target DC charging pile and preprocess the historical charging data; the historical charging data includes output voltage, output current, state of charge of the battery, ripple coefficient, and electric energy indication error; a training module 12, configured to train an LSTM model improved based on the coefficient of variation based on the preprocessed historical charging data; a prediction module 13, configured to input the charging data of the target DC charging pile obtained in real time into the trained LSTM model to obtain the predicted value of the electric energy indication error.

[0058] Further, the acquisition module 11 includes: a preprocessing unit, configured to perform outlier cleaning and standardization processing on the historical charging data.

[0059] Embodiment 3

[0060] Referring to Figure 5, a third aspect of the present invention provides an electronic device, comprising: one or more processors; a storage device for storing one or more programs, which when executed by the one or more processors, cause the one or more processors to implement the charging pile metering error prediction method of the first aspect of the present invention.

[0061] The electronic device 500 may include a processing device (such as a central processing unit, a graphics processing unit, etc.) 501, which may perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 502 or the program loaded from the storage device 508 into the random access memory (RAM) 503. In the random access memory 503, various programs and data required for the operation of the electronic device 500 are also stored. The processing device 501, the read-only memory 502, and the random access memory 503 are connected to each other through a bus 504. The input / output (I / O) interface 505 is also connected to the bus 504.

[0062] Generally, the following devices may be connected to the I / O interface 505: an input device 506 including, for example, a touch screen, a touchpad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 507 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 508 including, for example, a hard disk, etc.; and a communication device 509. The communication device 509 may allow the electronic device 500 to communicate with other devices wirelessly or wiredly to exchange data. Although Figure 5 the electronic device 500 with various devices is shown, it should be understood that it is not required to implement or include all the shown devices. Instead, more or fewer devices may be implemented or included. Figure 5 Each block shown in may represent a device or, as needed, multiple devices.

[0063] In particular, according to an embodiment of the present disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, an embodiment of the present disclosure includes a computer program product that includes a computer program carried on a computer-readable medium, and the computer program includes program code for performing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via a communication device 509, or installed from a storage device 508, or installed from a read-only memory 502. When the computer program is executed by a processing device 501, the above-described functions defined in the method of the embodiment of the present disclosure are performed. It should be noted that the computer-readable medium described in the embodiments of the present disclosure can be a computer-readable signal medium, a computer-readable storage medium, or any combination of the two. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples of the computer-readable storage medium can include, but are not limited to: an electrical connection having one or more wires, 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), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the embodiments of the present disclosure, the computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In the embodiments of the present disclosure, the computer-readable signal medium can include a data signal propagated in a baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer-readable signal medium can also be any computer-readable medium other than the computer-readable storage medium, and the computer-readable signal medium can send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted by any appropriate medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0064] The above computer-readable medium can be included in the above electronic device; or it can exist separately and not be assembled into the electronic device. The above computer-readable medium carries one or more computer programs, and when the above one or more programs are executed by the electronic device, the electronic device is caused to:

[0065] Computer program code for performing the operations of the embodiments of the present disclosure may be written in one or more programming languages or combinations thereof. The programming languages include object-oriented programming languages such as Java, Smalltalk, C++, Python, and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code may execute entirely on the user's computer, partially on the user's computer, execute as a stand-alone software package, execute partially on the user's computer and partially on a remote computer, or execute entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any kind of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0066] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagram may represent a module, a program segment, or a portion of code that contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than marked in the accompanying drawings. For example, two consecutive blocks shown may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should be noted that each block in the block diagram and / or flowchart, and combinations of blocks in the block diagram and / or flowchart, may be implemented by a dedicated hardware-based system that performs the specified functions or operations, or may be implemented by a combination of dedicated hardware and computer instructions.

[0067] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A DC charging pile measurement error prediction method based on coefficient of variation and LSTM, characterized in that: include: Acquire historical charging data of a target DC charging pile, and preprocess the historical charging data; The historical charging data includes output voltage, output current, battery charge state, ripple factor and electric energy indication error; Based on the preprocessed historical charging data, train the LSTM model based on the improved coefficient of variation; The LSTM model includes: multiple LSTM networks, which are used to extract time series features of multiple scales from preprocessed historical charging data; a coefficient of variation calculation module, which is used to calculate the coefficient of variation of each LSTM network according to the preprocessed historical charging data, and dynamically adjust the output gate of each LSTM network by the coefficient of variation: divide the activation value of the output gate into at least two parts; determine the gate weight matrix of each part respectively by the coefficient of variation; calculate the activation value of the output gate based on the gate weight matrix of each part and the Logistic activation function: ， in represents the activation value of the output gate, and Respectively represent the input data as The corresponding coefficient of variation the transformed dynamic gating weight matrix; Indicates the hidden state value at the previous moment; and Through and Generate the first part The weight matrix corresponding to the value and the second part are generated The weight matrix corresponding to the value; and Through and Generate the first part The bias term corresponding to the value and the second part of the generation The bias item corresponding to the value; a fully connected layer, used to fuse time series features of multiple scales, and predict the electric energy indication error according to the fused time series features; wherein, the calculation of the coefficient of variation of each LSTM network according to the preprocessed historical charging data includes: extracting multiple electric energy indication error sequences of each LSTM network from the preprocessed historical charging data; calculating the coefficient of variation of each LSTM network based on multiple electric energy indication error sequences: calculating the mean and standard deviation based on multiple electric energy indication error sequences; calculating the coefficient of variation based on the mean and standard value; The charging data of the target DC charging pile acquired in real time is input into the trained LSTM model to obtain the predicted value of the electric energy indication error.

2. A DC charging pile measurement error prediction device based on coefficient of variation and LSTM, characterized in that: include: An acquisition module, used to acquire historical charging data of a target DC charging pile and pre-process the historical charging data; The historical charging data includes output voltage, output current, battery charge state, ripple factor and electric energy indication error; The training module is used to train the LSTM model based on the improved coefficient of variation based on the preprocessed historical charging data; Based on the preprocessed historical charging data, train the LSTM model based on the improved coefficient of variation; The LSTM model includes: multiple LSTM networks, which are used to extract time series features of multiple scales from preprocessed historical charging data; a coefficient of variation calculation module, which is used to calculate the coefficient of variation of each LSTM network according to the preprocessed historical charging data, and dynamically adjust the output gate of each LSTM network by the coefficient of variation: divide the activation value of the output gate into at least two parts; determine the gate weight matrix of each part respectively by the coefficient of variation; calculate the activation value of the output gate based on the gate weight matrix of each part and the Logistic activation function: ， in represents the activation value of the output gate, and Respectively represent the input data as The corresponding coefficient of variation the transformed dynamic gating weight matrix; Indicates the hidden state value at the previous moment; and Through and Generate the first part The weight matrix corresponding to the value and the second part are generated The weight matrix corresponding to the value; and Through and Generate the first part The bias term corresponding to the value and the second part of the generation The bias item corresponding to the value; a fully connected layer, used to fuse time series features of multiple scales, and predict the electric energy indication error according to the fused time series features; wherein, the calculation of the coefficient of variation of each LSTM network according to the preprocessed historical charging data includes: extracting multiple electric energy indication error sequences of each LSTM network from the preprocessed historical charging data; calculating the coefficient of variation of each LSTM network based on multiple electric energy indication error sequences: calculating the mean and standard deviation based on multiple electric energy indication error sequences; calculating the coefficient of variation based on the mean and standard value; The prediction module is used to input the charging data of the target DC charging pile acquired in real time into the trained LSTM model to obtain the predicted value of the electric energy indication error.

3. The DC charging pile measurement error prediction device based on coefficient of variation and LSTM according to claim 2 is characterized in that: The acquisition module comprises: The preprocessing unit is used to perform outlier cleaning and standardization processing on the historical charging data.

4. An electronic device comprising: one or more processors; A storage device for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the DC charging pile metering error prediction method based on coefficient of variation and LSTM as claimed in claim 1.

5. A computer readable medium having a computer program stored thereon, wherein: When the computer program is executed by the processor, the DC charging pile metering error prediction method based on coefficient of variation and LSTM as claimed in claim 1 is implemented.

Images