Remote monitoring method for charging equipment
Through multi-sensor signal fusion and fault diagnosis model and combined with greedy algorithm to optimize the maintenance solution, the problems of insufficient fault diagnosis accuracy and simple maintenance plan in traditional charging equipment monitoring methods are solved, and efficient and intelligent equipment status monitoring and maintenance are achieved.
Patent Information
- Application Number
- CN202510398550.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-31
- Publication Date
- 2025-07-18
AI Technical Summary
Traditional charging equipment monitoring methods rely on a single signal source, resulting in insufficient accuracy of fault diagnosis and lack of intelligent prediction mechanisms. Manual intervention affects accuracy and efficiency, and the maintenance plan is simple, ignoring the complexity and differences of equipment.
Multiple vibration sensors and current sensors are used to collect signals, and the envelope curve is extracted through vibration signal fusion and Hilbert transform, and the fault level is identified in combination with a pre-trained fault diagnosis model, and the maintenance plan is optimized through a greedy algorithm.
Improve the accuracy and robustness of fault detection, reduce manual intervention, quickly respond to equipment abnormalities, optimize maintenance plans, improve equipment reliability and operational efficiency, and reduce long-term maintenance costs.
Smart Images
Figure CN120334623A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of charging device monitoring, and specifically to a remote monitoring method for charging devices. Background Art
[0002] Traditional methods usually rely only on a single signal source, such as only using current signals or vibration signals. This method may not be able to comprehensively capture the state information of the device. Mechanical failures and electrical failures often exhibit differently in different signal dimensions. A single signal source may lead to insufficient accuracy in fault diagnosis. Moreover, in traditional methods, the processing of vibration signals often fails to effectively remove noise and redundant information, thus affecting the extraction of fault features. This will result in a weak ability to identify fault patterns, possibly missing key fault information, reducing the robustness and accuracy of diagnosis. And traditional methods rely on manual intervention for fault judgment and classification. Especially in complex fault situations, manual analysis may be affected by subjective factors, leading to incorrect judgments or omissions, thus delaying the fault response time, increasing labor costs and operation risks. And traditional methods usually lack an intelligent fault prediction mechanism. They often rely on regular inspections or empirical maintenance cycles to arrange maintenance, making it difficult to detect potential fault risks in a timely manner, lacking dynamic monitoring and prediction of the device state, which may result in the device not being effectively repaired before a fault occurs, thereby affecting the normal operation of the charging device. And traditional methods are usually relatively simple in formulating maintenance plans. They may only handle according to fixed maintenance cycles or simple fault situations of the device, ignoring the complexity of device faults and the differences between different devices. This may lead to waste of resources or faults not being processed in a timely and effective manner. Summary of the Invention
[0003] The technical problem to be solved by the present invention is to overcome the above-mentioned shortcomings of the prior art and provide a remote monitoring method for charging devices.
[0004] The technical solution adopted to solve the above technical problem is: A remote monitoring method for charging devices, including:
[0005] Collecting signals from each charging device in the target area based on a preset plurality of vibration sensors to obtain a plurality of vibration signal sequences of the charging device, wherein the vibration signal is composed of a plurality of vibration data points, and collecting signals from each charging device in the target area based on a current sensor to obtain the current signal of the charging device;
[0006] Calculating the action coefficient of the vibration data points in the vibration signal sequence, and fusing the plurality of vibration signal sequences based on the action coefficient to obtain a fused vibration signal;
[0007] Perform Hilbert transform on the current signal and the fused vibration signal to obtain the first envelope curve of the current signal and the second envelope curve of the fused vibration signal;
[0008] Perform fault diagnosis on the first envelope curve and the second envelope curve based on a pre-trained fault diagnosis model to obtain the fault level label of the charging device;
[0009] Construct a maintenance optimization model for the maintenance of the charging device in the target area based on the fault level label of the charging device;
[0010] Solve the maintenance optimization model based on the greedy algorithm to obtain the optimal maintenance plan for the charging devices in the target area.
[0011] Preferably, calculating the action coefficient of the vibration data points in the vibration signal sequence includes:
[0012] For the total energy of each vibration signal in the vibration signal sequence, where the calculation formula for the total energy of the vibration signal is as follows:
[0013]
[0014] where E i represents the total energy of the i-th vibration signal in the vibration signal sequence, n represents the number of vibration data points in the vibration signal, and s iq represents the q-th vibration data point in the i-th vibration signal in the vibration signal sequence;
[0015] Correct the vibration data points of the vibration signal based on the total energy of the vibration signal to obtain corrected vibration data points, where the modification formula is as follows:
[0016]
[0017] where, represents the q-th corrected vibration data point in the i-th vibration signal in the vibration signal sequence;
[0018] Calculate the action coefficient of the vibration data points in the vibration signal based on the corrected vibration data points, where the calculation formula for the action coefficient is as follows:
[0019]
[0020] where Eff iq represents the action coefficient of the q-th vibration data point in the i-th vibration signal in the vibration signal sequence.
[0021] Preferably, based on the action coefficient, the multiple vibration signal sequences are fused to obtain a fused vibration signal, including:
[0022] Calculating the fusion weights of the vibration data points in the multiple vibration signals according to the action coefficients of the vibration data points in the vibration signals, wherein the calculation formula of the fusion weights is as follows:
[0023]
[0024] Where ω iq represents the fusion weight of the q-th vibration data point in the i-th vibration signal in the vibration signal sequence;
[0025] Fusing each vibration data point in the multiple vibration signals according to the fusion weights of the vibration data points in the multiple vibration signal sequences to obtain fused vibration data points, and obtaining the fused vibration signal based on the fused vibration data points, wherein the fusion formula of the fused vibration data points is as follows:
[0026]
[0027] Where S q represents the q-th fused vibration data point in the fused vibration signal, and I represents the number of vibration signal sequences.
[0028] Preferably, performing Hilbert transform on the current signal and the fused vibration signal to obtain a first envelope curve of the current signal and a second envelope curve of the fused vibration signal, including:
[0029] Performing Hilbert transform on the current signal to obtain a first transformed signal, wherein the expression of the first transformed signal is as follows:
[0030]
[0031] Where represents the first transformed signal of the current signal x1(t);
[0032] Calculating the first envelope curve of the current signal based on the first transformed signal, wherein the expression of the first envelope curve is as follows:
[0033]
[0034] Where c1(t) represents the first envelope curve;
[0035] Performing Hilbert transform on the fused vibration signal to obtain a second transformed signal, wherein the expression of the second transformed signal is as follows:
[0036]
[0037] Among them, represents the second transformed signal of the fused vibration signal x2(t);
[0038] Calculate the second envelope curve of the fused vibration signal based on the second transformed signal, where the expression of the second envelope curve is as follows:
[0039]
[0040] Among them, c2(t) represents the second envelope curve.
[0041] Preferably, the fault diagnosis model includes a first feature extraction module, a second feature extraction module, a feature connection module, and a fault classification module. The first feature extraction module is used to perform convolution on the first envelope curve to obtain a first feature vector, perform a max pooling operation on the first feature vector to obtain a first dimensionality-reduced feature vector, perform batch normalization processing on the first dimensionality-reduced feature vector to obtain a first normalized feature vector, estimate the feature importance of the first normalized feature vector through global average pooling and two fully connected layers to obtain a first evaluation vector. Among them, the two fully connected layers respectively adopt ReLU and Softmax activation functions, perform a vector dot product on the first normalized feature vector and the first evaluation vector to obtain a comprehensive feature. The second feature extraction module is used to perform convolution on the second envelope curve to obtain a second feature vector, perform a max pooling operation on the second feature vector to obtain a second dimensionality-reduced feature vector, perform batch normalization processing on the second dimensionality-reduced feature vector to obtain a second normalized feature vector. The feature connection module is used to perform a connection operation on the comprehensive feature and the second normalized feature vector to obtain a connection vector. The fault classification module is used to perform fault classification on the connection vector through the Softmax activation function to obtain a fault probability, and obtain the fault level label of the charging device based on the fault probability.
[0042] Preferably, the calculation formula of the first normalized feature vector is as follows:
[0043]
[0044] Among them, represents the first normalized feature vector, BN represents the batch normalization processing function, MP represents the max pooling operation function, f represents a preset activation function, and represent the first neuron bias and the kernel of the first neuron of the convolution, and cov1D represents the convolution pooling function;
[0045] The calculation formula of the second normalized feature vector is as follows:
[0046]
[0047] Among them, represents the second normalized eigenvector, and represent the second neuron bias of the convolution and the kernel of the second neuron;
[0048] The calculation formula of the connection vector is as follows:
[0049]
[0050] Among them, represents the connection vector, FC represents the fully connected layer, and GA represents the global average pooling function.
[0051] Preferably, a maintenance optimization model for repairing charging devices in the target area is constructed based on the fault level labels of the charging devices, including:
[0052] Construct an objective function based on the maintenance cost corresponding to each fault level label of the repaired charging device, where the expression of the objective function is as follows:
[0053]
[0054] Among them, MIN: L represents the objective function, θ i,j represents the fault level of the charging device, C i，j represents the cost required for maintenance with a specific priority, d i，j represents whether the implementation of the corresponding maintenance work requires downtime. If d i,j = 1, it means that the implementation of the maintenance measure requires downtime. If d i,j = 0, it means that the implementation of the maintenance measure does not require downtime, C D represents the fixed startup cost after the charging device is shut down, c e represents the opportunity cost of downtime per unit time, c o represents the downtime loss cost per unit time, s i represents the downtime of the corresponding charging device, c i，j，s represents the unit fixed setting cost of the charging device under the corresponding maintenance work, C i，j，L represents the loss cost caused by the incomplete utilization of the charging device, I represents the total number of charging devices to be repaired, and J represents the total number of types of fault level labels;
[0055] Construct a constraint condition based on the maximum charging device downtime, where the expression of the first constraint condition is as follows:
[0056] s i ≤ STi ;
[0057] Among them, ST i represents the maximum downtime of the charging device;
[0058] Construct a maintenance optimization model based on the objective function and the constraint conditions.
[0059] Preferably, solve the maintenance optimization model based on the greedy algorithm to obtain the optimal maintenance plan for the charging devices in the target area, including:
[0060] Randomly select a charging device to be repaired, calculate the maintenance cost of the charging device to be repaired within the time window, find the time window with the minimum maintenance cost, and set the latest time point within the time window as the maintenance start time of the charging device to be repaired;
[0061] Calculate the downtime of each charging device to be repaired, and judge whether the downtime of the charging device to be repaired meets the constraint conditions. If the constraint conditions are met, adjust the maintenance start time. The adjustment formula for the maintenance start time is as follows:
[0062] t s，i = max((min(C i (1), C i (1),..., C i (Q)))-(s i - ST i ));
[0063] Among them, t s，i represents the maintenance start time of the i-th charging device to be repaired, min represents the minimization function, and C i (Q) represents the maintenance cost of the Q-th time window of the i-th charging device to be repaired;
[0064] Calculate the total maintenance cost of all charging devices to be repaired, continuously adjust the initial scheduling of the charging devices to be repaired, generate a new maintenance plan, and calculate the total cost;
[0065] Compare the total maintenance costs corresponding to all maintenance plans to obtain the optimal maintenance plan.
[0066] The beneficial effects of the present invention are as follows: (1) By collecting both vibration signals and current signals, the present invention can obtain the status information of the charging device from different dimensions, improving the accuracy of fault detection. The vibration signal can reveal the details of mechanical faults in the device, while the current signal reflects the electrical performance of the device. Combining the information of both can provide a more comprehensive understanding of the device operation status. Moreover, by fusing multiple vibration signal sequences into a comprehensive vibration signal through the action coefficient, it helps to remove noise and redundant information, highlighting valuable fault features, thus enhancing the robustness of fault diagnosis. The fused signal undergoes Hilbert transform to further extract envelope features and improve the ability to identify fault patterns; (2) By analyzing the envelope curve based on a pre-trained fault diagnosis model, the present invention can accurately identify the fault type and level of the device, reducing the need for manual intervention. It can quickly respond when the device shows abnormalities and avoid fault expansion. Moreover, based on the fault level label of the device, a maintenance optimization model can be further constructed. Combining the status of charging devices in the target area, it can predict the maintenance needs of the device, prioritize the handling of serious faults or high-risk devices, reduce downtime, and improve the overall reliability of the system; (3) By using the greedy algorithm to solve the maintenance optimization model, the present invention can quickly obtain the optimal maintenance plan. This method rationally allocates resources by minimizing maintenance costs and fault impacts, making the maintenance work more efficient, avoiding conflicts in maintenance plans, and improving the overall operation efficiency of charging devices in the area. The entire process is highly automated, from signal acquisition, fault diagnosis to maintenance plan optimization, which can be achieved through an intelligent system, reducing the dependence on manual monitoring and judgment, improving the operation efficiency. Moreover, through accurate fault diagnosis and timely maintenance optimization, it can effectively reduce the device failure rate, extend the service life of the charging device, thereby reducing long-term maintenance costs and improving the operation efficiency of the device. BRIEF DESCRIPTION OF THE DRAWINGS
[0067] Figure 1 It is a schematic flowchart of the steps of the overall method in an embodiment proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0068] Embodiment 1, as Figure 1 shown, the remote monitoring method for a charging device proposed by the present invention includes:
[0069] S1. Based on a preset plurality of vibration sensors, signal acquisition is performed on each charging device in the target area to obtain a plurality of vibration signal sequences of the charging device. Among them, the vibration signal is composed of a plurality of vibration data points. Based on a current sensor, signal acquisition is performed on each charging device in the target area to obtain the current signal of the charging device;
[0070] S2. Calculate the action coefficients of the vibration data points in the vibration signal sequence, and fuse multiple vibration signal sequences based on the action coefficients to obtain a fused vibration signal;
[0071] S3. Perform Hilbert transform on the current signal and the fused vibration signal to obtain the first envelope curve of the current signal and the second envelope curve of the fused vibration signal;
[0072] S4. Perform fault diagnosis on the first envelope curve and the second envelope curve based on a pre-trained fault diagnosis model to obtain the fault level label of the charging device;
[0073] S5. Construct a maintenance optimization model for maintaining the charging device in the target area based on the fault level label of the charging device;
[0074] S6. Solve the maintenance optimization model based on the greedy algorithm to obtain the optimal maintenance plan for the charging devices in the target area.
[0075] In the present invention, the vibration sensor is used to measure the vibration condition of an object, and usually can capture the mechanical vibration generated during the operation of the device. Through these vibration data, the operation state and possible faults of the device can be inferred; the current sensor is a sensor used to measure the magnitude of the current. In this method, the current sensor is used to obtain the current signal of the charging device, and the change of the current may reveal the working state or fault information of the device; the action coefficient refers to the weight value of each data point or signal in the vibration signal, and is usually used to characterize the contribution of each vibration signal to the fault diagnosis; signal fusion refers to combining multiple signals (here are vibration signals) into a more comprehensive signal according to certain rules or methods, which helps to improve the accuracy of the analysis result, because different sensor signals can provide different device state information; the Hilbert transform is a mathematical transform method used to extract the envelope information of a signal. Through the Hilbert transform, a signal can be transformed into its amplitude envelope curve to more clearly analyze the dynamic change of the signal; the envelope curve is the outer boundary of the signal, used to represent the amplitude change trend of the signal. For the current signal and the vibration signal, the envelope curve can help identify abnormal changes in the device, so as to judge whether a fault occurs; the fault level label assigns a fault level label to the charging device according to the output of the fault diagnosis model, indicating the fault severity or state of the device; the greedy algorithm is an algorithm for solving optimization problems, and the goal is to approach the global optimal solution step by step through local optimal selection. The greedy algorithm is used to solve the maintenance optimization model to help determine which devices should be repaired first to achieve the optimal maintenance plan.
[0076] Embodiment 2. The remote monitoring method for the charging device proposed by the present invention. Compared with Embodiment 1, this embodiment further includes: calculating the action coefficients of the vibration data points in the vibration signal sequence, including:
[0077] A1. For the total energy of each vibration signal in the vibration signal sequence, where the calculation formula for the total energy of the vibration signal is as follows:
[0078]
[0079] where E i represents the total energy of the i-th vibration signal in the vibration signal sequence, n represents the number of vibration data points in the vibration signal, and s iq represents the q-th vibration data point in the i-th vibration signal in the vibration signal sequence;
[0080] A2. Based on the total energy of the vibration signal, correct the vibration data points of the vibration signal to obtain corrected vibration data points, where the correction formula is as follows:
[0081]
[0082] where represents the q-th corrected vibration data point in the i-th vibration signal in the vibration signal sequence;
[0083] A3. Calculate the influence coefficient of the vibration data points in the vibration signal based on the corrected vibration data points, where the calculation formula for the influence coefficient is as follows:
[0084]
[0085] where Eff iq represents the influence coefficient of the q-th vibration data point in the i-th vibration signal in the vibration signal sequence.
[0086] In this embodiment, the total energy of the vibration signal is a parameter used to measure the intensity of the vibration signal; in order to analyze the vibration signal more accurately, it may be necessary to correct the original vibration data points. The purpose of the correction is to remove noise or standardize the signal to make it more in line with the actual analysis requirements.
[0087] In an alternative embodiment, fuse multiple vibration signal sequences based on the influence coefficient to obtain a fused vibration signal, including:
[0088] B1. Calculate the fusion weight of the vibration data points in the vibration signals based on the influence coefficient of the vibration data points in the vibration signal, where the calculation formula for the fusion weight is as follows:
[0089]
[0090] where ω iq represents the fusion weight of the q-th vibration data point in the i-th vibration signal in the vibration signal sequence;
[0091] B2. Fuse each vibration data point in multiple vibration signals based on the fusion weights of the vibration data points in multiple vibration signal sequences to obtain fused vibration data points, and obtain a fused vibration signal based on the fused vibration data points. The fusion formula for the fused vibration data points is as follows:
[0092]
[0093] where S q represents the q-th fused vibration data point in the fused vibration signal, and I represents the number of vibration signal sequences.
[0094] In an optional embodiment, perform Hilbert transform on the current signal and the fused vibration signal to obtain a first envelope curve of the current signal and a second envelope curve of the fused vibration signal, including:
[0095] C1. Perform Hilbert transform on the current signal to obtain a first transformed signal. The expression of the first transformed signal is as follows:
[0096]
[0097] where represents the first transformed signal of the current signal x1(t);
[0098] C2. Calculate the first envelope curve of the current signal based on the first transformed signal. The expression of the first envelope curve is as follows:
[0099]
[0100] where c1(t) represents the first envelope curve;
[0101] C3. Perform Hilbert transform on the fused vibration signal to obtain a second transformed signal. The expression of the second transformed signal is as follows:
[0102]
[0103] where represents the second transformed signal of the fused vibration signal x2(t);
[0104] C4. Calculate the second envelope curve of the fused vibration signal based on the second transformed signal. The expression of the second envelope curve is as follows:
[0105]
[0106] where c2(t) represents the second envelope curve.
[0107] In an alternative embodiment, the fault diagnosis model includes a first feature extraction module, a second feature extraction module, a feature connection module, and a fault classification module. The first feature extraction module is used to perform convolution on the first envelope curve to obtain a first feature vector, perform a max pooling operation on the first feature vector to obtain a first dimensionality-reduced feature vector, perform batch normalization on the first dimensionality-reduced feature vector to obtain a first normalized feature vector, estimate the feature importance of the first normalized feature vector through global average pooling and two fully-connected layers to obtain a first evaluation vector, where the two fully-connected layers respectively adopt ReLU and Softmax activation functions, perform a vector dot product on the first normalized feature vector and the first evaluation vector to obtain a comprehensive feature. The second feature extraction module is used to perform convolution on the second envelope curve to obtain a second feature vector, perform a max pooling operation on the second feature vector to obtain a second dimensionality-reduced feature vector, perform batch normalization on the second dimensionality-reduced feature vector to obtain a second normalized feature vector. The feature connection module is used to perform a connection operation on the comprehensive feature and the second normalized feature vector to obtain a connection vector. The fault classification module is used to perform fault classification on the connection vector through the Softmax activation function to obtain a fault probability, and obtain a fault level label of the charging device based on the fault probability.
[0108] It should be noted that convolution is a commonly used signal processing and image processing technology, widely used in convolutional neural networks (CNNs). It performs a convolution operation on the input signal through a filter (or convolution kernel) to extract features in the signal; batch normalization is a technology in deep learning used to accelerate the training process and improve the stability of the model; global average pooling is a pooling method that reduces the dimensionality of data by calculating the average value of the entire feature map; a fully-connected layer is a neural network layer where each input node is connected to the output node; feature connection is to splice feature vectors from different sources into a new vector.
[0109] In an alternative embodiment, the calculation formula for the first normalized feature vector is as follows:
[0110]
[0111] where represents the first normalized feature vector, BN represents the batch normalization processing function, MP represents the max pooling operation function, f represents a preset activation function, and represent the first neuron bias and the kernel of the first neuron of the convolution, and cov1D represents the convolution pooling function;
[0112] The calculation formula for the second normalized feature vector is as follows:
[0113]
[0114] Among them, represents the second normalized eigenvector, and represent the second neuron bias of the convolution and the kernel of the second neuron;
[0115] The calculation formula of the connection vector is as follows:
[0116]
[0117] Among them, represents the connection vector, FC represents the fully connected layer, and GAP represents the global average pooling function.
[0118] In an alternative embodiment, a maintenance optimization model for maintaining a charging device in a target area is constructed based on the fault level label of the charging device, including:
[0119] D1. Construct an objective function based on the maintenance cost corresponding to each fault level label of the maintenance charging device, where the expression of the objective function is as follows:
[0120]
[0121] Among them, MIN: L represents the objective function, θ i,j represents the fault level of the charging device, C i,j represents the cost required for maintenance of a specific priority, d i，j represents whether the implementation of the corresponding maintenance work requires downtime. If d i,j = 1, it means that the implementation of the maintenance measure requires downtime. If d i,j = 0, it means that the implementation of the maintenance measure does not require downtime, C D represents the fixed start-up cost after the charging device is shut down, c e represents the opportunity cost of downtime per unit time, c o represents the downtime loss cost per unit time, s i represents the downtime of the corresponding charging device, c i，j，s represents the unit fixed setting cost of the charging device under the corresponding maintenance work, C i，j，L represents the loss cost caused by the incomplete utilization of the charging device, I represents the total number of charging devices to be maintained, and J represents the total number of types of fault level labels;
[0122] D2. Construct a constraint condition based on the maximum charging device downtime, where the expression of the first constraint condition is as follows:
[0123] s i ≤ ST i ;
[0124] Among them, STi Represents the maximum charging device downtime;
[0125] D3. Construct a maintenance optimization model based on the objective function and constraints.
[0126] In an alternative embodiment, solve the maintenance optimization model based on the greedy algorithm to obtain the optimal maintenance plan for the charging devices in the target area, including:
[0127] E1. Randomly select a charging device to be repaired, calculate the maintenance cost of the charging device to be repaired within the time window, find the time window with the minimum maintenance cost, and set the latest time point within the time window as the maintenance start time of the charging device to be repaired;
[0128] E2. Calculate the downtime of each charging device to be repaired, and determine whether the downtime of the charging device to be repaired meets the constraints. If it meets the constraints, adjust the maintenance start time. The adjustment formula for the maintenance start time is as follows:
[0129] t s，i = max((min(C i (1), C i (1),..., C i (Q)))-(s i - ST i ));
[0130] where t s，i represents the maintenance start time of the i-th charging device to be repaired, min represents the minimization function, and C i (Q) represents the maintenance cost of the Q-th time window of the i-th charging device to be repaired;
[0131] E3. Calculate the total maintenance cost of all charging devices to be repaired, continuously adjust the initial scheduling of the charging devices to be repaired, generate a new maintenance plan, and calculate the total cost;
[0132] E4. Compare the total maintenance costs corresponding to all maintenance plans to obtain the optimal maintenance plan.
[0133] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made without departing from the spirit of the present invention within the knowledge of those skilled in the art.
Claims
1. A remote monitoring method for a charging device, characterized in that, Including: Collecting signals of each charging device in the target area based on a plurality of preset vibration sensors to obtain a plurality of vibration signal sequences of the charging device, wherein the vibration signal is composed of a plurality of vibration data points, and collecting signals of each charging device in the target area based on a current sensor to obtain the current signal of the charging device; Calculating the action coefficient of the vibration data points in the vibration signal sequence, and fusing the plurality of vibration signal sequences based on the action coefficient to obtain a fused vibration signal; Performing Hilbert transform on the current signal and the fused vibration signal to obtain a first envelope curve of the current signal and a second envelope curve of the fused vibration signal; Performing fault diagnosis on the first envelope curve and the second envelope curve based on a pre-trained fault diagnosis model to obtain a fault level label of the charging device; Constructing a maintenance optimization model for the maintenance charging devices in the target area based on the fault level label of the charging device; Solving the maintenance optimization model based on the greedy algorithm to obtain an optimal maintenance plan for the charging devices in the target area.
2. The remote monitoring method of the charging device according to claim 1, characterized in that, Calculating the action coefficient of the vibration data points in the vibration signal sequence includes: Calculating the total energy of each vibration signal in the vibration signal sequence, wherein the calculation formula of the total energy of the vibration signal is as follows: Among them, E i represents the total energy of the i-th vibration signal in the vibration signal sequence, n represents the number of vibration data points in the vibration signal, and s iq represents the q-th vibration data point in the i-th vibration signal in the vibration signal sequence; Correcting the vibration data points of the vibration signal based on the total energy of the vibration signal to obtain corrected vibration data points, wherein the correction formula is as follows: Among them, represents the q-th corrected vibration data point in the i-th vibration signal in the vibration signal sequence; Calculating the action coefficient of the vibration data points in the vibration signal based on the corrected vibration data points, wherein the calculation formula of the action coefficient is as follows: Among them, Eff iq represents the action coefficient of the q-th vibration data point in the i-th vibration signal in the vibration signal sequence.
3. The remote monitoring method of the charging device according to claim 2, characterized in that, Fusing the plurality of vibration signal sequences based on the action coefficient to obtain a fused vibration signal includes: Calculating the fusion weight of the vibration data points in the plurality of vibration signals based on the action coefficient of the vibration data points in the vibration signal, wherein the calculation formula of the fusion weight is as follows: where ω iq represents the fusion weight of the q-th vibration data point in the i-th vibration signal in the vibration signal sequence; Fusing each vibration data point in the plurality of vibration signals based on the fusion weight of the vibration data points in the plurality of vibration signal sequences to obtain fused vibration data points, and obtaining the fused vibration signal based on the fused vibration data points, wherein the fusion formula of the fused vibration data points is as follows: Among them, S q represents the q-th fused vibration data point in the fused vibration signal, and I represents the number of vibration signal sequences.
4. The remote monitoring method of the charging device according to claim 1, characterized in that, Performing Hilbert transform on the current signal and the fused vibration signal to obtain a first envelope curve of the current signal and a second envelope curve of the fused vibration signal includes: Performing Hilbert transform on the current signal to obtain a first transformed signal, wherein the expression of the first transformed signal is as follows: Among them, represents the first transformed signal of the current signal x1(t); Calculating the first envelope curve of the current signal based on the first transformed signal, wherein the expression of the first envelope curve is as follows: Wherein, c1(t) represents the first envelope curve; Performing Hilbert transform on the fused vibration signal to obtain a second transformed signal, wherein the expression of the second transformed signal is as follows: Among them, represents the second transformed signal of the fused vibration signal x2(t); Calculating the second envelope curve of the fused vibration signal based on the second transformed signal, wherein the expression of the second envelope curve is as follows: Among them, c2(t) represents the second envelope curve.
5. The remote monitoring method of the charging device according to claim 4, characterized in that, The fault diagnosis model includes a first feature extraction module, a second feature extraction module, a feature connection module, and a fault classification module. The first feature extraction module is used to perform convolution on the first envelope curve to obtain a first feature vector, perform a max pooling operation on the first feature vector to obtain a first dimensionality-reduced feature vector, perform batch normalization processing on the first dimensionality-reduced feature vector to obtain a first normalized feature vector, estimate the feature importance of the first normalized feature vector through global average pooling and two fully connected layers to obtain a first evaluation vector. Among them, the two fully connected layers respectively adopt ReLU and Softmax activation functions, perform a vector dot product on the first normalized feature vector and the first evaluation vector to obtain a comprehensive feature. The second feature extraction module is used to perform convolution on the second envelope curve to obtain a second feature vector, perform a max pooling operation on the second feature vector to obtain a second dimensionality-reduced feature vector, and perform batch normalization processing on the second dimensionality-reduced feature vector to obtain a second normalized feature vector. The feature connection module is used to perform a connection operation on the comprehensive feature and the second normalized feature vector to obtain a connection vector. The fault classification module is used to perform fault classification on the connection vector through the Softmax activation function to obtain a fault probability, and obtain the fault level label of the charging device based on the fault probability.
6. The remote monitoring method of the charging device according to claim 5, characterized in that, The calculation formula of the first normalized feature vector is as follows: Among them, represents the first normalized eigenvector, BN represents the batch normalization processing function, MP represents the max pooling operation function, and f represents a preset activation function. and represent the first neuron bias of the convolution and the kernel of the first neuron, and cov1D represents the convolution pooling function. The calculation formula of the second normalized feature vector is as follows: Among them, represents the second normalized eigenvector, and represent the second neuron bias and the kernel of the second neuron for convolution; The calculation formula of the connection vector is as follows: Among them, represents the connection vector, FC represents the fully connected layer, and GA represents the global average pooling function.
7. The remote monitoring method of the charging device according to claim 1, characterized in that Construct a maintenance optimization model for the repaired charging devices in the target area based on the fault level labels of the charging devices, including: Construct an objective function based on the maintenance cost corresponding to each fault level label of the repaired charging devices. Among them, the expression of the objective function is as follows: Among them, MIN:L represents the objective function, θ i,j represents the fault level of the charging device, C i，j represents the cost required for maintenance with a specific priority, d i,j represents whether the implementation of the corresponding repair work requires downtime. If d i,j = 1, it means that the implementation of the repair measure requires downtime. If d i,j = 0, it means that the implementation of the repair measure does not require downtime, C D represents the fixed startup cost after the charging device is shut down, c e represents the opportunity cost of downtime per unit time, c o represents the loss cost of downtime per unit time, s i represents the downtime of the corresponding charging device, c i，j，s represents the unit fixed setting cost of the charging device under the corresponding repair work, C i，j，L represents the loss cost caused by the incomplete utilization of the charging device. I represents the total number of charging devices to be repaired, and J represents the total number of types of fault level labels; Construct a constraint condition based on the maximum charging device downtime. Among them, the expression of the first constraint condition is as follows: s i ≤ ST i ; Among them, ST i represents the maximum charging device downtime; Construct a maintenance optimization model based on the objective function and the constraint condition.
8. The remote monitoring method of the charging device according to claim 7, characterized in that, Solve the maintenance optimization model based on the greedy algorithm to obtain the optimal maintenance plan for the charging devices in the target area, including: Randomly select a charging device to be repaired, calculate the maintenance cost of the charging device to be repaired within the time window, find the time window with the minimum maintenance cost, and set the latest time point within the time window as the maintenance start time of the charging device to be repaired; Calculate the downtime of each charging device to be repaired, and judge whether the downtime of the charging device to be repaired meets the constraint condition. If it meets the constraint condition, adjust the maintenance start time. Among them, the adjustment formula of the maintenance start time is as follows: t s，i = max((min(C i (1), C i (1),..., C i (Q)))-(s i - ST i )); where, t s，i represents the repair start time of the i-th charging device to be repaired, min represents the minimization function, C i (Q) represents the repair cost of the Q-th time window of the i-th charging device to be repaired; Calculate the total maintenance cost of all charging devices to be repaired, continuously adjust the initial scheduling of the charging devices to be repaired, generate a new maintenance plan, and calculate the total cost; Compare the total maintenance costs corresponding to all maintenance plans to obtain the optimal maintenance plan.