Remote interaction and intelligent operation and maintenance system and method for charging pile of electric two-wheeled vehicle

Abstract

The invention belongs to the technical field of electric vehicle charging management, and particularly relates to a remote interaction and intelligent operation and maintenance system and method for an electric two-wheeled vehicle charging pile, a charging pile terminal, an edge computing device, a cloud server and a user interaction terminal. The edge computing device is used for operating the LSTM neural network model to perform real-time anomaly detection on the frequency domain feature vector; the edge computing device further comprises a feature extraction unit and a data encryption module. The feature extraction unit generates a frequency domain feature vector from the current spectrum data collected by the charging pile terminal through fast Fourier transform; according to the invention, a feature extraction unit in an edge computing device uses fast Fourier transform to generate a frequency domain feature vector from current spectrum data collected by a charging pile terminal, and then an LSTM neural network model is operated to carry out real-time anomaly detection on the frequency domain feature vector. Therefore, the real-time abnormity monitoring of the current during the charging of the two-wheeled electric vehicle is realized, and the safety of the two-wheeled electric vehicle during the charging is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of electric vehicle charging management, and specifically relates to a remote interaction and intelligent operation and maintenance system and method for an electric two-wheeled vehicle charging pile. Background Art

[0002] At present, the smart charging pile system for electric two-wheeled vehicles generally adopts an independently operated localized control architecture. Its current monitoring module only has basic waveform acquisition functions and is limited by the following technical defects: First, the charging pile body is not equipped with a spectrum feature extraction unit and a wireless communication module, resulting in the inability to upload the time-frequency domain characteristic parameters such as the non-steady-state AC ripple and harmonic distortion rate generated during the charging process to the cloud analysis platform in real time. This data link break makes the battery equivalent series resistance (ESR) prediction model based on deep learning only rely on historical data for offline training, and cannot realize the online correlation analysis of charging conditions and polarization impedance, which ultimately leads to the technical bottleneck that the error rate of battery health status (SOH) assessment has been maintained at 12%-15% for a long time.

[0003] Secondly, the discrete threshold trigger alarm mechanism used in existing charging piles has significant limitations. Its preset fixed current threshold cannot effectively capture the waveform envelope distortion caused by contact impedance degradation during dynamic charging, and it is difficult to identify intermittent pulse spikes caused by imbalance of battery pack cells. Experimental data show that when the contact resistance exceeds 30mΩ due to contactor oxidation, the traditional alarm system has a miss rate of up to 42.7% for abnormal waveforms. What is more serious is that in the constant current-constant voltage conversion stage, the existing system lacks the real-time analysis capability of the current differential eigenvector, and cannot distinguish between the normal third-order Bessel filtering process and the real poor contact fault, which directly causes the overcharge warning delay to an average of 8.5 seconds, significantly increasing the risk of thermal runaway.

[0004] Therefore, the existing electric two-wheeled vehicle charging station system adopts an independent operation mode and is unable to upload current spectrum characteristics in real time. In addition, the traditional threshold alarm mechanism is also unable to identify dynamically changing current waveform characteristics, which leads to significant limitations. Summary of the invention

[0005] In order to make up for the deficiencies of the prior art, the present invention proposes a remote interaction and intelligent operation and maintenance system and method for an electric two-wheeled vehicle charging pile. The present invention is mainly used to solve the problem that the existing electric two-wheeled vehicle charging pile system adopts an independent operation mode and cannot upload current spectrum characteristics in real time, and the traditional threshold alarm mechanism cannot identify the dynamically changing current waveform characteristics, thus having significant limitations.

[0006] The technical solution adopted by the present invention to solve its technical problems is as follows: The present invention provides a remote interaction and intelligent operation and maintenance system for an electric two-wheeler charging pile, including: a charging pile terminal, an edge computing device, a cloud server, and a user interaction terminal; The charging pile terminal collects charging current spectrum data during the charging process through a Hall sensor; The edge computing device is communicatively connected to the charging pile terminal and is used to run an LSTM neural network model to perform real-time anomaly detection on the frequency domain feature vector and generate an anomaly detection result; specifically, it can be connected to the charging pile terminal through the LoRaWAN protocol; The cloud server is used to store a fault knowledge base and generate operation and maintenance instructions; the fault knowledge base is a decision tree model containing more than 20,000 fault cases, and can thus generate operation and maintenance instructions according to the anomaly detection result; The user interaction terminal is used to receive and visually display the operation and maintenance instructions; Among them, the edge computing device further includes a feature extraction unit and a data encryption module; The feature extraction unit generates a frequency domain feature vector from the current spectrum data collected by the charging pile terminal through fast Fourier transform; The data encryption module encrypts the frequency domain feature vector using the SM4 algorithm.

[0007] The data processing logic steps of the system are as follows: Step S1: The Hall sensor of the charging pile terminal captures charging current waveform data at a sampling rate of 10 kHz; Step S2: The feature extraction unit in the edge computing device calls the FFT algorithm to perform fast Fourier transform on the original waveform to generate a 128-dimensional frequency domain feature vector; Step S3: The environment perception module synchronously collects temperature, humidity, and smoke concentration data to generate an equipment operation environment index; Step S4: The edge computing device performs data obfuscation processing on the feature vector and uses a dynamic key seed to call the SM4 encryption algorithm to generate an encrypted data packet; Step S5: Upload {{device ID, timestamp, encrypted data}} to the edge gateway through the LoRaWAN communication protocol; Step S6: The edge computing device receives the encrypted data packet and performs SM4 decryption operation using the preset key seed; Step S7: Perform Z-score standardization processing on the frequency domain feature vector to generate a standardized input matrix; Step S8: The edge computing device inputs the standardized matrix into a pre-trained LSTM neural network model to generate an anomaly detection result.

[0008] Preferably, the real-time anomaly detection of the frequency-domain feature vector includes: Training an LSTM neural network model; The input layer receives the normalized frequency-domain feature vector; specifically, the frequency-domain feature vector is processed by Z-score normalization to generate a normalized input matrix; The normalized matrix is input into the pre-trained LSTM neural network model, and according to the LSTM neural network model, the output layer generates a prediction error rate; Condition for triggering a warning: When the error rate > 15% for three consecutive sampling periods, an alarm is triggered. Specifically, a JSON alarm packet containing {{fault code, confidence level, recommended measures}} can be generated.

[0009] Preferably, the training method of the LSTM neural network model includes: Construct a training data set containing normal / anomalous current waveforms, and use the sliding window method to generate time series samples; Dynamically allocate spectral feature weights through the attention mechanism; apply a weight of less than 0.3 to suppress the high-frequency noise features of 500 - 800 Hz; apply a weight of more than 0.7 to strengthen the effective features of 100 - 300 Hz; Set an early stopping mechanism, and terminate the training when the loss function of the validation set drops by less than 1% for five consecutive epochs.

[0010] Preferably, the cloud server is also used to train a demand prediction model based on historical charging data, and optimize the power distribution strategy of the regional charging pile group using the NSGA-II algorithm.

[0011] The demand prediction model: Input: {{historical charging volume, weather data, user behavior characteristics}}; Output: A heat map of the demand distribution for the next 2 hours.

[0012] Preferably, the parameter configuration of the NSGA-II algorithm includes: Set the population size to 3 times the number of charging piles; Adopt an adaptive crossover probability adjustment mechanism, with a range of 0.7 - 0.9; Introduce an elite retention strategy, and retain the top 10% of the optimal solutions for the next generation of iteration.

[0013] Preferably, the working process of the data encryption module includes: Before encrypting with the SM4 algorithm, perform data obfuscation processing on the frequency-domain feature vector; Adopt a dynamic key distribution mechanism to update the encryption key seed every 24 hours; the dynamic key distribution mechanism can be an elliptic curve encryption algorithm; Add metadata fields of {{checksum, protocol version, encryption flag}} to the data packet header.

[0014] Preferably, the user interaction terminal further includes a charging reservation module, an emergency instruction channel, and a charging policy generation unit; The charging reservation module generates a recommended optimal charging time period based on historical charging data analysis; The emergency instruction channel triggers an audible and visual alarm when a thermal runaway risk is detected; The charging policy generation unit dynamically adjusts the tiered electricity price coefficient according to the power grid load status.

[0015] Among them, the charging policy generation unit calculates the real-time electricity price through the following formula: P_t = P_base × (1 + α·(L_t / L_max)) Where P_base is the base electricity price, that is, the initial electricity price when the power grid load is in the reference state, α is the adjustment factor (0.2 - 0.5), L_t is the current regional power grid load value, and L_max is the power grid carrying threshold.

[0016] P_base can be trained based on historical data through the demand prediction model, thereby optimizing the setting of the initial value of P_base; and by displaying the comparison between P_base and the real-time electricity price to the user through the user interaction terminal, thereby guiding the charging behavior, and then the user can reduce the cost by choosing to charge during low-load periods, and at the same time can effectively relieve the power grid pressure.

[0017] Preferably, the user interaction terminal optimizes the interface interaction in the following ways: The charging status visualization uses HSV color space mapping, so that the overload warning color gamut covers wavelengths of 540 - 580 nm; The emergency interface enables a high-contrast mode, and the font size is not less than 18 pt; The voice control interface supports semantic parsing and execution of voice commands. For example, it supports semantic parsing of commands such as "pause charging" and "extend duration"; Preferably, the charging pile terminal further includes a mechanical locking mechanism, an anti-theft detection unit, and an anti-theft linkage module; The mechanical locking mechanism controls the tongue to extend and retract by an electromagnetic driving device; The anti-theft detection unit collects the three-dimensional displacement signal of the device through a vibration sensor and generates a smooth displacement trajectory through Kalman filtering; When unauthorized movement is detected, an audible and visual alarm is triggered, and the device ID and displacement trajectory are uploaded to the cloud server; The anti-theft linkage module compares displacement trajectories. When the similarity between the displacement trajectory and the theft case database is greater than 85%, it pushes a warning to the operation and maintenance terminal within 5 kilometers.

[0018] The displacement trajectory comparison uses the Hausdorff distance algorithm to compare the similarity between the currently obtained displacement trajectory and the historical theft case database.

[0019] A remote interaction and intelligent operation and maintenance method for an electric two-wheeler charging pile includes the following steps: S1: The Hall sensor of the charging pile terminal captures charging current waveform data at a sampling rate of 10 kHz; S2: The feature extraction unit in the edge computing device calls the FFT algorithm to perform a fast Fourier transform on the original waveform and generates a 128-dimensional frequency domain feature vector; S3: The environmental perception module synchronously collects temperature, humidity, and smoke concentration data and generates an equipment operation environment index; S4: The edge computing device performs data confusion processing on the feature vector and uses a dynamic key seed to call the SM4 encryption algorithm to generate an encrypted data packet; S5: Uploads {{device ID, timestamp, encrypted data}} to the edge gateway through the LoRaWAN communication protocol; S6: The edge computing device receives the encrypted data packet and performs an SM4 decryption operation using the preset key seed; S7: Performs Z-score normalization processing on the frequency domain feature vector to generate a normalized input matrix; S8: The edge computing device inputs the normalized input matrix into a pre-trained LSTM neural network model; the LSTM neural network model dynamically assigns frequency domain feature weights through an attention mechanism: applies a weight of less than 0.3 to suppress high-frequency noise features of 500 - 800 Hz; applies a weight of more than 0.7 to strengthen effective features of 100 - 300 Hz; S9: The edge computing device calculates the prediction error rate between the output value of the LSTM neural network model and the actual value. When the error rate exceeds 15% for three consecutive cycles, a third-level warning signal is triggered; S10: The cloud server first trains a demand prediction model based on historical charging data and predicts the regional charging demand for the next 2 hours; S11: The cloud server then uses the NSGA-II algorithm to optimize the power distribution strategy of the regional charging pile group; S12: The cloud server generates a scheduling instruction set containing {{device ID, power limit, time window}} and issues it to the edge gateway.

[0020] The beneficial effects of the present invention are as follows: 1. In the present invention, the feature extraction unit in the edge computing device uses the fast Fourier transform to generate a frequency-domain feature vector from the current spectrum data collected by the charging pile terminal, and then runs the LSTM neural network model to perform real-time anomaly detection on the frequency-domain feature vector, thereby realizing real-time anomaly monitoring of the current during the charging of two-wheeled electric vehicles, and improving the safety of two-wheeled electric vehicles during charging. Moreover, by storing the decision tree model of more than 20,000 historical fault cases in the cloud server, it can match the frequency-domain feature vector and quickly output a repair plan, thereby shortening the operation and maintenance response time from an average of 4.2 hours to 18 minutes.

[0021] 2. In the LSTM neural network model of the present invention, the attention mechanism is used to dynamically allocate the spectrum feature weights, applying a weight of less than 0.3 to suppress the high-frequency noise features of 500 - 800 Hz, and applying a weight of more than 0.7 to strengthen the effective features of 100 - 300 Hz. As a result, the fault prediction accuracy rate of the LSTM neural network model is as high as 92.3%, greatly improving the real-time anomaly monitoring ability of two-wheeled electric vehicles during charging, and further ensuring the safety of two-wheeled electric vehicles during charging.

[0022] 3. In the present invention, the cloud server trains a demand prediction model based on historical charging data and uses the NSGA-II algorithm to optimize the power distribution strategy of the regional charging pile group. Specifically, by introducing an elite retention strategy, the top 10% of the optimal solutions are retained for the next generation of iteration, so that the charging energy in this region can be fully utilized, and the regional energy utilization rate is increased by more than 40%.

[0023] 4. In the present invention, the anti-theft linkage module uses the Hausdorff distance algorithm to compare the similarity between the currently obtained displacement trajectory and the historical theft case library. When the similarity between the displacement trajectory and the theft case library is greater than 85%, a warning is pushed to the operation and maintenance terminal within 5 kilometers, further shortening the operation and maintenance response time. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The present invention will be further described below with reference to the accompanying drawings.

[0025] Figure 1 is the overall framework diagram of the system of the present invention; Figure 2 is the data flow topology diagram of the system of the present invention; Figure 3 is the data processing flow chart of the system of the present invention; Figure 4 is the flow chart of LSTM model training in the present invention; Figure 5 is the data encryption flow chart shown from the perspective of each terminal in the present invention; Figure 6It is the data encryption flowchart shown from the perspective of the data end in the present invention; Figure 7 It is the flowchart of the power optimization algorithm in the present invention; Figure 8 It is the user interface diagram in the present invention; Figure 9 It is the timing diagram of anti-theft detection in the present invention. Detailed implementation manners

[0026] In order to make the technical means, creative features, achieved purposes and effects realized by the present invention easy to understand, the present invention will be further described below in conjunction with specific implementation manners.

[0027] As Figures 1 to 3 shown, a remote interaction and intelligent operation and maintenance system for an electric two-wheeler charging pile includes: a charging pile terminal, an edge computing device, a cloud server and a user interaction terminal; The charging pile terminal collects charging current spectrum data during the charging process through a Hall sensor; The edge computing device is communicatively connected to the charging pile terminal and is used to run an LSTM neural network model to perform real-time anomaly detection on the frequency domain feature vector and generate an anomaly detection result; specifically, it can be connected to the charging pile terminal through the LoRaWAN protocol; The cloud server is used to store a fault knowledge base and generate operation and maintenance instructions; the fault knowledge base is a decision tree model containing more than 20,000 fault cases, and can further generate operation and maintenance instructions according to the anomaly detection result; The user interaction terminal is used to receive and visually display the operation and maintenance instructions; Among them, the edge computing device further includes a feature extraction unit and a data encryption module; The feature extraction unit generates a frequency domain feature vector by performing a fast Fourier transform on the current spectrum data collected by the charging pile terminal through the fast Fourier transform; The data encryption module encrypts the frequency domain feature vector using the SM4 algorithm.

[0028] The data processing logic steps of the system are as follows: Step S1: The Hall sensor of the charging pile terminal captures charging current waveform data at a sampling rate of 10 kHz; Step S2: The feature extraction unit in the edge computing device calls the FFT algorithm to perform a fast Fourier transform on the original waveform to generate a 128-dimensional frequency domain feature vector; Step S3: The environment perception module synchronously collects temperature, humidity and smoke concentration data to generate an equipment operation environment index; Step S4: The edge computing device performs data obfuscation processing on the feature vector, and uses a dynamic key seed to call the SM4 encryption algorithm to generate an encrypted data packet; Step S5: Upload {{device ID, timestamp, encrypted data}} to the edge gateway through the LoRaWAN communication protocol; Step S6: The edge computing device receives the encrypted data packet and performs SM4 decryption operation using the preset key seed; Step S7: Perform Z-score normalization processing on the frequency domain feature vector to generate a normalized input matrix; Step S8: The edge computing device inputs the normalized matrix into the pre-trained LSTM neural network model to generate an anomaly detection result.

[0029] As Figures 1 to 3 shown, the real-time anomaly detection of the frequency domain feature vector includes: Training the LSTM neural network model; The input layer receives the normalized frequency domain feature vector; specifically, perform Z-score normalization processing on the frequency domain feature vector to generate a normalized input matrix; Input the normalized matrix into the pre-trained LSTM neural network model, and according to the LSTM neural network model, the output layer generates a prediction error rate; Condition for triggering a warning: When the error rate > 15% for three consecutive sampling periods, a warning is triggered. Specifically, a JSON warning packet containing {{fault code, confidence level, recommended measures}} can be generated.

[0030] As Figure 4 shown, the training method of the LSTM neural network model includes: Construct a training data set containing normal / anomalous current waveforms, and use the sliding window method to generate time series samples; Dynamically allocate spectrum feature weights through the attention mechanism; apply a weight below 0.3 to suppress the high-frequency noise features of 500 - 800 Hz; apply a weight above 0.7 to strengthen the effective features of 100 - 300 Hz; Set an early stopping mechanism, and terminate the training when the decrease amplitude of the validation set loss function is less than 1% for five consecutive epochs.

[0031] As Figure 7 shown, the cloud server is also used to train a demand prediction model based on historical charging data, and optimize the power allocation strategy of the regional charging pile group using the NSGA-II algorithm.

[0032] The demand prediction model: Input: {{historical charging volume, weather data, user behavior characteristics}}; Output: Heat map of demand distribution in the next 2 hours.

[0033] As Figure 7 shown, the parameter configuration of the NSGA-II algorithm includes: Set the population size to 3 times the number of charging piles; Adopt an adaptive crossover probability adjustment mechanism with a range of 0.7 - 0.9; Introduce an elite retention strategy to retain the top 10% of the optimal solutions for the next generation of iteration.

[0034] As Figures 5 to 6 shown, the working process of the data encryption module includes: Before encrypting with the SM4 algorithm, perform data confusion processing on the frequency domain feature vector; Adopt a dynamic key distribution mechanism to update the encryption key seed every 24 hours; the dynamic key distribution mechanism can be an elliptic curve encryption algorithm; Add metadata fields of {{check code, protocol version, encryption identifier}} to the packet header.

[0035] As Figure 1 shown, the user interaction terminal further includes a charging reservation module, an emergency instruction channel, and a billing policy generation unit; The charging reservation module generates a recommended optimal charging time period based on historical charging data analysis; The emergency instruction channel triggers an audible and visual alarm when a thermal runaway risk is detected; The billing policy generation unit dynamically adjusts the ladder electricity price coefficient according to the grid load status.

[0036] Among them, the billing policy generation unit calculates the real-time electricity price through the following formula: P_t = P_base × (1 + α·(L_t / L_max)) Where P_base is the base electricity price, that is, the initial electricity price when the grid load is in the benchmark state, α is the adjustment factor (0.2 - 0.5), L_t is the current regional grid load value, and L_max is the grid carrying threshold.

[0037] P_base can be trained based on historical data through the demand prediction model, thereby optimizing the initial value setting of P_base; and by showing the comparison between P_base and the real-time electricity price to users through the user interaction terminal, thereby guiding the charging behavior, and then users can reduce costs by choosing to charge during low-load periods, and at the same time can effectively relieve the grid pressure.

[0038] As Figure 1 shown, the user interaction terminal optimizes the interface interaction in the following ways: The charging status visualization uses HSV color space mapping, enabling the overload warning color gamut to cover wavelengths of 540 - 580 nm; The emergency interface enables a high - contrast mode with a font size not less than 18 pt; The voice control interface supports semantic parsing and execution of voice commands. For example, it supports semantic parsing of commands such as "pause charging" and "extend duration"; As Figure 9 As shown, the charging pile terminal further includes a mechanical locking mechanism, an anti - theft detection unit, and an anti - theft linkage module; The mechanical locking mechanism controls the retraction and extension of the locking tongue by an electromagnetic driving device; The anti - theft detection unit collects three - dimensional displacement signals of the device through a vibration sensor and generates a smooth displacement trajectory through Kalman filtering; When unauthorized movement is detected, an audible and visual alarm is triggered, and the device ID and displacement trajectory are uploaded to the cloud server; The anti - theft linkage module compares displacement trajectories. When the similarity between the displacement trajectory and the theft case library is greater than 85%, a warning is pushed to the operation and maintenance terminal within 5 kilometers.

[0039] The displacement trajectory comparison uses the Hausdorff distance algorithm to compare the similarity between the currently obtained displacement trajectory and the historical theft case library.

[0040] As Figures 1 to 9 As shown, a remote interaction and intelligent operation and maintenance method for an electric two - wheeler charging pile includes the following steps: S1: The Hall sensor of the charging pile terminal captures charging current waveform data at a sampling rate of 10 kHz; S2: The feature extraction unit in the edge computing device calls the FFT algorithm to perform a fast Fourier transform on the original waveform, generating a 128 - dimensional frequency - domain feature vector; S3: The environment perception module synchronously collects temperature, humidity, and smoke concentration data to generate an equipment operation environment index; S4: The edge computing device performs data confusion processing on the feature vector and uses a dynamic key seed to call the SM4 encryption algorithm to generate an encrypted data packet; S5: Upload {{device ID, timestamp, encrypted data}} to the edge gateway through the LoRaWAN communication protocol; S6: The edge computing device receives the encrypted data packet and performs an SM4 decryption operation using a preset key seed; S7: Perform Z - score normalization processing on the frequency - domain feature vector to generate a normalized input matrix; S8: The edge computing device inputs the standardized input matrix into the pre-trained LSTM neural network model; the LSTM neural network model dynamically allocates frequency-domain feature weights through the attention mechanism: applies a weight below 0.3 to suppress the high-frequency noise features of 500 - 800 Hz; applies a weight above 0.7 to strengthen the effective features of 100 - 300 Hz. S9: The edge computing device calculates the prediction error rate between the output value of the LSTM neural network model and the actual value. When the error rate exceeds 15% for three consecutive cycles, a level-three alarm signal is triggered. S10: The cloud server first trains a demand prediction model based on historical charging data to predict the regional charging demand for the next 2 hours. S11: The cloud server then optimizes the power distribution strategy of the regional charging pile group using the NSGA-II algorithm. S12: The cloud server generates a scheduling instruction set containing {{device ID, power upper limit, time window}} and sends it to the edge gateway.

[0041] Embodiment 1: Abnormal current spectrum detection Specific implementation steps: Data collection: Configure an HX10-NP type Hall sensor at the charging pile terminal to collect charging current waveform data at a sampling rate of 10 kHz, and synchronously obtain environmental parameters such as temperature (accuracy of ±0.5°C) and smoke concentration (range of 0 - 20% LEL). Feature extraction: The edge device uses an STM32F407 chip to perform a 1024-point FFT operation to generate a 128-dimensional feature vector containing the fundamental component (50 Hz) and 3rd / 5th / 7th harmonics (150 / 250 / 350 Hz). Encrypted transmission: Dynamically scramble the feature vector (insert random noise of ±5%), encrypt it using the SM4-CTR mode, and update the key seed every 24 hours through the ECDH protocol. Abnormal determination: The LSTM model sets the sliding window length to 30 seconds. When the prediction error > 15% for three consecutive cycles, a JSON alarm packet containing a fault location code (e.g., E102 represents an overcurrent fault) is generated.

[0042] The following technical effects are obtained from Embodiment 1: Compared with the traditional threshold detection method, the false alarm rate is reduced by 62% (measured data), and the high-frequency noise suppression effect is improved by 40%.

[0043] Embodiment 2: Dynamic key Specific implementation plan: Key management: Adopt a two-level key architecture. The main key is stored in the HSM encryption machine through the PKI system, and the session key is dynamically generated through the ECIES algorithm.

[0044] Data packet structure: { "header": { "protocol_ver": "1.2", "checksum": "9A3F", "key_index": 0x1B }, "payload": "7A2E...D9F4" / / Feature vector after SM4 encryption } Anti-theft linkage: When the displacement speed detected by the triaxial accelerometer > 0.5 m / s², activate the trajectory comparison algorithm and calculate the similarity with the theft case database using the Hausdorff distance.

[0045] The following measured data are obtained from Example 2: In the simulated attack test, 99.3% of replay attacks and man-in-the-middle attacks are successfully blocked.

[0046] Example 3: Grid collaborative optimization Specific implementation process: Demand forecasting: Input historical charging amounts (time series decomposition), weather data (temperature / humidity / wind speed), and user behavior (charging duration preference) to train the TCN time series model, and output a demand heat map at a 2-hour granularity; NSGA-II optimization algorithm: Objective function: min (grid load volatility) & max (charging pile utilization rate) Constraints: Single pile power ≤ 3 kW, total load ≤ 80% of the regional transformer capacity Dynamic charging: Generate a stepped electricity price coefficient according to the optimization results, and the value of α in the formula is dynamically adjusted (0.5 during peak hours, 0.3 during flat hours, 0.1 during valley hours) The following application effects are obtained from Example 3: In a pilot area (50 charging piles), the grid peak load is reduced by 27% and the average charging cost of users is decreased by 18%.

[0047] The embodiments of the present invention have been described above in conjunction with the accompanying drawings. However, the present invention is not limited to the above specific implementation manners. The above specific implementation manners are merely illustrative and not restrictive. Under the inspiration of the present invention, those of ordinary skill in the art can also make many forms without departing from the purpose of the present invention and the scope protected by the claims. All these are within the protection scope of the present invention.

Claims

1. A remote interaction and intelligent operation and maintenance system for an electric two-wheeled vehicle charging pile, characterized in that: include: Charging pile terminals, edge computing devices, cloud servers and user interaction terminals; The charging pile terminal collects charging current spectrum data during the charging process through a Hall sensor; The edge computing device is communicatively connected to the charging pile terminal, and is used to run an LSTM neural network model to perform real-time anomaly detection on the frequency domain feature vector; The cloud server is used to store the fault knowledge base and generate operation and maintenance instructions; The user interaction terminal is used to receive and visually display the operation and maintenance instructions; Wherein, the edge computing device also includes a feature extraction unit and a data encryption module; The feature extraction unit generates a frequency domain feature vector from the current spectrum data collected by the charging pile terminal through fast Fourier transform; The data encryption module uses the SM4 algorithm to encrypt the frequency domain feature vector.

2. The remote interaction and intelligent operation and maintenance system for an electric two-wheeled vehicle charging pile according to claim 1, characterized in that: The real-time anomaly detection of the frequency domain feature vector includes: Train the LSTM neural network model; The input layer receives the normalized frequency domain feature vector; According to the LSTM neural network model, the output layer generates the prediction error rate; Conditions for triggering a warning: The alarm is triggered when the error rate for three consecutive sampling periods is greater than 15%.

3. The remote interaction and intelligent operation and maintenance system for an electric two-wheeled vehicle charging pile according to claim 2 is characterized in that: The training method of the LSTM neural network model includes: Construct a training data set containing normal / abnormal current waveforms and use the sliding window method to generate time series samples; Dynamically allocate spectral feature weights through the attention mechanism; apply weight suppression below 0.3 to 500-800Hz high-frequency noise features; apply weight enhancement above 0.7 to 100-300Hz effective features; Set the early stopping mechanism to terminate the training when the validation set loss function decreases by less than 1% for 5 consecutive epochs.

4. The remote interaction and intelligent operation and maintenance system for an electric two-wheeled vehicle charging pile according to claim 1, characterized in that: The cloud server is also used to train a demand prediction model based on historical charging data, and use the NSGA-II algorithm to optimize the power allocation strategy of the regional charging pile group.

5. The remote interaction and intelligent operation and maintenance system for an electric two-wheeled vehicle charging pile according to claim 4 is characterized in that: The parameter configuration of the NSGA-II algorithm includes: Set the population size to 3 times the number of charging posts; Adopt adaptive crossover probability adjustment mechanism, ranging from 0.7-0.9; An elite retention strategy is introduced to retain the top 10% optimal solutions for the next generation of iterations.

6. The remote interaction and intelligent operation and maintenance system for an electric two-wheeled vehicle charging pile according to claim 1, characterized in that: The workflow of the data encryption module includes: Before encryption by the SM4 algorithm, data obfuscation processing is performed on the frequency domain feature vector; Adopt a dynamic key distribution mechanism to update the encryption key seed every 24 hours; Add {{checksum, protocol version, encryption identifier}} metadata field to the data packet header.

7. The remote interaction and intelligent operation and maintenance system for an electric two-wheeled vehicle charging pile according to claim 1, characterized in that: The user interaction terminal also includes a charging reservation module, an emergency instruction channel and a charging strategy generation unit; The charging reservation module generates an optimal charging time period recommendation based on historical charging data analysis; The emergency command channel triggers an audible and visual alarm when a thermal runaway risk is detected; The billing strategy generation unit dynamically adjusts the tiered electricity price coefficient according to the power grid load status.

8. The remote interaction and intelligent operation and maintenance system for an electric two-wheeled vehicle charging pile according to claim 1, characterized in that: The user interaction terminal optimizes the interface interaction in the following ways: The charging status visualization uses HSV color space mapping to make the overload warning color range cover 540-580nm wavelength; Enable high contrast mode on the emergency interface; The voice control interface supports semantic parsing and execution of voice commands.

9. The remote interaction and intelligent operation and maintenance system for an electric two-wheeled vehicle charging pile according to claim 1, characterized in that: The charging pile terminal also includes a mechanical locking mechanism, an anti-theft detection unit and an anti-theft linkage module; The mechanical locking mechanism is controlled by an electromagnetic drive device to control the extension and retraction of the lock tongue; The anti-theft detection unit collects the three-dimensional displacement signal of the device through a vibration sensor and generates a smooth displacement trajectory through Kalman filtering; When unauthorized movement is detected, an audible and visual alarm is triggered, and the device ID and displacement trajectory are uploaded to the cloud server; The anti-theft linkage module compares the displacement trajectory. When the similarity between the displacement trajectory and the theft case library is greater than 85%, an early warning is pushed to the operation and maintenance terminal within 5 kilometers. The displacement trajectory comparison uses a Hausdorff distance algorithm to compare the similarity between the currently obtained displacement trajectory and a historical theft case library.

10. A remote interaction and intelligent operation and maintenance method for an electric two-wheeled vehicle charging pile, characterized in that: Applicable to any one of claims 1 to 9, comprising the steps of: S1: The Hall sensor of the charging pile terminal captures the charging current waveform data at a sampling rate of 10kHz; S2: The feature extraction unit in the edge computing device uses the FFT algorithm to perform fast Fourier transform on the original waveform to generate a 128-dimensional frequency domain feature vector; S3: The environmental perception module simultaneously collects temperature, humidity and smoke concentration data to generate the equipment operating environment index; S4: The edge computing device performs data obfuscation processing on the feature vector and uses the dynamic key seed to call the SM4 encryption algorithm to generate an encrypted data packet; S5: Upload {{device ID, timestamp, encrypted data}} to the edge gateway via the LoRaWAN communication protocol; S6: The edge computing device receives the encrypted data packet and performs SM4 decryption operation using the preset key seed; S7: Perform Z-score normalization on the frequency domain feature vector to generate a normalized input matrix; S8: The edge computing device inputs the standardized input matrix into the pre-trained LSTM neural network model; the LSTM neural network model dynamically allocates frequency domain feature weights through the attention mechanism: a weight suppression of less than 0.3 is applied to the 500-800Hz high-frequency noise features; a weight enhancement of more than 0.7 is applied to the 100-300Hz effective features; S9: The edge computing device calculates the prediction error rate between the output value of the LSTM neural network model and the actual value. When the error rate exceeds 15% for three consecutive cycles, a level 3 alarm signal is triggered; S10: The cloud server first trains the demand prediction model based on historical charging data to predict the regional charging demand in the next 2 hours; S11: The cloud server then uses the NSGA-II algorithm to optimize the power allocation strategy of the regional charging pile group; S12: The cloud server generates a scheduling instruction set including {{device ID, power limit, time window}} and sends it to the edge gateway.

Images