A new energy vehicle charging pile remote monitoring management control method and system

Abstract

This invention provides a remote monitoring, management, and control method and system for new energy vehicle charging piles. It achieves real-time monitoring of charging gun status, abnormal noise, and electrical parameters by deploying multimodal sensors. Edge computing and dynamic threshold cleaning reduce data redundancy and improve transmission efficiency. A self-attention mechanism and Drools rule engine are combined for multi-source data fusion and root cause analysis to dynamically assess risk levels. Dynamic power allocation based on health status and grid load optimizes resource utilization. Peak-valley dynamic pricing and an incentive mechanism guide off-peak charging. A hash chain ensures the immutability of maintenance data. Scheduling paths are optimized based on urgency and engineer resources. User feedback and extreme weather data iterative models enhance system robustness. This achieves intelligent management of the entire charging pile lifecycle, improving operation and maintenance efficiency, safety, and user charging experience.

Description

Technical Field

[0001] This invention relates to the field of new energy vehicle charging pile management technology, and in particular to a remote monitoring, management and control method and system for new energy vehicle charging piles. Background Technology

[0002] Existing charging pile monitoring systems mostly rely on single sensors for data collection, lacking multimodal data fusion capabilities. Data transmission often adopts a full-data reporting mode without lightweight processing and dynamic cleaning at the edge, resulting in high transmission bandwidth consumption and poor real-time performance.

[0003] Existing charging pile monitoring systems mostly rely on preset thresholds or fixed rules for fault diagnosis, lack multimodal data fusion and adaptive mechanisms, and risk assessment is mostly based on static indicators, without integrating environmental parameters and dynamic power grid load, making it difficult to achieve graded early warning.

[0004] Traditional dynamic power allocation schemes are mainly based on grid load or charging pile physical parameters, without considering the multi-dimensional constraints of equipment health status, energy storage capacity and real-time user demand. They lack a real-time dynamic adjustment mechanism for the available power of the grid, resulting in low resource utilization.

[0005] Current maintenance scheduling relies heavily on manual dispatching, without optimizing based on the urgency of the fault, the matching of engineer skills, and real-time task load. Maintenance records are mostly stored in a centralized manner, lacking immutable technologies such as hash chains to ensure data integrity. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for remote monitoring, management and control of new energy vehicle charging piles.

[0007] The problem this invention aims to solve is to address the multi-dimensional issues in real-time monitoring, fault early warning, resource optimization, and operation and maintenance management of new energy vehicle charging piles, and to resolve the problems faced by existing charging piles such as incomplete data collection, lagging fault detection, inefficient resource allocation, high user charging pressure during peak hours, slow maintenance response, and insufficient model adaptability.

[0008] A remote monitoring, management, and control method for new energy vehicle charging piles, employing the following technical solution:

[0009] S1: Deploy multimodal sensors, install cameras to monitor the charging gun status, integrate MEMS microphones to monitor abnormal noise, and deploy current, voltage, and temperature sensors to monitor the current, voltage, and temperature of the charging pile;

[0010] S2: Run a lightweight classification model on the edge gateway to detect the charging gun insertion status in real time. Only upload the charging gun feature vector, including the binary result of the charging gun insertion status. Use a GPS module to synchronize the clocks of all sensors, transmit data based on a compression algorithm, and perform dynamic threshold cleaning on the data.

[0011] S3: Real-time assessment of charging pile health status and fault location based on multi-source data, establishment of multi-modal feature fusion model for real-time assessment, construction of fault cause-effect graph based on Drools, and dynamic assessment of risk level;

[0012] S4: Based on real-time data and prediction models, adaptive control and resource optimization of charging piles are performed, and dynamic power allocation and load balancing are carried out.

[0013] S5: Dynamic pricing and guidance based on user behavior, increasing the unit price during peak charging hours to guide users to charge during off-peak hours;

[0014] S6: Establish a maintenance resource scheduling algorithm based on the urgency of the fault and the status of maintenance resources, dynamically generate scheduling paths, and use a hash chain to store the hash value of the data. The data digest includes the fault type, maintenance engineer ID, and maintenance time.

[0015] S7: Based on keywords in user reviews, such as slow charging and interface lag, we correlate charging pile performance parameters, regularly update model parameters, and add new data, including fault data under extreme weather conditions.

[0016] Furthermore, in step S1, a camera is installed to monitor the charging gun status, detect abnormal noise, and monitor the current, voltage, and temperature of the charging pile, including:

[0017] S11: The camera is installed on the top of the charging pile, facing the charging gun interface, with a field of view covering the area where the charging gun is inserted. It captures 1 frame of image per second, converts the image into a black and white image based on the Otsu algorithm, highlights the metal part of the charging gun, and eliminates noise through dilation and corrosion operations.

[0018] S12: An integrated MEMS microphone is installed around relays and fan components at a distance of 5-10cm to extract voiceprint features. ,in The original audio signal. Twelve coefficients were extracted for the Mel frequency cepstral coefficients.

[0019] Anomaly detection and labeling are performed based on the extracted voiceprint feature threshold. ,in This is the average of historical data. The standard deviation of historical data indicates that abnormal sound patterns include relay jamming and fan malfunction.

[0020] S13: A Hall effect sensor is used to monitor the current of the charging pile, with a sampling frequency of 1kHz; a voltage divider resistor network is used to monitor the voltage of the charging pile, with a sampling frequency of 1kHz; and a DS18B20 temperature sensor is used to monitor the temperature of the charging pile, with a sampling frequency of 1Hz.

[0021] Furthermore, in S2, the charging gun insertion status is detected in real time, only the feature vector is uploaded, the GPS module is used to synchronize the clocks of all sensors, and dynamic threshold cleaning is performed on the data, including:

[0022] S21: Establish an insertion state determination model to detect the charging gun insertion state in real time. ,in This represents the area of ​​the metal region of the charging gun when it is normally inserted.

[0023] S22: Only the feature vector of the charging gun is transmitted. The resistance value is calculated based on the current and voltage of the charging pile monitored as described in S1;

[0024] S23: Each sensor data packet is appended with a GPS timestamp, aligned with the GPS time server based on the NTP protocol, the time offset is calculated, and all sensor clocks are synchronized.

[0025] S24: Calculate the moving average and standard deviation of the sensor data, and set the adaptive cleaning threshold. , ,in This represents the average data traffic. The standard deviation of data traffic This is for the safety factor.

[0026] Furthermore, in S3, a multimodal feature fusion model is established for real-time evaluation, and a fault cause-effect graph is constructed based on Drools to perform dynamic risk level assessment, including:

[0027] S31: Perform weighted fusion of multi-source data. The multi-source data are the electrical data, visual data and environmental data mentioned in S1 and S2, respectively. The electrical data includes the current and voltage of the charging pile, the visual data includes the status of the charging gun, and the environmental data includes the temperature of the charging pile.

[0028] S32: Data feature interaction based on self-attention mechanism ,in For attention mechanism functions, For query, key, value vector, For function, As a dimension, output a health status score. Label the health status;

[0029] S33: Define rule templates, deduce fault root causes based on Drools rule chains, including abnormal insulation resistance → leakage protection triggering, and combine the output health status score. Combine with real-time data to update rule weights. ,in For the updated rule weights, As the initial rule weights, The attenuation coefficient is... Threshold for health status scoring;

[0030] S34: Calculate risk indicators and establish a risk value calculation formula. , ,in The environmental risk coefficient is calculated by weighting the charging pile temperature mentioned in S1 and the GPS module location information mentioned in S2. .

[0031] Furthermore, in S4, dynamic power allocation and load balancing are performed, including:

[0032] S41: Real-time health status assessment of each charging station based on the description in S3 Remaining capacity of energy storage equipment and the number of users currently charging The allocation weight of each charging station , ,in To prevent tiny constants with a denominator of zero;

[0033] S42: Allocate power to each charging station according to its weight. , ,in , For real-time load of the power grid, The upper limit of the power grid capacity is determined by dynamically adjusting the allocation ratio based on the available power in the power grid.

[0034] Furthermore, the dynamic pricing and guidance based on user behavior in S5 includes:

[0035] By combining real-time grid load, historical peak demand, and user behavior preferences, the charging unit price is dynamically adjusted, and users are guided to charge during off-peak hours through points rewards and priority enhancement mechanisms.

[0036] Establish a dynamic pricing model , ,in The benchmark electricity price, The load sensitivity coefficient for the power grid is set to 0.4. The user peak preference penalty coefficient is set to 0.3. For real-time load of the power grid, This is the upper limit of the power grid capacity. The percentage of charging during peak hours;

[0037] Charging during off-peak hours earns points, which are calculated based on the product of the charging amount and the dynamically priced electricity price. Charging stations are allocated to users based on their points during historical off-peak charging periods.

[0038] Furthermore, in the S6 state-generated scheduling path, the hash value of the data is stored using a hash chain, including:

[0039] S61: Combine the risk level definition in S3 to define the urgency score E of the fault. If the risk level is high, E=1.0; if the risk level is medium, E=0.8; if the risk level is low, E=0.5; otherwise, E=0.3.

[0040] S62: Combining skill matching Distance from the fault point Current number of tasks Define the availability score for maintenance engineers. , ;

[0041] S63: Prioritize assigning high-urgency faults to engineers. Initial solution = select from high to low by pressing E. Assign tasks from highest to lowest priority, randomly remove n low-priority tasks, and then assign tasks based on the remaining engineers. And E reassign tasks, fitness function ,in As weight, ,in Total scheduling time, To improve resource utilization, a tabu list is set to prevent duplicate solutions, and a maximum number of iterations is set.

[0042] S64: The data digest includes the fault type, maintenance engineer ID, and maintenance time. Each maintenance record generates a hash value and links it to the previous block to form an immutable chain. A digest is extracted from each maintenance record, and the digest = {fault type, maintenance engineer ID, maintenance time}. The hash value of the current block is generated using the SHA-256 hash function.

[0043] Furthermore, the performance parameters of the associated charging piles in S7 are periodically updated in the model parameters, including:

[0044] S71: Correlate keywords in user reviews, including slow charging and interface lag, with the physical parameters of the charging pile, and quantify the impact of user feedback on the performance of the charging pile based on NLP technology;

[0045] S72: Add new data, including fault data under extreme weather conditions, combined with user feedback, and update model parameters regularly with new data. Update model parameters with new data every 24 hours, and retain only fault data related to extreme weather.

[0046] Furthermore, a remote monitoring and management control system for new energy vehicle charging piles is provided to implement the remote monitoring and management control method for new energy vehicle charging piles described in any of the above-mentioned embodiments. The remote monitoring and management control system for new energy vehicle charging piles includes: a system management module, a charging pile management module, a charging pile repair management module, a user management module, a repairman management module, and a personal center module.

[0047] System management module: Deploys multimodal sensors, installs cameras to monitor charging gun status, integrates MEMS microphones to monitor abnormal noise, deploys current, voltage, and temperature sensors to monitor the current, voltage, and temperature of the charging pile, runs models on the edge gateway to detect the charging gun insertion status in real time, uploads only the binary results and feature vectors of the charging gun status, aligns all sensor clocks, eliminates time offset, and performs dynamic threshold cleaning.

[0048] Charging pile management module: Performs weight calculations to achieve dynamic adjustment based on the available power of the power grid, prioritizing the allocation of charging piles with high health status and sufficient remaining capacity, combined with the real-time load and capacity limit of the power grid;

[0049] Charging pile repair management module: It adopts a self-attention mechanism to integrate electrical data, visual data and environmental data to output a health status score, and dynamically adjusts the rule weights based on predefined rules to conduct risk level assessment;

[0050] User management module: Establishes a dynamic pricing model and points reward mechanism, integrating grid load, user preferences, and peak-hour penalties to guide user behavior;

[0051] Repairman Management Module: Establishes an urgency score to optimize resource utilization. Combines urgency with resource availability to achieve collaborative optimization of dynamic path planning and resource utilization. Hash chain stores data summaries, including fault type, repair engineer ID, and repair time.

[0052] The user center module extracts keywords from user reviews, quantifies the impact of user feedback on charging pile performance based on NLP technology, updates model parameters with new data including extreme weather fault data every 24 hours, and retains only data related to environmental risks.

[0053] The beneficial effects of this invention are: it enables accurate positioning of health status scores and root causes of faults, improves the accuracy and response speed of fault identification, and reduces unplanned downtime;

[0054] Dynamically allocate charging pile power, prioritize the efficient operation of equipment in high health status, avoid overload of faulty equipment, optimize the load balance of charging piles, alleviate grid pressure, guide users to charge during off-peak hours, reduce peak grid pressure, improve the utilization rate of charging piles during off-peak hours, and achieve dual optimization of resource utilization efficiency and user energy costs.

[0055] By employing an adaptive path planning algorithm, high-risk faults are prioritized and maintenance resources are dynamically allocated, shortening the average maintenance time. Hash chain technology ensures the integrity and traceability of data through the immutable storage of maintenance records. Model parameters are continuously optimized based on real user experience, and an incremental learning mechanism based on extreme weather fault data is added to enhance the system's adaptability to atypical scenarios. Through closed-loop optimization based on user feedback and environmental data, the system possesses self-evolution capabilities, adapting to the ever-changing operating environment and user needs. Attached Figure Description

[0056] Figure 1 A flowchart of a remote monitoring, management and control method for new energy vehicle charging piles;

[0057] Figure 2 This is a module diagram of a remote monitoring, management and control system for new energy vehicle charging piles. Detailed Implementation

[0058] The present invention will be further described clearly and completely below, but the scope of protection of the present invention is not limited thereto.

[0059] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article, unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship.

[0060] Example 1

[0061] A remote monitoring, management, and control method for new energy vehicle charging piles, employing the following technical solution:

[0062] S1: Deploy multimodal sensors, install cameras to monitor the charging gun status, integrate MEMS microphones to monitor abnormal noise, and deploy current, voltage, and temperature sensors to monitor the current, voltage, and temperature of the charging pile;

[0063] S2: Run a lightweight classification model on the edge gateway to detect the charging gun insertion status in real time. Only upload the charging gun feature vector, including the binary result of the charging gun insertion status. Use a GPS module to synchronize the clocks of all sensors, transmit data based on a compression algorithm, and perform dynamic threshold cleaning on the data.

[0064] S3: Real-time assessment of charging pile health status and fault location based on multi-source data, establishment of multi-modal feature fusion model for real-time assessment, construction of fault cause-effect graph based on Drools, and dynamic assessment of risk level;

[0065] S4: Based on real-time data and prediction models, adaptive control and resource optimization of charging piles are performed, and dynamic power allocation and load balancing are carried out.

[0066] S5: Dynamic pricing and guidance based on user behavior, increasing the unit price during peak charging hours to guide users to charge during off-peak hours;

[0067] S6: Establish a maintenance resource scheduling algorithm based on the urgency of the fault and the status of maintenance resources, dynamically generate scheduling paths, and use a hash chain to store the hash value of the data. The data digest includes the fault type, maintenance engineer ID, and maintenance time.

[0068] S7: Based on keywords in user reviews, such as slow charging and interface lag, we correlate charging pile performance parameters, regularly update model parameters, and add new data, including fault data under extreme weather conditions.

[0069] refer to Figure 1 The diagram shown is a flowchart of a remote monitoring, management and control method for new energy vehicle charging piles.

[0070] Furthermore, in step S1, a camera is installed to monitor the charging gun status, detect abnormal noise, and monitor the current, voltage, and temperature of the charging pile, including:

[0071] S11: The camera is installed on the top of the charging pile, facing the charging gun interface, with a field of view covering the area where the charging gun is inserted. It captures 1 frame of image per second, converts the image into a black and white image based on the Otsu algorithm, highlights the metal part of the charging gun, and eliminates noise through dilation and corrosion operations.

[0072] S12: An integrated MEMS microphone is installed around relays and fan components at a distance of 5-10cm to extract voiceprint features. ,in The original audio signal. The Mel-frequency cepstral coefficients (MFCCs) are extracted into 12 coefficients. MFCCs are a commonly used feature in audio processing and can capture the unique sound signature of sound, such as a relay jam or a fan malfunction.

[0073] Anomalies are detected by comparing the current MFCC coefficient with historical data, which is based on the normal distribution assumption: if a data point deviates from the mean by more than 3 times the standard deviation, it is considered an anomaly (corresponding to a 99.7% confidence interval).

[0074] Anomaly detection and labeling are performed based on the extracted voiceprint feature threshold. ,in This is the average of historical data. The historical data standard deviation is used to identify abnormal sound patterns, including relay jamming and fan failure. The maximum MFCC coefficient represents the main frequency components of the sound, and abnormal noise usually causes significant changes in this coefficient. The historical data mean and historical data standard deviation are calculated by collecting audio data during normal operation. A baseline is established by first recording the MFCC coefficient of 100 hours of normal sound, and then calculating the mean and standard deviation.

[0075] S13: A Hall effect sensor is used to monitor the current of the charging pile, with a sampling frequency of 1kHz; a voltage divider resistor network is used to monitor the voltage of the charging pile, with a sampling frequency of 1kHz; and a DS18B20 temperature sensor is used to monitor the temperature of the charging pile, with a sampling frequency of 1Hz.

[0076] Furthermore, in S2, the charging gun insertion status is detected in real time, only the feature vector is uploaded, the GPS module is used to synchronize the clocks of all sensors, and dynamic threshold cleaning is performed on the data, including:

[0077] S21: Establish an insertion state determination model to detect the charging gun insertion state in real time. ,in This represents the area of ​​the metal region of the charging gun when it is normally inserted.

[0078] Based on the adaptive threshold principle of computer vision, an environmental interference coefficient, such as changes in lighting, is introduced to dynamically adjust the threshold and avoid misjudgment. Fixed thresholds are unreliable in variable environments, and multiplying by a coefficient can compensate for environmental influences. Through experiments, the average area of ​​the metal region was calculated by repeatedly inserting the charging gun; the environmental interference coefficient was initially set to 1 and then finely adjusted based on real-time lighting data, such as the camera brightness value.

[0079] S22: Only the feature vector of the charging gun is transmitted. The resistance value is calculated based on the current and voltage of the charging pile monitored as described in S1;

[0080] S23: Each sensor data packet is appended with a GPS timestamp, aligned with the GPS time server based on the NTP protocol, the time offset is calculated, and all sensor clocks are synchronized.

[0081] S24: Calculate the moving average and standard deviation of the sensor data, and set the adaptive cleaning threshold. , ,in This represents the average data traffic. To define the standard deviation of data flow, we follow the statistical 3σ principle. To ensure safety, only obviously abnormal data, such as sudden noise, is cleaned, balancing sensitivity and stability. and The calculation is performed using a real-time sliding window, specifically the last 100 data points.

[0082] Furthermore, in S3, a multimodal feature fusion model is established for real-time evaluation, and a fault cause-effect graph is constructed based on Drools to perform dynamic risk level assessment, including:

[0083] S31: Perform weighted fusion of multi-source data. The multi-source data are the electrical data, visual data and environmental data mentioned in S1 and S2, respectively. The electrical data includes the current and voltage of the charging pile, the visual data includes the status of the charging gun, and the environmental data includes the temperature of the charging pile.

[0084] S32: Data feature interaction based on self-attention mechanism ,in For attention mechanism functions, For query, key, value vector, For function, As a dimension, based on a preset number of features, all features of the sensor, including current, voltage, etc., are denoted as . Value, output health status score Label the health status;

[0085] S33: Define rule templates, deduce fault root causes based on Drools rule chains, including abnormal insulation resistance → leakage protection triggering, and combine the output health status score. Based on reinforcement learning, the rules are updated with real-time data. If a rule frequently misreports (i.e., its score is close to the threshold), its weight is reduced to avoid overfitting. An exponential function ensures smooth adjustment.

[0086] ,in For the updated rule weights, As the initial rule weights, The attenuation coefficient was set experimentally. The threshold for health status scoring is usually set at an empirical value of 0.7, which is calibrated using historical fault data.

[0087] S34: Calculate risk indicators and establish a risk value calculation formula. , ,in The weights are set based on the analytic hierarchy process (AHP). The environmental risk coefficient is calculated by averaging the charging pile temperature mentioned in S1 and the GPS module location information mentioned in S2. .

[0088] Furthermore, in S4, dynamic power allocation and load balancing are performed, including:

[0089] S41: Real-time health status assessment of each charging station based on the description in S3 Remaining capacity of energy storage equipment and the number of users currently charging The allocation weight of each charging station , ,in To prevent tiny constants with a denominator of zero, the weights are proportional to the health status and remaining capacity, and inversely proportional to the number of connected users, ensuring that power is prioritized for charging stations with good health, sufficient power, and few users, thereby improving overall efficiency.

[0090] S42: Allocate power to each charging station according to its weight. , ,in , For real-time load of the power grid, The upper limit of the power grid capacity is determined by the power system parameters provided in real time. The upper limit is based on infrastructure settings such as transformer capacity and the allocation ratio is dynamically adjusted through the available power of the power grid.

[0091] Furthermore, the dynamic pricing and guidance based on user behavior in S5 includes:

[0092] By combining real-time grid load, historical peak demand, and user behavior preferences, the charging unit price is dynamically adjusted, and users are guided to charge during off-peak hours through points rewards and priority enhancement mechanisms.

[0093] Electricity prices increase linearly with grid load and users’ peak preferences. Based on demand response theory, high prices suppress peak demand, while low prices encourage off-peak charging.

[0094] Establish a dynamic pricing model , ,in The benchmark electricity price, The load sensitivity coefficient for the power grid was determined to be 0.4 through regression analysis of historical data. The user peak preference penalty coefficient was determined to be 0.3 through historical data regression. For real-time load of the power grid, This is the upper limit of the power grid capacity. The percentage of charging during peak hours is calculated from user data.

[0095] Charging during off-peak hours earns points, which are calculated based on the product of the charging amount and the dynamically priced electricity price. Charging stations are allocated to users based on their points during historical off-peak charging periods.

[0096] Furthermore, in the S6 state-generated scheduling path, the hash value of the data is stored using a hash chain, including:

[0097] S61: Combine the risk level definition in S3 to define the urgency score E of the fault. If the risk level is high, E=1.0; if the risk level is medium, E=0.8; if the risk level is low, E=0.5; otherwise, E=0.3.

[0098] S62: The score is directly proportional to skill level and inversely proportional to distance and number of tasks, reflecting the priority of proximity, off-peak hours, and high skills, combined with skill matching. Distance from the fault point Current number of tasks Define the availability score for maintenance engineers. , , Based on engineer certification level assignment, distance and number of tasks are directly based on real-time data;

[0099] S63: Prioritize assigning high-urgency faults to engineers. Initial solution = select from high to low by pressing E. Assign tasks from highest to lowest priority, randomly remove n low-priority tasks, and then assign tasks based on the remaining engineers. Tasks are reassigned to E, and a fitness function is established based on multi-objective optimization theory. ,in The weights are determined based on experimental adjustments. ,in Total scheduling time, To improve resource utilization, a tabu list is set to prevent duplicate solutions, and a maximum number of iterations is set.

[0100] S64: The data digest includes the fault type, maintenance engineer ID, and maintenance time. Each maintenance record generates a hash value and links it to the previous block to form an immutable chain. A digest is extracted from each maintenance record, and the digest = {fault type, maintenance engineer ID, maintenance time}. The hash value of the current block is generated using the SHA-256 hash function.

[0101] Furthermore, the performance parameters of the associated charging piles in S7 are periodically updated in the model parameters, including:

[0102] S71: Correlate keywords in user reviews, including slow charging and interface lag, with the physical parameters of the charging pile, and quantify the impact of user feedback on the performance of the charging pile based on NLP technology;

[0103] S72: Add new data, including fault data under extreme weather conditions, combined with user feedback, and update model parameters regularly with new data. Update model parameters with new data every 24 hours, and retain only fault data related to extreme weather.

[0104] Furthermore, a remote monitoring and management control system for new energy vehicle charging piles is provided to implement the remote monitoring and management control method for new energy vehicle charging piles described in any of the above-mentioned embodiments. The remote monitoring and management control system for new energy vehicle charging piles includes: a system management module, a charging pile management module, a charging pile repair management module, a user management module, a repairman management module, and a personal center module.

[0105] System management module: Deploys multimodal sensors, installs cameras to monitor charging gun status, integrates MEMS microphones to monitor abnormal noise, deploys current, voltage, and temperature sensors to monitor the current, voltage, and temperature of the charging pile, runs models on the edge gateway to detect the charging gun insertion status in real time, uploads only the binary results and feature vectors of the charging gun status, aligns all sensor clocks, eliminates time offset, and performs dynamic threshold cleaning.

[0106] Charging pile management module: Performs weight calculations to achieve dynamic adjustment based on the available power of the power grid, prioritizing the allocation of charging piles with high health status and sufficient remaining capacity, combined with the real-time load and capacity limit of the power grid;

[0107] Charging pile repair management module: It adopts a self-attention mechanism to integrate electrical data, visual data and environmental data to output a health status score, and dynamically adjusts the rule weights based on predefined rules to conduct risk level assessment;

[0108] User management module: Establishes a dynamic pricing model and points reward mechanism, integrating grid load, user preferences, and peak-hour penalties to guide user behavior;

[0109] Repairman Management Module: Establishes an urgency score to optimize resource utilization. Combines urgency with resource availability to achieve collaborative optimization of dynamic path planning and resource utilization. Hash chain stores data summaries, including fault type, repair engineer ID, and repair time.

[0110] The user center module extracts keywords from user reviews, quantifies the impact of user feedback on charging pile performance based on NLP technology, updates model parameters with new data including extreme weather fault data every 24 hours, and retains only data related to environmental risks.

[0111] refer to Figure 2 The diagram shown is a module diagram of a remote monitoring, management and control system for new energy vehicle charging piles.

[0112] Example 2

[0113] The charging station is located in the city center, within the maximum grid capacity. =100kW, real-time load =70kW, peak hours. Recently, some users have complained about slow charging on charging station CZ-01, CZ-02 is overheating, and CZ-03 is normal. Ambient temperature 35°C, extreme weather, requiring dynamic system response.

[0114] Sensors are deployed at each charging station: cameras, MEMS microphones, and electrical parameter sensors.

[0115] Camera: Top-mounted, captures 1 frame per second, converts to black and white images using the Otsu algorithm to highlight the metal area of ​​the charging gun. (Image showing the metal area of ​​the CZ-01 during normal insertion.) Calibration is performed using historical data.

[0116] MEMS microphone: 5cm from the relay, voiceprint features extracted. Assuming historical MFCC coefficient average... =0.5, standard deviation =0.1, based on 100 hours of normal data.

[0117] Electrical parameter sensor: current / voltage sampling 1kHz, temperature sampling 1Hz. The real-time temperature value of CZ-02 is 50°C, and the normal threshold is 45°C.

[0118] Charging gun insertion status detection: The edge gateway runs a lightweight model. For CZ-01, the area of ​​the metal region in the current image is 480px², and the interference coefficient under normal lighting conditions is 1.0. Calculate the insertion state: Since 480 < 500 × 1.0, the insertion state = 0, no insertion. Only upload the binary results and feature vectors to reduce the amount of data.

[0119] Average data traffic =10MB / s, standard deviation =2MB / s, security factor k=3, cleaning threshold =10 + 3 × 2 = 16 MB / s. Current traffic is 15 MB / s < 16 MB / s, so the data is retained; if there is a sudden surge in traffic of 20 MB / s, it will be filtered.

[0120] Data from CZ-02 was integrated: electrical data: current = 30A, voltage = 220V; visual data: insertion state = 1; environmental data: temperature = 50°C. A self-attention mechanism was used, with [dimensions not specified]. =3, calculate health score =0.6, in the range of 0-1, the lower the value, the worse.

[0121] Initial rule weights =1.0, attenuation coefficient α=0.5, health threshold =0.7. Rule weight update: =1.0× =1.0× The weighting is approximately 0.95, with the weighting slightly reduced as the health score approaches the threshold.

[0122] parameter: The historical failure frequency of CZ-02 is 0.2, and the environmental risk coefficient after high temperature weighting is 0.8. Risk value: =0.4×0.6+0.3×0.2+0.3×0.8=0.24+0.06+0.24=0.54. Risk level: Since 0.3<0.54≤0.7, the level is medium.

[0123] Charging pile data: CZ-01: =0.8, =80%, =2; CZ-02: =0.6, =60%, =3; CZ-03: =0.9, =90%, =1; constant .

[0124] Weight calculation: , , The total weights are approximately 0.32 + 0.12 + 0.81 = 1.25.

[0125] Available power of the grid =100-70=30kW, power of each pile: ≈7.68kW ≈2.88kW ≈19.44kW.

[0126] Parameter: Benchmark electricity price =1.0 yuan / kWh, grid load sensitivity coefficient γ=0.4, user peak preference penalty coefficient δ=0.3, peak proportion β=0.7.

[0127] Electricity price calculation: =1.0×(1+0.28+0.21)=1.49 yuan / kWh. Points are awarded for charging during off-peak hours. For example, if a user charges 10kWh during off-peak hours, they will receive 10×1.0=10 points, which are used to prioritize the allocation of charging stations.

[0128] Urgency Score: In risk level CZ-02, the urgency level E = 0.8. Engineer Availability Score: Assuming Engineer A: Skill Matching... =1.0, distance =5km, number of tasks =1, Usability score .

[0129] Scheduling path optimization: Using the tabu search algorithm, initially assign high-E tasks, and iteratively optimize the fitness function (weights). =0.7), minimize the total time.

[0130] Hash chain storage: Repair record summary = {fault type: "overheating", engineer ID: "A001", repair time: "2023-10-01 10:00"}, generated with SHA-256 hash value, and the chained blocks ensure immutability.

[0131] User feedback correlation: CZ-01 user comments mention slow charging, NLP technology correlates this with low current in physical parameters, quantifies the impact and adjusts the model accordingly.

[0132] Model Update: New data is added every 24 hours. Due to the extreme weather of 35°C, CZ-02 malfunctioned, so the model parameters were updated, and only the relevant data were retained.

[0133] The above formulas are all dimensionless calculations. Dimensionless calculations can be performed using various methods such as standardization, which will not be elaborated here. The formulas are derived from software simulations based on a large amount of collected data, and the preset parameters in the formulas can be set by those skilled in the art according to the actual situation.

[0134] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0135] This invention provides a remote monitoring, management, and control method and system for new energy vehicle charging piles. It achieves real-time monitoring of charging gun status, abnormal noise, and electrical parameters by deploying multimodal sensors. Edge computing and dynamic threshold cleaning reduce data redundancy and improve transmission efficiency. A self-attention mechanism and Drools rule engine are combined for multi-source data fusion and root cause analysis to dynamically assess risk levels. Dynamic power allocation based on health status and grid load optimizes resource utilization. Peak-valley dynamic pricing and an incentive mechanism guide off-peak charging. A hash chain ensures the immutability of maintenance data. Scheduling paths are optimized based on urgency and engineer resources. User feedback and extreme weather data iterative models enhance system robustness. This achieves intelligent management of the entire charging pile lifecycle, improving operation and maintenance efficiency, safety, and user charging experience.

Claims

1. A new energy vehicle charging pile remote monitoring management control method, characterized in that, include: S1: Deploy multimodal sensors, install cameras to monitor the charging gun status, integrate MEMS microphones to monitor abnormal noise, and deploy current, voltage, and temperature sensors to monitor the current, voltage, and temperature of the charging pile; S2: Run a lightweight classification model on the edge gateway to detect the charging gun insertion status in real time. Only upload the charging gun feature vector, including the binary result of the charging gun insertion status. Use a GPS module to synchronize the clocks of all sensors, transmit data based on a compression algorithm, and perform dynamic threshold cleaning on the data. In step S2, the charging gun insertion status is detected in real time, only the feature vector is uploaded, all sensor clocks are synchronized using a GPS module, and dynamic threshold cleaning is performed on the data, including: S21: Establishing an insertion state judgment model to detect the insertion state of the charging gun in real time, wherein is the area of the metal region of the charging gun when normally inserted; S22: only the upper charging gun transmission eigenvector, The resistance value is calculated based on the current and voltage of the charging pile monitored in S1. S23: Each sensor data packet is appended with a GPS timestamp, aligned with the GPS time server based on the NTP protocol, the time offset is calculated, and all sensor clocks are synchronized. S24: Calculate the moving average and standard deviation of the sensor data, and set the adaptive cleaning threshold. , ,in This represents the average data traffic. The standard deviation of data traffic. For safety factor; S3: Real-time assessment of charging pile health status and fault location based on multi-source data, establishment of multi-modal feature fusion model for real-time assessment, construction of fault cause-effect graph based on Drools, and dynamic assessment of risk level; S4: Based on real-time data and prediction models, adaptive control and resource optimization of charging piles are performed, and dynamic power allocation and load balancing are carried out. The dynamic power allocation and load balancing in S4 includes: S41: Real-time health status assessment of each charging pile based on S3 Remaining capacity of energy storage equipment and the number of users currently charging The allocation weight of each charging station , ,in To prevent tiny constants with a denominator of zero; S42: Allocate power to each charging station according to its weight. , ,in , For real-time load of the power grid, The upper limit of the power grid capacity is determined by dynamically adjusting the allocation ratio based on the available power in the power grid. S5: Dynamic pricing and guidance based on user behavior, increasing the unit price during peak charging hours to guide users to charge during off-peak hours; S6: Establish a maintenance resource scheduling algorithm based on the urgency of the fault and the status of maintenance resources, dynamically generate scheduling paths, and use a hash chain to store the hash value of the data. The data digest includes the fault type, maintenance engineer ID, and maintenance time. S7: Based on keywords in user reviews, such as slow charging and interface lag, we correlate charging pile performance parameters, regularly update model parameters, and add new data, including fault data under extreme weather conditions.

2. The new energy vehicle charging pile remote monitoring management control method according to claim 1, characterized in that, The camera installed in S1 monitors the status of the charging gun, detects abnormal noise, and monitors the current, voltage, and temperature of the charging pile, including: S11: The camera is installed on the top of the charging pile, facing the charging gun interface, and the field of view covers the area where the charging gun is inserted. Based on the Otsu algorithm, the image is converted into a black and white image to highlight the metal part of the charging gun. Noise is eliminated through dilation and erosion operations. S12: integrated MEMS microphone, extract voiceprint features, wherein is the original audio signal, is the mel-frequency cepstral coefficient, 12 coefficients are extracted; threshold decision based on the extracted voiceprint features and flagging, wherein is the historical data mean, is the historical data standard deviation, the abnormal voiceprint includes a relay stuck and a fan failure; S13: A Hall effect sensor is used to monitor the current of the charging pile, with a sampling frequency of 1kHz; a voltage divider resistor network is used to monitor the voltage of the charging pile, with a sampling frequency of 1kHz; and a DS18B20 temperature sensor is used to monitor the temperature of the charging pile, with a sampling frequency of 1Hz.

3. The new energy vehicle charging pile remote monitoring management control method according to claim 1, characterized in that, The S3 section establishes a multimodal feature fusion model for real-time evaluation, constructs a fault cause-effect graph based on Drools, and performs dynamic risk level assessment, including: S31: Weighted fusion of multi-source data, which includes electrical data, visual data and environmental data in S1. Electrical data includes the current and voltage of the charging pile, visual data includes the status of the charging gun, and environmental data includes the temperature of the charging pile. S32: Data feature interaction based on self-attention mechanism ,in For attention mechanism functions, For query, key, value vector, For function, As a dimension, output a health status score. Label the health status. S33: Define rule templates, deduce fault root causes based on Drools rule chains, including abnormal insulation resistance → leakage protection triggering, and combine the output health status score. Combine with real-time data to update rule weights. ,in For the updated rule weights, As the initial rule weights, The attenuation coefficient is... Threshold for health status scoring; S34: Calculate risk indicators and establish a risk value calculation formula. , ,in The environmental risk coefficient is calculated by weighting the charging pile temperature mentioned in S1 and the GPS module location information mentioned in S2. .

4. The new energy vehicle charging pile remote monitoring management control method according to claim 1, characterized in that, The dynamic pricing and guidance based on user behavior in S5 includes: By combining real-time grid load, historical peak demand, and user behavior preferences, the charging unit price is dynamically adjusted, and users are guided to charge during off-peak hours through points rewards and priority enhancement mechanisms. Establish a dynamic pricing model , ,in The benchmark electricity price, The load sensitivity coefficient for the power grid is set to 0.

4. The user peak preference penalty coefficient is set to 0.

3. For real-time load of the power grid, This is the upper limit of the power grid capacity. The percentage of charging during peak hours; Charging during off-peak hours earns points, which are calculated based on the product of the charging amount and the dynamically priced electricity price. Charging stations are allocated to users based on their points during historical off-peak charging periods.

5. The new energy vehicle charging pile remote monitoring management control method according to claim 1, characterized in that, The S6 process dynamically generates scheduling paths, using a hash chain to store the hash values ​​of the data, including: S61: Combine the risk level definition in S3 to define the urgency score E of the fault. If the risk level is high, E=1.0; if the risk level is medium, E=0.8; if the risk level is low, E=0.5; otherwise, E=0.

3. S62: Combining skill matching Distance from the fault point Current number of tasks Define the availability score for maintenance engineers. , ; S63: Prioritize assigning high-urgency faults to engineers. Initial solution = select from high to low by pressing E. Assign tasks from highest to lowest priority, randomly remove n low-priority tasks, and then assign tasks based on the remaining engineers. And E reassign tasks, fitness function ,in As weight, ,in Total scheduling time, To improve resource utilization, a tabu list is set to prevent duplicate solutions, and a maximum number of iterations is set. S64: The data digest includes the fault type, maintenance engineer ID, and maintenance time. Each maintenance record generates a hash value and links it to the previous block to form an immutable chain. A digest is extracted from each maintenance record, and the digest = {fault type, maintenance engineer ID, maintenance time}. The hash value of the current block is generated using the SHA-256 hash function.

6. The new energy vehicle charging pile remote monitoring management control method according to claim 1, characterized in that, The performance parameters of the associated charging piles in S7 are updated periodically, including: S71: Correlate keywords in user reviews, including slow charging and interface lag, with the physical parameters of the charging pile, and quantify the impact of user feedback on the performance of the charging pile based on NLP technology; S72: Add new data, including fault data under extreme weather conditions, combined with user feedback, and update model parameters regularly with new data. Update model parameters with new data every 24 hours, and retain only fault data related to extreme weather.

7. A new energy vehicle charging pile remote monitoring management control system, characterized in that, To implement the remote monitoring, management, and control method for new energy vehicle charging piles as described in any one of claims 1-6, the remote monitoring, management, and control system for new energy vehicle charging piles includes: a system management module, a charging pile management module, a charging pile repair management module, a user management module, a repairman management module, and a personal center module. System management module: Deploys multimodal sensors, installs cameras to monitor charging gun status, integrates MEMS microphones to monitor abnormal noise, deploys current, voltage, and temperature sensors to monitor the current, voltage, and temperature of the charging pile, runs models on the edge gateway to detect the charging gun insertion status in real time, uploads only the binary results and feature vectors of the charging gun status, aligns all sensor clocks, eliminates time offset, and performs dynamic threshold cleaning. Charging pile management module: Performs weight calculations to achieve dynamic adjustment based on the available power of the power grid, prioritizing the allocation of charging piles with high health status and sufficient remaining capacity, combined with the real-time load and capacity limit of the power grid; Charging pile repair management module: It adopts a self-attention mechanism to integrate electrical data, visual data and environmental data to output a health status score, and dynamically adjusts the rule weights based on predefined rules to conduct risk level assessment; User management module: Establishes a dynamic pricing model and points reward mechanism, integrating grid load, user preferences, and peak-hour penalties to guide user behavior; Repairman Management Module: Establishes an urgency score to optimize resource utilization. Combines urgency with resource availability to achieve collaborative optimization of dynamic path planning and resource utilization. Hash chain stores data summaries, including fault type, repair engineer ID, and repair time. The user center module extracts keywords from user reviews, quantifies the impact of user feedback on charging pile performance based on NLP technology, updates model parameters with new data including extreme weather fault data every 24 hours, and retains only data related to environmental risks.

Images