New energy charging pile operation data analysis processing method

Abstract

The invention discloses a new energy charging pile operation data analysis processing method, and relates to the technical field of new energy charging pile data analysis, and the method comprises the following steps: S1, collecting multi-source operation data such as charging power, current and voltage of a charging pile, and storing the multi-source operation data according to a timestamp; s2, preprocessing data, removing abnormal values, complementing missing values and standardizing; s3, extracting instantaneous operation, time dimension and equipment state core features; s4, user charging behaviors are clustered and analyzed to generate a classification label library; s5, analyzing an operation load situation and identifying a load fluctuation factor; s6, evaluating the health state of the equipment, grading and performing early warning; s7, generating optimization strategies such as load scheduling; and S8, collecting new data to iteratively optimize model parameters and algorithm logic. According to the method, multi-source data fusion and accurate analysis are realized, the fault identification accuracy and the load balance degree are improved, and a landing optimization strategy is generated; different scenes are adapted through iterative optimization, and scientific support is provided for fine management of the charging pile.

Description

Technical Field

[0001] This invention relates to the field of new energy charging pile data analysis technology, and in particular to a method for analyzing and processing new energy charging pile operation data. Background Technology

[0002] With the continuous increase in the number of new energy vehicles, charging piles, as core supporting facilities, directly impact the promotion and user experience of new energy vehicles through their operational stability and service efficiency. During long-term, high-frequency operation, charging piles face challenges such as large load fluctuations, rapid equipment aging, and diverse user needs, making them prone to charging failures, load imbalances, and untimely maintenance. These issues not only affect the user charging experience but may also pose safety hazards. Therefore, achieving real-time monitoring, fault warning, and optimized scheduling of charging pile operation status through precise data analysis and processing has become crucial for improving the operation and maintenance management level and service quality of charging piles, and is also an important research direction in the field of smart energy.

[0003] While existing technologies have made some progress in data analysis and processing, several technical shortcomings remain. For example, patent application CN118885703B, entitled "A Method for Detecting and Analyzing the Operating Status of High-Power Devices in Charging Piles," mentions collecting detection data from components such as charging pile switches, electromagnetic locks, and resistors. After preprocessing, the least squares method is used to fit the optimal curve, which is then compared with current data to determine faults. However, this technology only focuses on fault detection of high-power devices and does not involve overall charging pile load analysis, user behavior feature mining, or operational strategy optimization. The data collection dimension is limited, and the application scenario is confined to fault diagnosis, failing to meet the needs of full lifecycle operation and maintenance and efficient scheduling of charging piles.

[0004] The patent application CN118132956A, entitled "A Method for Analyzing and Processing Operational Data of New Energy Charging Piles Based on Artificial Intelligence," mentions periodic data monitoring of charging piles in different charging areas. By calculating indicators such as daily charging continuity and regional charging saturation, it aims to identify charging pile anomalies and provide dynamic alerts for regional management. However, this technology focuses on overall regional status analysis, lacking precision in assessing the health status of individual charging piles. It does not introduce a dynamic optimization mechanism and fails to consider the impact of environmental factors and grid parameters on charging pile operation. The comprehensiveness and adaptability of the analysis results are insufficient, making it difficult to support the generation of personalized operation optimization strategies.

[0005] The patent application document with publication number CN117892155B and title "A Data Processing Method and System for Charging Piles" mentions building a predictive model based on a multi-view shapelet learning mechanism to predict charging and operation data of charging piles, and combining vehicle feature information to achieve matching recommendations between charging piles and vehicles. However, this technology focuses primarily on user matching recommendations, paying insufficient attention to core operation and maintenance needs such as the health status assessment and load balancing scheduling of the charging piles themselves. It lacks a complete health assessment indicator system and iterative optimization mechanism, and the data processing results are difficult to directly translate into practical strategies for equipment maintenance and load scheduling, thus limiting its practicality and applicability.

[0006] In summary, existing technologies suffer from three major limitations: First, data collection is fragmented, focusing primarily on single scenarios or indicators, failing to integrate multi-source data from equipment operation, user behavior, environmental parameters, and power grid status. Second, the depth of analysis is insufficient, often remaining at the level of fault identification or status statistics, lacking precise quantitative assessment and dynamic optimization of load status and health conditions. Third, the application orientation is unclear, with analysis results disconnected from actual operation and maintenance needs, making it difficult to generate directly implementable operation optimization strategies. These issues result in insufficient accuracy, comprehensiveness, and practicality in charging pile operation data analysis and processing, failing to meet the demands for efficient operation and maintenance and intelligent scheduling of charging piles in complex scenarios. Therefore, a multi-source fusion, precise analysis, and dynamic optimization data analysis and processing method is urgently needed to overcome the bottlenecks of existing technologies. Summary of the Invention

[0007] This invention proposes a method for analyzing and processing operational data of new energy charging piles to solve the problems mentioned in the prior art.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: a method for analyzing and processing operational data of new energy charging piles, comprising the following steps: S1: Multi-source operation data acquisition. Through the built-in metering module, communication module, environmental sensor and grid-side acquisition device of the charging pile, charging parameters, equipment status, environmental information and grid data are collected simultaneously and the collected data is structured and stored to form the original operation dataset. S2: Data quality preprocessing, outlier removal, missing value completion and noise smoothing are performed on the original running dataset, and then the data is standardized and normalized to generate a standardized running dataset. S3: Basic operation feature extraction. Based on a standardized dataset, instantaneous operation, time dimension and equipment status features are extracted. Redundant feature parameters are removed by feature dimensionality reduction algorithm, and core feature set that is strongly correlated with the charging pile's operation status and health is retained. S4: Charging behavior feature clustering. The core feature set is selected as input, and the density clustering algorithm is used to perform cluster analysis on user charging behavior, classify user types, and generate a charging behavior clustering model and user classification label library. S5: Operational load situation analysis. Based on standardized operation datasets, construct charging pile operation load time series, calculate multi-timescale load parameters, identify load influencing factors by combining clustering models, and quantitatively evaluate load balance and fluctuation characteristics. S6: Equipment health status assessment. Combine relevant equipment operation data to construct an assessment index system, determine the index weights and calculate the health index, and divide the equipment health status into four levels: healthy, sub-healthy, abnormal, and faulty. Automatically trigger early warning prompts for abnormal states and generate assessment reports and traceability information. S7: Optimization strategy generation. Based on charging behavior clustering, load situation analysis and equipment health assessment results, combined with charging pile geographical location and surrounding distribution data, it generates a set of optimization strategies for load scheduling, equipment maintenance and user charging guidance to address issues such as uneven load, equipment hazards and demand matching. S8: Iterative optimization of the data model. Repeat steps S1-S7 to evaluate the implementation effect, compare key indicators, optimize feature extraction, clustering algorithm and evaluation indicator weight parameters through gradient descent algorithm, iteratively improve the accuracy and applicability of the model, and generate the optimized model and strategy library.

[0009] Furthermore, it also includes a load balancing quantification calculation step, executed in S5, which quantifies the degree of load balancing of the charging piles using a formula, the specific formula being: in For load balance, To count the number of sampling points within the statistical time window, For the first The charging power at each sampling point This represents the average charging power within the statistical time window.

[0010] Furthermore, it includes a dynamic correction step for equipment health, executed in S6. This step dynamically adjusts the equipment health index using a formula that considers equipment runtime and environmental factors. The specific formula is as follows: in This is the dynamically adjusted equipment health index. For the initial calculation of the equipment health index, The cumulative runtime of the device. This is the difference between the ambient temperature and the standard operating temperature. This is the runtime impact coefficient. This represents the influence coefficient of ambient temperature.

[0011] Furthermore, in the S1 multi-source operation data acquisition step, the metering module adopts a high-precision power metering chip with a measurement accuracy level of not less than 0.5. The communication module supports 4G / 5G, Ethernet, and LoRa communication protocols. The environmental sensor adopts a temperature and humidity composite sensor, and the geographical location information is obtained through the Beidou positioning module.

[0012] Furthermore, in the S2 data quality preprocessing step, the threshold for identifying abnormal data is set based on the equipment's factory parameters and historical normal operating data. The linear interpolation method for missing value completion is applied to continuous data. The window size for the moving average filtering method is set to 5 sampling points. Data standardization uses the Z-score standardization method, with the formula: ,in The original data, The mean of the data. This represents the standard deviation of the data.

[0013] Furthermore, in the S3 basic operation feature extraction step, the window size of the time window sliding method is set to 10 minutes, the change rate of charging amount per unit time is calculated by the difference in charging amount between adjacent sampling points and the time interval, and the feature dimensionality reduction algorithm adopts the principal component analysis algorithm, retaining the principal components with a cumulative variance contribution rate of not less than 85% as the core feature parameters.

[0014] Furthermore, in the S4 charging behavior feature clustering step, the neighborhood radius of the density clustering algorithm is set to 0.5, the minimum number of neighborhood samples is set to 5, the clustering density is calculated by the number of samples within a unit time window, the user classification label library is updated in real time, and newly added user behavior data is automatically assigned to the corresponding cluster category or forms a new category.

[0015] Furthermore, in the S5 operation load situation analysis step, the calculation period for the load average, fluctuation rate, and peak value is set to 1 hour, 1 day, and 1 month, respectively. The load situation heat map is drawn using color depth to represent load intensity. The horizontal axis of the load change trend curve is time, and the vertical axis is charging power. The load change trend is displayed through a curve fitting algorithm.

[0016] Furthermore, in the S6 equipment health status assessment step, the health assessment index system includes equipment operating status indicators, failure frequency indicators, offline duration indicators, and maintenance record indicators. The judgment matrix of the analytic hierarchy process is constructed based on expert experience and historical data. Fault risk tracing information is associated with the fault occurrence time, corresponding operating characteristics, and environmental conditions to achieve fault cause location.

[0017] Furthermore, in the S7 operation optimization strategy generation step, the load dispatch strategy includes peak-shaving and valley-filling strategies such as peak-shifting charging guidance, multi-pile power allocation, and interaction with the power grid; the equipment maintenance optimization strategy includes preventive maintenance plans based on health indices and rapid repair and replacement strategies for faulty equipment; and the user charging guidance strategy includes recommending the best charging pile, guiding users to idle charging piles, and providing charging discount information. The operation optimization strategy set is displayed through a visual interface, which supports operators to view and execute it directly. In the S8 data model iterative optimization steps, the indicators for quantitatively evaluating the implementation effect include the load balance improvement rate, equipment health index improvement rate, user average waiting time for charging reduction rate, and charging failure rate reduction rate. The initial learning rate of the gradient descent algorithm is set to 0.01. The iteration stopping condition is that the improvement rate of the evaluation indicators is less than 1% for three consecutive times or the maximum number of iterations of 1000 is reached. The optimized operation strategy library is stored on the cloud server, supporting sharing and synchronous updates of multiple charging piles.

[0018] Compared with existing technologies, the beneficial effects of this invention are: First, multi-source data fusion enhances the comprehensiveness of the analysis. This invention overcomes the limitations of existing technologies that rely on a single data collection dimension. It simultaneously collects multi-source data, including charging pile operating parameters, user charging behavior, ambient temperature and humidity, and power grid status. Combined with redundant dual-sampling channels and standardized preprocessing, it ensures data integrity and reliability. The deep fusion of multi-dimensional data comprehensively depicts the complex operating scenarios of charging piles, encompassing both the equipment's own status and the influence of the external environment and user needs. This lays a solid foundation for subsequent feature extraction, situational analysis, and strategy generation, solving the core problem of inaccurate analysis results caused by incomplete data in existing technologies.

[0019] Secondly, precise modeling and quantitative assessment enhance the depth of analysis. This invention extracts multi-dimensional core features using a sliding time window method, employs density clustering algorithms to mine user behavior patterns, and constructs a scientific health assessment indicator system and load balancing quantitative model to achieve precise quantitative assessment of the charging pile's operating status, health level, and load situation. Compared to the simple status statistics or fault judgments of existing technologies, this invention can deeply mine the correlation patterns behind the data, accurately identify problems such as equipment aging hazards, load imbalance, and mismatch between user needs, providing a clear direction for targeted optimization and significantly improving the depth and accuracy of data analysis.

[0020] Furthermore, dynamic optimization and strategy implementation enhance the practicality of the application. Based on data analysis results, this invention generates multi-dimensional operational optimization strategies for load scheduling, equipment maintenance, and user guidance. Through an iterative optimization mechanism, model parameters are continuously updated based on the strategy implementation results, ensuring the adaptability and effectiveness of the strategies. The optimized strategies closely align with the operation and maintenance needs of charging piles and can be directly transformed into practical solutions. This not only solves the problem of the disconnect between existing technical analysis results and actual needs but also continuously improves the quality of the strategies through dynamic optimization, significantly improving the operational stability and service efficiency of charging piles and reducing operation and maintenance costs.

[0021] Finally, the end-to-end closed-loop management expands application value. This invention constructs an end-to-end closed-loop processing mechanism encompassing "data acquisition - preprocessing - feature extraction - analysis and evaluation - strategy generation - iterative optimization," covering multiple stages such as charging pile operation monitoring, fault early warning, load scheduling, and user services, achieving refined management of the entire lifecycle of charging piles. This mechanism not only promptly identifies and resolves various problems in charging pile operation but also continuously improves data analysis and processing capabilities through data accumulation and model optimization, adapting to the application needs of different scenarios and types of charging piles, providing core technical support for the construction of smart charging networks, and possessing significant economic and social value. Attached Figure Description

[0022] Figure 1 A process flow diagram for analyzing and processing operational data of new energy charging piles; Figure 2 Line graph showing changes in device health index over different operating periods; Figure 3 A line graph showing the relationship between the number of model iterations and the analysis error; Figure 4 A bar chart comparing the operation and maintenance costs of charging piles in different scenarios; Figure 5 A radar chart for multi-dimensional performance evaluation of charging pile data analysis methods. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0025] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified. Furthermore, the terms "installed," "connected," and "linked" should be interpreted broadly; for example, they may refer to a fixed connection, a detachable connection, or an integral connection; they may refer to a mechanical connection or an electrical connection; they may refer to a direct connection or an indirect connection through an intermediate medium; and they may refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances. The invention will now be described in further detail with reference to the accompanying drawings.

[0026] Reference Figures 1 to 5 A method for analyzing and processing operational data of new energy charging piles, comprising the following steps: S1: Multi-source operation data acquisition. Through the metering module, communication module, environmental sensors and grid-side acquisition device built into the new energy charging pile, the real-time charging power, current and voltage, charging time, cumulative charging amount, equipment operation status, gun plugging and unplugging status, ambient temperature and humidity, geographical location information and grid power supply voltage and frequency data of the charging pile are collected simultaneously. The data acquisition frequency is set to 1 second / time. Dual sampling channels are used for redundant acquisition of key status parameters to ensure data integrity and real-time performance. The collected data is stored in a structured manner according to timestamps to form the original operation dataset. S2: Data quality preprocessing involves cleaning the collected raw running dataset, using the Grubbs criterion to identify and remove outlier data points, and using linear interpolation to complete missing sampled data. For data that is highly susceptible to external interference, such as environmental temperature and humidity, a moving average filtering method is used for smoothing to eliminate random noise interference. After data cleaning, data standardization is performed to map various types of data with different dimensions to a unified numerical range, thereby normalizing the data format and generating a standardized running dataset. S3: Basic operation feature extraction. Based on the standardized operation dataset, the instantaneous operation features of the charging pile are extracted using the time window sliding method, including instantaneous peak charging power, average charging power, and rate of change of charging amount per unit time. Time dimension features are extracted, including daily charging frequency, monthly charging saturation, and peak period charging ratio. Equipment status features are extracted, including offline duration, number of failures, and normal operation rate within the maintenance cycle. Redundant feature parameters are removed through feature dimensionality reduction algorithm, retaining the core feature set that is strongly correlated with the operation status and health of the charging pile. S4: Charging behavior feature clustering. The core feature set is selected as input, and the density clustering algorithm is used to perform clustering analysis on the charging user behavior corresponding to the charging pile. According to the dimensions of charging time period preference, charging power demand, charging duration, charging frequency, etc., the charging users are divided into multiple categories such as fixed time period type, random demand type, high power fast charging type, and low frequency maintenance type. The behavior feature center and cluster density of different categories of users are calculated to generate charging behavior clustering model and user classification label library. S5: Operational load situation analysis. Based on a standardized operational dataset, a time series of charging pile operational load is constructed. The sliding window method is used to calculate the average load, load fluctuation rate, and load peak at different time scales. The load change trend of charging piles within days, weeks, and months is analyzed. Combined with a charging behavior clustering model, the influencing factors of different user categories on the load situation are identified. A load situation heat map and load change trend curve are generated to quantitatively evaluate the balance and fluctuation characteristics of the charging pile operational load. S6: Equipment health status assessment. Combining equipment operation status data, number of failures, and offline duration data from the core feature set, a health assessment index system is introduced. The weight of each index is determined by the analytic hierarchy process. The health index of the charging pile equipment is calculated by the weighted summation method. The equipment health status is divided into four levels: healthy, sub-healthy, abnormal, and faulty. An early warning prompt is automatically triggered for abnormal status, and an equipment health status assessment report and fault risk tracing information are generated. S7: Operation optimization strategy generation. Based on the charging behavior clustering model, load situation analysis results, and equipment health status assessment report, it analyzes the load imbalance problem, equipment aging potential problem, and user demand matching problem during the operation of charging piles. Combining the geographical location information of charging piles and the distribution data of similar charging piles in the surrounding area, it generates load scheduling strategy, equipment maintenance optimization strategy, and user charging guidance strategy. The output is a set of operation optimization strategies including optimal charging time period suggestions, maintenance cycle adjustment schemes, and power allocation optimization parameters. S8: Iterative optimization of the data model. Collect actual operation data after the implementation of the charging pile operation optimization strategy as a new original operation dataset. Repeat steps S1 to S7 to quantitatively evaluate the implementation effect of the operation optimization strategy set. By comparing the load balance, equipment health index, and user satisfaction index before and after the strategy implementation, use the gradient descent algorithm to iteratively optimize the basic operation feature extraction parameters, charging behavior clustering algorithm parameters, and health assessment index weights to continuously improve the accuracy and applicability of the data analysis and processing model, and generate the iteratively optimized data analysis and processing model and operation strategy library.

[0027] This invention also includes a load balancing quantification calculation step, executed in S5, which quantifies the degree of load balancing of the charging piles using a formula, the specific formula being: in This represents the load balance degree, with a value ranging from 0 to 1. A higher value indicates better load balance. To count the number of sampling points within the statistical time window, For the first The charging power at each sampling point The average charging power within the statistical time window is calculated to accurately quantify load fluctuations and equilibrium states, providing a quantitative basis for subsequent load scheduling strategies.

[0028] This invention also includes a dynamic equipment health correction step, executed in S6, which dynamically adjusts the equipment health index using a formula that considers equipment runtime and environmental influencing factors. The specific formula is as follows: in This is the dynamically adjusted equipment health index. For the initial calculation of the equipment health index, The cumulative runtime of the device. This is the difference between the ambient temperature and the standard operating temperature. This is the runtime impact coefficient. The environmental temperature influence coefficient is used to calculate and achieve dynamic and accurate assessment of the equipment's health status, adapting to health determination under different operating durations and environmental conditions.

[0029] In this invention, during the S1 multi-source operation data acquisition step, the metering module uses a high-precision power metering chip with a measurement accuracy level of not less than 0.5. The communication module supports multiple communication protocols such as 4G / 5G, Ethernet, and LoRa. The environmental sensor uses a temperature and humidity composite sensor with a measurement range of -40℃ to 85℃ and 0 to 100%RH, with accuracies of ±0.5℃ and ±3%RH, respectively. The geographical location information is obtained through the Beidou positioning module with a positioning accuracy of not less than 10 meters.

[0030] In this invention, in the S2 data quality preprocessing step, the threshold for identifying abnormal data is set based on the equipment's factory parameters and historical normal operation data. The linear interpolation method for missing value completion is applied to continuous data. The window size for the moving average filtering method is set to 5 sampling points. Data standardization uses the Z-score standardization method, with the formula: ,in The original data, The mean of the data. This represents the standard deviation of the data.

[0031] In this invention, in the S3 basic operation feature extraction step, the window size of the time window sliding method is set to 10 minutes, the change rate of charging amount per unit time is calculated by the difference of charging amount between adjacent sampling points and the time interval, and the feature dimensionality reduction algorithm adopts the principal component analysis algorithm, retaining the principal components with a cumulative variance contribution rate of not less than 85% as the core feature parameters.

[0032] In this invention, in the S4 charging behavior feature clustering step, the neighborhood radius of the density clustering algorithm is set to 0.5, the minimum number of neighborhood samples is set to 5, the clustering density is calculated by the number of samples within a unit time window, the user classification label library is updated in real time, and newly added user behavior data is automatically assigned to the corresponding cluster category or forms a new category.

[0033] In this invention, in the S5 load status analysis step, the calculation period for the average load, fluctuation rate, and peak load is set to 1 hour, 1 day, and 1 month, respectively. The load status heat map is drawn using color depth to represent load intensity. The horizontal axis of the load change trend curve is time, and the vertical axis is charging power. The load change trend is displayed through a curve fitting algorithm.

[0034] In this invention, the health assessment index system in the S6 equipment health status assessment step includes equipment operating status index, failure frequency index, offline duration index, and maintenance record index. The judgment matrix of the analytic hierarchy process is constructed based on expert experience and historical data. The fault risk tracing information is associated with the fault occurrence time, corresponding operating characteristics, and environmental conditions to achieve rapid location of the fault cause.

[0035] In this invention, the S7 operation optimization strategy generation step includes load scheduling strategies such as peak-shaving charging guidance, multi-pile power allocation, and peak-shaving and valley-filling strategies that interact with the power grid; equipment maintenance optimization strategies include preventive maintenance plans based on health indices and rapid repair and replacement strategies for faulty equipment; user charging guidance strategies include recommending the best charging piles, guiding users to idle charging piles, and providing charging discount information; the operation optimization strategy set is displayed through a visual interface, supporting operators to directly view and execute it. In the S8 data model iterative optimization steps, the indicators for quantitatively evaluating the implementation effect include the load balance improvement rate, equipment health index improvement rate, user average waiting time for charging reduction rate, and charging failure rate reduction rate. The initial learning rate of the gradient descent algorithm is set to 0.01. The iteration stopping condition is that the improvement rate of the evaluation indicators is less than 1% for three consecutive times or the maximum number of iterations of 1000 is reached. The optimized operation strategy library is stored on the cloud server, supporting sharing and synchronous updates of multiple charging piles.

[0036] The following two examples further illustrate specific embodiments of the present invention: Example 1: Application of centralized charging pile clusters in urban commercial areas This embodiment is applied to a centralized charging pile cluster in the core business district of a city. The area has 20 DC fast charging piles, serving new energy vehicle users in surrounding office buildings and shopping malls. The average daily charging demand is concentrated in the morning peak from 8-10 am, noon peak from 12-2 pm, and evening peak from 6-8 pm. It is characterized by large load fluctuations, diverse user charging needs, and high-frequency equipment operation. The goal is to achieve accurate monitoring of the charging pile operation status, load balancing scheduling, and equipment health management through this method.

[0037] S1: Multi-source operational data acquisition. Each charging pile is equipped with a metering module and a status monitoring module, and external environmental sensors are installed, synchronously connected to the power grid-side data acquisition device. The metering module collects real-time charging power, current, voltage, charging time, and cumulative charging amount. The status monitoring module collects equipment operating status and charging gun insertion / removal status. The environmental sensors collect ambient temperature, humidity, and rainfall data. The power grid-side device collects power supply voltage and frequency data. All data acquisition is set to a frequency of 1 second per acquisition. Dual sampling channels are used for redundant acquisition of key parameters such as charging power, current, and voltage to avoid data loss due to single-channel failure. The acquired data is stored in a structured JSON format with timestamps on a local server, forming the original operational dataset.

[0038] S2: Data quality preprocessing involves a comprehensive cleaning of the original operational dataset. The Grubbs criterion is used to identify and remove anomalous data points caused by momentary sensor malfunctions, such as data where charging power suddenly drops to 0 or far exceeds the device's rated power. Missing sampling data due to communication delays are supplemented using linear interpolation. For data highly susceptible to external interference, such as ambient temperature and humidity, a moving average filter with a window size of 5 sampling points is used for smoothing to eliminate random noise interference. After data cleaning, Z-score normalization is used to map various data of different dimensions to the [-1,1] numerical range, achieving data format normalization and generating a standardized operational dataset.

[0039] S3: Basic operational feature extraction. Based on a standardized operational dataset, a sliding time window method with a window size of 10 minutes is used to extract instantaneous operational features, including peak instantaneous charging power, average charging power, and rate of change of charging amount per unit time within each window. Time-dimensional features are extracted, including daily charging frequency, monthly charging saturation, and peak-hour charging percentage. Charging saturation is the ratio of actual charging time to available device time. Device status features are extracted, including offline time, number of failures, and normal operation rate within the maintenance cycle. Principal component analysis (PCA) is used for feature dimensionality reduction, retaining principal components with a cumulative variance contribution rate of no less than 85%. Finally, 12 core feature parameters, including average charging power, peak-hour charging percentage, and normal operation rate, are retained to form the core feature set.

[0040] S4: Charging Behavior Feature Clustering. Charging time preference, charging power demand, charging duration, and charging frequency from the core feature set are selected as clustering inputs. A density-based clustering algorithm is used for user behavior analysis, with a neighborhood radius of 0.5 and a minimum neighborhood sample size of 5. Through cluster analysis, charging users are divided into four categories: fixed-time users, random-demand users, high-power fast-charging users, and low-frequency maintenance users. Fixed-time users are mostly office commuters, charging during the morning rush hour; high-power fast-charging users are mostly ride-hailing drivers, with high charging power demand and short charging durations; low-frequency maintenance users are mostly local residents, charging infrequently but for longer durations per charge. The behavioral feature centers and cluster densities for different user categories are calculated, generating a charging behavior clustering model and a user classification label library. The label library is updated in real-time with newly added user data.

[0041] S5: Operational Load Situation Analysis. Based on a standardized operational dataset, a time series of charging pile operational loads was constructed. The average load, load fluctuation rate, and peak load were calculated for 1-hour, 1-day, and 1-month periods. Analysis revealed a three-peak distribution in the daily load of the charging pile group: weekday loads were significantly higher than weekend loads, and monthly loads peaked around holidays. Combined with a charging behavior clustering model, high-power fast-charging users were identified as the main factor causing load fluctuations, while users operating at fixed times exacerbated the load pressure during peak hours. A load situation heatmap was generated, visually displaying the load intensity of different charging piles at different times using color depth. Load change trend curves were plotted, and a quantitative assessment revealed poor load balance in the charging pile group, with some charging piles operating at full capacity and others idle during peak hours.

[0042] S6: Equipment Health Status Assessment. Combining equipment operating status data, failure frequency, and offline duration data from the core feature set, a health assessment indicator system is constructed, comprising four primary indicators: equipment operating status, failure frequency, offline duration, and maintenance records. The analytic hierarchy process (AHP) is used to determine the weights of each indicator: equipment operating status accounts for 40%, failure frequency for 30%, offline duration for 20%, and maintenance records for 10%. A weighted summation method is used to calculate the equipment health index for each charging pile. A health index above 0.8 is classified as healthy, 0.6-0.8 as sub-healthy, 0.4-0.6 as abnormal, and below 0.4 as faulty. The assessment identified three charging piles in an abnormal state due to prolonged high-load operation, primarily manifested as large fluctuations in charging power and increased offline frequency. This automatically triggered warnings, generating an equipment health status assessment report and fault risk tracing information. The tracing results showed that the fault risk was related to continuous high-load operation during peak hours.

[0043] S7: Operation optimization strategy generation. Based on charging behavior clustering models, load situation analysis results, and equipment health status assessment reports, three types of operation optimization strategies are formulated. Regarding load scheduling strategies, to address the issue of uneven load during peak hours, a staggered charging guidance mechanism is adopted, pushing off-peak charging discount information to users with fixed time slots. The output power of each charging pile is dynamically adjusted through a multi-pile power allocation algorithm to avoid long-term full-load operation of a single device. Regarding equipment maintenance optimization strategies, preventative maintenance plans are formulated for three abnormal-level charging piles, scheduled for maintenance during off-peak hours, and aging components are replaced. Regarding user charging guidance strategies, currently idle charging piles are recommended to users with random demand via a mobile app, and devices with sufficient power redundancy are prioritized for high-power fast-charging users. The output includes a set of operation optimization strategies containing optimal charging time suggestions, maintenance cycle adjustment schemes, and power allocation optimization parameters.

[0044] S8: Iterative optimization of the data model. One month after the implementation of the optimization strategy, actual operational data was collected as a new raw operational dataset. Steps S1 to S7 were repeated to quantitatively evaluate the effectiveness of the optimization strategy. Comparison revealed a significant improvement in load balance during peak hours, an average increase of 0.15 in the equipment health index, a reduction in average user waiting time for charging, and a decrease in the charging failure rate. The gradient descent algorithm was used to iteratively optimize the core feature extraction parameters, clustering algorithm parameters, and health assessment indicator weights. The initial learning rate of the gradient descent algorithm was set to 0.01. Iteration stopped after the improvement rate of the evaluation indicators fell below 1% for three consecutive times. The iteratively optimized data analysis and processing model and operational strategy library were generated and stored on a cloud server for shared use by the charging pile group.

[0045] Table 1: Comparison of Operational Indicators Before and After Optimization of Charging Pile Clusters in Commercial Areas Table 1 clearly demonstrates the optimization effect of this method on charging pile clusters in commercial areas. Before optimization, the load balance during peak hours was low, with some charging piles operating at full capacity while others were idle. The equipment health index was at a moderate level due to the long-term high load, resulting in long user waiting times and a high failure rate. Through multi-source data fusion analysis, user behavior clustering, and the implementation of precise optimization strategies using this method, the load balance during peak hours was significantly improved, the equipment health status was significantly improved, user waiting times were significantly shortened, the failure rate was reduced to a low level, and the proportion of equipment operating at full capacity was reasonably reduced. This fully demonstrates that this method can effectively solve the problems of uneven load, rapid equipment aging, and poor user experience in centralized charging pile clusters, thereby improving the operating efficiency and service quality of charging piles.

[0046] Example 2: Application of distributed charging piles in residential communities This embodiment is applied to a distributed charging pile system in a large residential community. The community has deployed 30 AC slow charging piles, distributed in the parking lots of various buildings, to serve the charging needs of residents' new energy vehicles. Users' charging behavior is mostly concentrated between 8 pm and 8 am the next day, with a long charging time. The system is characterized by dispersed equipment distribution, high maintenance difficulty, and concentrated load periods. The goal of this method is to achieve refined management, early warning of faults, and optimization of operating efficiency of the distributed charging pile system.

[0047] S1: Multi-source operational data acquisition. A metering module, a status monitoring module, and environmental sensors are installed on each distributed charging pile, and connected to the data management platform via the community's IoT network. The metering module collects real-time charging power, current, voltage, charging time, and cumulative charging amount. The status monitoring module collects equipment operating status, charging gun insertion / removal status, and grounding status. The environmental sensors collect ambient temperature, humidity, and dust concentration around the charging pile, and simultaneously access the community power grid's power supply voltage and frequency data. The data acquisition frequency is once per second. Dual sampling channels are used for redundant acquisition of safety-related parameters such as charging gun insertion / removal status and grounding status to ensure data reliability. The collected data is structured and stored after being timestamped and associated with the charging pile's location information, forming the original operational dataset.

[0048] S2: Data quality preprocessing involves cleaning the original dataset, using the Grubbs criterion to identify and remove abnormal charging power fluctuations caused by the moment the charging head is plugged in or unplugged, and using linear interpolation to fill in missing data caused by network interruptions. For data with small ambient temperature fluctuations but sporadic noise at night, a moving average filter is used for smoothing, with a window size of 5 sampling points. Z-score normalization is used to map data of different dimensions such as charging power, current, voltage, ambient temperature, and humidity to the [-1,1] interval, achieving data format normalization and generating a standardized dataset to ensure the accuracy of subsequent analysis.

[0049] S3: Basic operational feature extraction. Based on a standardized operational dataset, a sliding time window method with a window size of 10 minutes is used to extract instantaneous operational features, including peak instantaneous charging power, average charging power, and rate of change of charging amount per unit time. Time-dimensional features are extracted, including daily charging frequency, monthly charging saturation, and the proportion of nighttime charging (the ratio of charging time from 8 PM to 8 AM the next day to the total charging time). Equipment status features are extracted, including offline time, number of failures, normal operation rate within the maintenance cycle, and number of grounding anomalies. Principal component analysis (PCA) is used for feature dimensionality reduction, retaining principal components with a cumulative variance contribution rate of no less than 85%, ultimately determining 10 core feature parameters, including average charging power, nighttime charging proportion, normal operation rate, and number of grounding anomalies, forming a core feature set.

[0050] S4: Charging Behavior Feature Clustering. This section selects charging time preferences, charging power requirements, charging duration, and charging frequency from the core feature set as inputs. A density-based clustering algorithm is used for user behavior analysis, with a neighborhood radius of 0.5 and a minimum neighborhood sample size of 5. Cluster analysis categorizes users in the community into three types: long-duration nighttime users, short-duration daytime users, and random supplementary users. Long-duration nighttime users account for 75%, charging mostly at night for more than 8 hours. Short-duration daytime users mostly charge temporarily during the day for 1-3 hours. Random supplementary users have inconsistent charging times and lower frequency. The behavioral feature centers and cluster densities for different user categories are calculated, generating a charging behavior clustering model and a user classification label library, providing a basis for subsequent optimization strategy development.

[0051] S5: Operational Load Situation Analysis. Based on a standardized operational dataset, a time series of charging pile operational loads was constructed. The average load, load fluctuation rate, and peak load were calculated for periods of 1 hour, 1 day, and 1 month. Analysis revealed that the charging pile load in this community exhibits a typical nighttime single-peak distribution, with the peak load occurring from 10 PM to 2 AM the following day. During this period, most charging piles operate at full capacity, placing significant pressure on the power grid. Daytime loads are lower, with most charging piles idle. Combined with a charging behavior clustering model, long-term nighttime users were identified as the main factor causing the concentrated load during certain periods. A load situation heatmap was generated, visually displaying the load distribution of charging piles in different times and buildings. Load change trend curves were plotted, and a quantitative assessment revealed the uneven distribution of charging pile loads during different time periods, high pressure on the power grid at night, and low equipment utilization during the day.

[0052] S6: Equipment Health Status Assessment. This assessment combines core feature set data on equipment operating status, number of failures, offline duration, and number of grounding anomalies to construct a health assessment index system comprising five primary indicators: equipment operating status, number of failures, offline duration, maintenance records, and safety status. The analytic hierarchy process (AHP) is used to determine the weights of each indicator: equipment operating status (30%), number of failures (25%), offline duration (20%), maintenance records (10%), and safety status (15%). A weighted summation method is used to calculate the equipment health index for each charging pile, classifying them into four levels: healthy, sub-healthy, abnormal, and faulty. The assessment identified five charging piles at the abnormal level due to their long service life and prolonged full-load operation at night. These abnormalities were mainly manifested as decreased charging efficiency and occasional grounding anomalies, automatically triggering early warning prompts and generating equipment health status assessment reports and fault risk tracing information. The tracing results showed that the fault risk was related to decreased equipment heat dissipation efficiency and aging wiring.

[0053] S7: Operation optimization strategy generation. Based on charging behavior clustering models, load situation analysis results, and equipment health status assessment reports, targeted operation optimization strategies are formulated. Regarding load scheduling strategies, to address the issue of concentrated nighttime loads, a peak-valley electricity price guidance mechanism is adopted to push off-peak (0:00 to 6:00) charging discount information to long-duration nighttime users, encouraging users to charge during off-peak hours and reducing grid pressure. Regarding equipment maintenance optimization strategies, differentiated maintenance plans are formulated for five abnormal-level charging piles, prioritizing the replacement of aging lines and heat dissipation components, and scheduling maintenance during daytime low-load periods to avoid affecting users' nighttime charging. Regarding user charging guidance strategies, nearby idle charging piles are recommended to short-duration daytime users through community announcements and the APP, and real-time charging pile status query services are provided for users who randomly supplement their charging. The output includes a set of operation optimization strategies containing optimal charging time suggestions, maintenance cycle adjustment plans, and user guidance scripts.

[0054] S8: Iterative optimization of the data model. One month after the implementation of the optimization strategy, actual operational data is collected as a new raw operational dataset. Steps S1 to S7 are repeated to quantitatively evaluate the effectiveness of the optimization strategy. Comparison reveals a significant reduction in load pressure during nighttime peak hours, an average increase of 0.18 in the equipment health index, a more balanced utilization rate of charging piles, and a significant decrease in the failure rate. The gradient descent algorithm is used to iteratively optimize the core feature extraction parameters, clustering algorithm parameters, and health assessment index weights. The initial learning rate is set to 0.01. Iteration stops when the improvement rate of the evaluation index falls below 1% for three consecutive times. The iteratively optimized data analysis and processing model and operational strategy library are generated and stored in the community data management platform to achieve dynamic optimization management of distributed charging piles.

[0055] Table 2: Comparison of Operational Indicators of Charging Piles in Residential Communities Before and After Optimization Table 2 data fully demonstrates the optimization value of this method for distributed charging piles in residential communities. Before optimization, the nighttime peak load balance was low, the grid load peak was high, the equipment health index was less affected by long-term full-load operation at night, the utilization rate of charging piles was uneven, and the failure rate was high. Through multi-source data collection, precise feature extraction, user behavior clustering, and optimization strategy implementation using this method, the nighttime peak load balance was significantly improved, the grid load peak was significantly reduced, the equipment health status was significantly improved, the utilization rate of charging piles was evenly improved, and the failure rate was reduced to an extremely low level. This shows that this method can effectively solve the problems of difficult maintenance, concentrated load, and low operating efficiency of distributed charging piles, achieve refined management and optimized operation, improve the charging experience for residents, and reduce the operating pressure on the power grid and the operating costs of equipment.

[0056] refer to Figure 2This line graph illustrates the advantages of this invention in dynamically tracking equipment health status. As operating time increases, the health index assessed by various methods all show a downward trend, but the method of this invention exhibits the smallest decrease and the highest overall index. No assessment method relies entirely on the natural operating state of the equipment, failing to consider aging and environmental impacts, resulting in the fastest decline in the health index. Traditional assessment methods rely solely on static judgments using fixed indicators, failing to reflect the dynamic changes in equipment status. This invention, by constructing a scientific health assessment index system and dynamically correcting the health index based on operating time and environmental factors, can accurately depict the aging trend of equipment. Even after 3 years of operation, the health index remains above 0.68, far exceeding other methods, providing accurate basis for preventative maintenance, avoiding safety accidents caused by excessive equipment aging, and solving the problems of static and lagging health assessments in existing technologies.

[0057] refer to Figure 3 This figure illustrates the role of the iterative optimization mechanism of this invention in improving analysis accuracy. In the initial state, without data feedback optimization, the analysis error reached as high as 8.5%, mainly due to insufficient adaptability of the model parameters to the actual operating scenario. As the number of iterations increased, the model continuously optimized its parameters and algorithm logic by absorbing new operating data, leading to a rapid decrease in analysis error. After 6 iterations, the error dropped to 1.8%, and after 8 iterations, it stabilized. This trend indicates that the iterative optimization mechanism can effectively compensate for the limitations of the initial model, enabling the data analysis model to continuously adapt to the dynamic changes in the charging pile's operating status and continuously improve analysis accuracy. Compared to existing technologies without an iterative mechanism, this invention, through a closed-loop process of "analysis-verification-optimization," ensures that the data analysis results always maintain high accuracy, enabling long-term adaptation to the data analysis needs of charging piles under different operating stages and environmental conditions, significantly improving the practicality and long-term effectiveness of the method.

[0058] refer to Figure 4 This diagram illustrates the economic value of the invention. Traditional operation and maintenance methods rely on periodic manual inspections and post-failure repairs, which are not only costly in terms of labor but also exacerbate equipment wear and tear due to untimely fault handling, resulting in persistently high operation and maintenance costs. Especially in high-frequency usage scenarios such as highway service areas, annual operation and maintenance costs can reach as high as 185,000 yuan. This invention, through proactive fault warnings, preventative maintenance strategies, and load balancing scheduling, significantly reduces the failure rate and repair costs, while also reducing excessive equipment wear and tear. Operation and maintenance costs are reduced by approximately 50% across various scenarios. This advantage stems from the optimized allocation of operation and maintenance resources through precise data analysis, avoiding blind inspections and reactive repairs, making operation and maintenance work more targeted and efficient. Whether in centralized or decentralized charging pile scenarios, this invention can significantly reduce operation and maintenance costs, providing strong support for charging pile operators to improve economic efficiency, while extending equipment lifespan, thus possessing both economic and practical value.

[0059] refer to Figure 5This radar chart visually compares the overall performance of this invention with traditional analysis methods across five core dimensions, comprehensively showcasing the overall advantages of this invention. Traditional analysis methods scored below 6.5 points in all dimensions, especially in strategy implementation (only 5.0 points), reflecting a disconnect between their analysis results and actual operation and maintenance needs, poor model adaptability, and difficulty in handling different charging pile operating scenarios. This invention maintained high scores above 9.0 in all dimensions, with near-perfect accuracy in fault identification and health assessment, thanks to multi-source data fusion and precise feature extraction algorithms. High scores in load balancing and model adaptability demonstrate the effectiveness of the load scheduling strategy and the adaptability value of the iterative optimization mechanism. A 9.0 score in strategy implementation proves that the optimization strategy generated by this invention can be directly transformed into a practical solution. The balanced and high scores across all dimensions indicate that this invention achieves a comprehensive performance improvement in charging pile data analysis and processing, solving the problems of single-dimensional optimization and insufficient overall performance of traditional methods, and possessing superior comprehensive application value.

[0060] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for analyzing and processing operational data of new energy charging piles, characterized in that, Includes the following steps: S1: Multi-source operation data acquisition. Through the built-in metering module, communication module, environmental sensor and grid-side acquisition device of the charging pile, charging parameters, equipment status, environmental information and grid data are collected simultaneously and the collected data is structured and stored to form the original operation dataset. S2: Data quality preprocessing, outlier removal, missing value completion and noise smoothing are performed on the original running dataset, and then the data is standardized and normalized to generate a standardized running dataset. S3: Basic operation feature extraction. Based on a standardized dataset, instantaneous operation, time dimension and equipment status features are extracted. Redundant feature parameters are removed by feature dimensionality reduction algorithm, and core feature set that is strongly correlated with the charging pile's operation status and health is retained. S4: Charging behavior feature clustering. The core feature set is selected as input, and the density clustering algorithm is used to perform cluster analysis on user charging behavior, classify user types, and generate a charging behavior clustering model and user classification label library. S5: Operational load situation analysis. Based on standardized operation datasets, construct charging pile operation load time series, calculate multi-timescale load parameters, identify load influencing factors by combining clustering models, and quantitatively evaluate load balance and fluctuation characteristics. S6: Equipment health status assessment. Combine relevant equipment operation data to construct an assessment index system, determine the index weights and calculate the health index, and divide the equipment health status into four levels: healthy, sub-healthy, abnormal, and faulty. Automatically trigger early warning prompts for abnormal states and generate assessment reports and traceability information. S7: Optimization strategy generation. Based on charging behavior clustering, load situation analysis and equipment health assessment results, combined with charging pile geographical location and surrounding distribution data, it generates a set of optimization strategies for load scheduling, equipment maintenance and user charging guidance to address issues such as uneven load, equipment hazards and demand matching. S8: Iterative optimization of the data model. Repeat steps S1-S7 to evaluate the implementation effect, compare key indicators, optimize feature extraction, clustering algorithm and evaluation indicator weight parameters through gradient descent algorithm, iteratively improve the accuracy and applicability of the model, and generate the optimized model and strategy library.

2. The method for analyzing and processing operational data of new energy charging piles according to claim 1, characterized in that, It also includes a load balancing quantification calculation step, executed in S5, which quantifies the degree of load balancing of the charging piles using a formula: in For load balance, To count the number of sampling points within the statistical time window, For the first The charging power at each sampling point This represents the average charging power within the statistical time window.

3. The method for analyzing and processing operational data of new energy charging piles according to claim 1, characterized in that, It also includes a dynamic equipment health correction step, executed in S6, which dynamically adjusts the equipment health index using a formula that considers equipment runtime and environmental factors. The specific formula is as follows: in This is the dynamically adjusted equipment health index. For the initial calculation of the equipment health index, The cumulative runtime of the device. This is the difference between the ambient temperature and the standard operating temperature. This is the runtime impact coefficient. This represents the influence coefficient of ambient temperature.

4. The method for analyzing and processing operational data of new energy charging piles according to claim 1, characterized in that, In the S1 multi-source operation data acquisition process, the metering module uses a high-precision power metering chip with a measurement accuracy level of no less than 0.

5. The communication module supports 4G / 5G, Ethernet, and LoRa communication protocols. The environmental sensor uses a temperature and humidity composite sensor, and the geographical location information is obtained through the Beidou positioning module.

5. The method for analyzing and processing operational data of new energy charging piles according to claim 1, characterized in that, In the S2 data quality preprocessing step, the threshold for identifying abnormal data is set based on the equipment's factory parameters and historical normal operating data. Linear interpolation for missing value completion is used for continuous data. The window size for the moving average filtering method is set to 5 sampling points. Data standardization employs the Z-score standardization method, with the formula: ,in This is the original data. The mean of the data. This represents the standard deviation of the data.

6. The method for analyzing and processing operational data of new energy charging piles according to claim 1, characterized in that, In the S3 basic operation feature extraction step, the window size of the sliding time window method is set to 10 minutes. The rate of change of charging amount per unit time is calculated by the difference in charging amount between adjacent sampling points and the time interval. The feature dimensionality reduction algorithm adopts the principal component analysis algorithm, and retains the principal components with a cumulative variance contribution rate of not less than 85% as the core feature parameters.

7. The method for analyzing and processing operational data of new energy charging piles according to claim 1, characterized in that, In the S4 charging behavior feature clustering step, the neighborhood radius of the density clustering algorithm is set to 0.5, the minimum number of neighborhood samples is set to 5, the clustering density is calculated by the number of samples within a unit time window, the user classification label library is updated in real time, and newly added user behavior data is automatically assigned to the corresponding cluster category or forms a new category.

8. The method for analyzing and processing operational data of new energy charging piles according to claim 1, characterized in that, In the S5 operation load situation analysis steps, the calculation period for the average load, fluctuation rate, and peak load is set to 1 hour, 1 day, and 1 month, respectively. The load situation heat map is drawn using color depth to represent load intensity. The horizontal axis of the load change trend curve is time, and the vertical axis is charging power. The load change trend is displayed through a curve fitting algorithm.

9. The method for analyzing and processing operational data of new energy charging piles according to claim 1, characterized in that, In the S6 equipment health status assessment process, the health assessment indicator system includes equipment operating status indicators, failure frequency indicators, offline duration indicators, and maintenance record indicators. The judgment matrix of the analytic hierarchy process is constructed based on expert experience and historical data. Fault risk tracing information is associated with the fault occurrence time, corresponding operating characteristics, and environmental conditions to achieve fault cause location.

10. The method for analyzing and processing operational data of new energy charging piles according to claim 1, characterized in that, In the S7 operation optimization strategy generation steps, the load dispatch strategy includes peak-shaving and valley-filling strategies such as peak-shifting charging guidance, multi-pile power allocation, and interaction with the power grid; the equipment maintenance optimization strategy includes preventive maintenance plans based on health indices and rapid repair and replacement strategies for faulty equipment; and the user charging guidance strategy includes recommending the best charging pile, guiding users to idle charging piles, and providing charging discount information. The operation optimization strategy set is displayed through a visual interface, which supports operators to view and execute it directly. In the S8 data model iterative optimization steps, the indicators for quantitatively evaluating the implementation effect include the load balance improvement rate, equipment health index improvement rate, user average waiting time for charging reduction rate, and charging failure rate reduction rate. The initial learning rate of the gradient descent algorithm is set to 0.

01. The iteration stopping condition is that the improvement rate of the evaluation indicators is less than 1% for three consecutive times or the maximum number of iterations of 1000 is reached. The optimized operation strategy library is stored on the cloud server, supporting sharing and synchronous updates of multiple charging piles.

Images