Method for calculating accuracy of charging voltage and current frequency data of charging pile through standard deviation

Abstract

The invention provides a method for calculating the accuracy of charging voltage and current frequency data of a charging pile through a standard deviation, and the method comprises the steps: extracting continuous and non-zero voltage and current data segments from original time sequence data collected from the charging pile, and obtaining an effective charging data sequence; for the effective charging data sequence, a time sequence segmentation method is adopted, charging starting time and charging ending time are taken as boundaries, and a complete charging process data segment is divided; under the determined time granularity, the standard deviation of the voltage and current data is calculated, and a data fluctuation index is obtained; for the zero charging data, a rule-based classification algorithm is adopted, and in combination with an equipment operation log and an environment parameter, the reason category of the zero charging data is judged; and inputting the power comparison reference, the data fluctuation index and the fault judgment result into a preset comprehensive evaluation model, and calculating the operating state score of the charging pile.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular, to a method for calculating the accuracy of charging voltage, current, and frequency data of a charging pile based on standard deviation. Background Art

[0002] When analyzing the accuracy of charging voltage and current data of a charging pile, technical problems in data collection and processing are faced. First, the charging process of a charging pile is a dynamically changing process, and the voltage and current data show complex and variable characteristics in the time dimension. How to extract continuous non-zero effective charging data from a large amount of time-series data is a major challenge. Second, there are differences in power performance among different models of charging pile devices. How to select appropriate power data as a comparison reference requires comprehensive consideration of factors such as device model and charging power range. Third, when calculating the standard deviation of voltage and current data in the same region and time period, a reasonable spatio-temporal granularity needs to be determined, which should not only ensure the representativeness and comparability of the data, but also take into account the requirements of calculation efficiency and real-time performance. Finally, for charging pile data that is zero, how to identify whether it is caused by equipment failure or other reasons requires establishing a perfect fault judgment mechanism and rules, and combining with the analysis results of normal charging data to comprehensively evaluate the operating status and data quality of the charging pile. Summary of the Invention

[0003] The present invention provides a method for calculating the accuracy of charging voltage, current, and frequency data of a charging pile based on standard deviation, mainly including: Extract continuous non-zero voltage and current data segments from the original time-series data collected from the charging pile to obtain an effective charging data sequence; According to the charging pile device model and charging power range, match the corresponding power reference value from the preset device power performance database to determine the power comparison benchmark; For the effective charging data sequence, adopt a time series segmentation method, and divide the complete charging process data segment with the charging start and end times as boundaries; For the divided charging process data segments, adopt an adaptive time granularity method, and dynamically adjust the time interval according to the data fluctuation degree to obtain a time scale division scheme; Under the determined time granularity, calculate the standard deviation of the voltage and current data to obtain a data fluctuation index; For the charging data that is zero, adopt a rule-based classification algorithm, and combine the device operation log and environmental parameters to judge the cause category of the charging data being zero; Input the power comparison benchmark, data fluctuation index, and fault judgment result into a preset comprehensive evaluation model to calculate the operating status score of the charging pile; According to the running status scoring results, the charging piles are divided into three levels: normal, warning, and fault, and the final data quality evaluation results are obtained. The evaluation results are associated and stored with the device operation parameters and updated to the charging pile running status database.

[0004] The technical solution provided by the embodiment of the present invention may include the following beneficial effects: The present invention discloses a method for calculating the accuracy of charging voltage, current, and frequency data of charging piles based on standard deviation. This method extracts an effective charging data sequence from the original time-series data and matches the power reference value in combination with the device power performance database, realizing the precise division and analysis of the charging process. Innovatively, an adaptive time granularity algorithm and a sliding standard deviation algorithm are adopted to dynamically adjust the data analysis granularity and accurately calculate the data fluctuation index. In the case of zero charging data, the reasons are classified by combining the device operation log and environmental parameters. Finally, the present invention inputs the power comparison benchmark, data fluctuation index, and fault judgment result into the comprehensive evaluation model, and uses the weighted scoring method to calculate the running status score of the charging pile, realizing the precise evaluation and classification of the running status of the charging pile. This method significantly improves the accuracy and efficiency of the charging pile data quality evaluation and provides strong support for the intelligent management and optimized operation of charging facilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0005] Figure 1 It is a flowchart of a method for calculating the accuracy of charging voltage, current, and frequency data of charging piles by standard deviation according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0006] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative work shall fall within the protection scope of the present invention.

[0007] Such as Figure 1 , a method for calculating the accuracy of charging voltage, current, and frequency data of charging piles by standard deviation in this embodiment may specifically include: Step S101, extract the voltage and current data segments that are continuously non-zero from the original time-series data collected from the charging pile to obtain an effective charging data sequence.

[0008] Obtain the original charging pile data set with timestamp identification from the time series database. After the original data set is sorted by timestamp and filled with missing values by a preprocessor, a preprocessed data set is obtained; based on the preprocessed data set, the data with abnormal marks removed is transformed using the min-max normalization method to obtain a normalized data set; for the normalized data set, a density-based spatial clustering algorithm is used for grouping. The Euclidean distance between the data points and the cluster center is calculated through the set clustering radius parameter, and the data points beyond the clustering radius are marked as noise data. A sliding window is used to extract the data segments where both the voltage value and the current value are non-zero; the data segments are smoothed using a Kalman filter. The state vector of the Kalman filter contains two components, namely the voltage value and the current value, to obtain an effective charging data sequence.

[0009] Exemplarily, the charging pile timestamps, area identification values, and power values are collected from the time series database to form an original data set. The original data set is sorted by timestamp, filled with missing values, and marked with outliers through a data preprocessor to obtain a preprocessed data set. The standard deviations of three indicators, namely power value, voltage value, and current value, are calculated for the preprocessed data set. Judgments are made based on the standard deviation threshold interval statistically obtained from historical data. If any one of the three indicators in a data record exceeds the corresponding threshold interval, the record is marked as abnormal data. The data in the preprocessed data set with abnormal marks removed is subjected to standardization transformation using the min-max normalization method to obtain a normalized data set. For the normalized data set, temporal continuity verification is performed. If there are time breaks in the data sequence, the linear interpolation method is used to supplement the data. The density-based spatial clustering algorithm is used to group and cluster the normalized data set according to the area identification. The clustering radius parameter is set to the median of the statistically obtained data fluctuation range, and the Euclidean distance between the data points and the cluster center is calculated. The data points beyond the clustering radius are marked as noise data. The sliding window size is set to the standard charging duration interval value, and the data set after removing noise data is scanned slidingly to extract continuous data segments where both the voltage value and the current value within the window are non-zero. The Kalman filter is used to smooth the extracted data segments. The filter state vector contains two components, namely voltage value and current value, and the observation matrix is determined by the state transition at adjacent moments. After iterative filtering, a smoothed charging data sequence is obtained. An index structure is established based on the data segment timestamps, and the smoothed charging data sequence is stored in the time series database in blocks according to the area identification and timestamp order. During the charging pile time series data collection process, the data preprocessing stage usually includes multiple links. Among them, timestamp sorting ensures that the data is organized in chronological order, avoiding subsequent analysis deviations caused by out-of-order data. When there is a delay in the charging pile network communication, out-of-order arrival of data packets may occur. At this time, it is necessary to reorder according to the timestamps in the data packets. In practical applications, for charging data with a sampling interval of 1 minute, if the timestamp of a certain piece of data is 10:15:00 and the timestamp of the previous piece of data is 10:17:00, it indicates that the data packets are out of order and the data needs to be reordered. When marking abnormal data, the standard deviations of the power value, voltage value, and current value are calculated to determine whether the data fluctuation is abnormal. Taking power data as an example, if the power standard deviation threshold statistically obtained from historical data is 2.5 kWh, when the power value at a certain moment exceeds the range of plus or minus 2.5 kWh from the average value, the data point is considered abnormal. The normal range of the voltage value fluctuates within 10% above and below 220 volts, and the reasonable change range of the current value is determined according to the rated current of the charging device. In the normalization process, the min-max method is used to linearly map the data to the interval from 0 to 1, which can maintain the relative magnitude relationship of the data. For charging power data, if the original data range is between 0 and 50 kWh, it is converted to a value between 0 and 1 through normalization processing.When there is a missing time-series data, linear interpolation method is used for data repair to maintain data continuity. In density-based spatial clustering of applications with noise (DBSCAN), the clustering radius parameter directly affects the clustering effect. By statistically analyzing the fluctuation range of charging data, if the median of the data fluctuation is 0.8 kWh, the clustering radius is set to 0.8. When the distance between a data point and the nearest cluster center exceeds 0.8, it is determined as a noise point. This method can effectively identify abnormal fluctuations during the charging process. For the standard charging duration interval, a reasonable range is determined according to the charging pile type and battery capacity. In the fast charging mode, the standard charging duration is usually between 30 minutes and 60 minutes. Therefore, the sliding window size is set to 30 minutes, and the time interval for each movement is 1 minute. Data segments where the voltage and current are continuously non-zero are extracted within the window, and these segments represent the actual charging process. In the data smoothing process using the Kalman filter, the state vector includes two components: voltage and current, and the observation noise is set to 0.1 volts and 0.1 amperes. The state estimate is continuously corrected through iterative calculations to obtain a smooth charging curve. The filtered data can better reflect the overall trend of the charging process and eliminate the influence of short-term fluctuations. In the data storage section, the processed charging data is indexed according to the region identifier and timestamp. The index uses a B+ tree structure, where the leaf nodes store the actual data records and the non-leaf nodes store the index key values. Each data block size is set to 4 megabytes and contains continuous 10-minute charging records. Through reasonable index design and data block storage, the data query efficiency is improved.

[0010] According to the battery type and charging mode, the sampling frequency at different stages is determined, and at the same time, the key change points during the charging process are judged. If a sudden change in voltage or current is detected, the sampling frequency is increased to form the original time-series data.

[0011] According to the battery model identifier in the battery management unit, the rated capacity and charging curve parameters corresponding to the battery model identifier are obtained from the charging parameter database; a basic sampler is used to divide the charging stage according to the charging curve parameters, and sampling period values are set respectively for the trickle charging stage, constant current charging stage, and constant voltage charging stage to obtain the initial sampling sequence; the voltage change rate and current change rate of adjacent sampling points in the initial sampling sequence are calculated by a differential calculator, and mutation points are judged by comparing with the standard change rate threshold in the charging parameter database, and the sampling period value is adjusted for the mutation points; a data validator is used to verify the validity of the sampled sequence, and the verified sampled data forms the original time-series data.

[0012] Exemplarily, the battery model identifier in the battery management unit is read according to the charging interface protocol. Through the identifier matcher, three parameters of the rated capacity, charging voltage curve, and charging current curve of the battery of this model are obtained from the charging parameter database. The charging process is divided into a trickle stage, a constant current stage, and a constant voltage stage according to the voltage inflection point value and current inflection point value in the parameters. The basic sampler is used to set the initial sampling time period. The sampling period in the trickle charging stage is set to T, the sampling period in the constant current charging stage is set to 2T, and the sampling period in the constant voltage charging stage is set to 4T to generate an initial sampling sequence. The differential calculator is used to calculate the voltage change rate and current change rate of adjacent sampling points in the initial sampling sequence, and compare them with the standard change rate thresholds in each stage in the charging parameter database. If the voltage change rate or current change rate of a certain sampling point exceeds the corresponding threshold, then this sampling point is marked as a mutation point. For the marked mutation points, the adaptive sampler is used to increase the sampling frequency in the time interval before and after this point. If the current sampling period is nT, then the sampling period after the mutation point is detected is adjusted to nT / 4 until the change rates of 32 consecutive sampling points do not exceed the threshold. The data validator is used to check the validity of all sampling data. By comparing whether the voltage value and current value are within the value range specified in the charging parameter database, abnormal data points are eliminated, and the segmented linear interpolation method is used to supplement the data at the eliminated points. The window smoother is used to process the verified sampling data. The sliding window size is set to match the current sampling period, and the weighted average calculation is performed on the data within the window to generate the smoothed charging curve data. The smoothed charging curve data is sorted according to the timestamp, and the data is written into the time series database in a block storage manner. Each data block contains the complete sampling sequence before and after the mutation point, and each data block forms the original time series data. From the perspective of the characteristics of the charging parameters, the battery charging process mainly includes three stages: trickle, constant current, and constant voltage, and each stage has its unique charging characteristics. Taking the lithium iron phosphate battery as an example, the rated capacity of this type of battery is recorded as 100 ampere-hours, the nominal voltage is 3.2 volts, and the charging cut-off voltage is 3.65 volts in the charging parameter database. In the trickle charging stage, the charging current does not exceed 0.1 times the rated capacity, that is, 10 amperes. In this stage, the battery voltage rises slowly, from 2.5 volts to 3.0 volts gradually. When entering the constant current charging stage, the charging current rises to 0.5 times the rated capacity, that is, 50 amperes and remains constant, and the battery voltage rises rapidly, from 3.0 volts to 3.4 volts. In the constant voltage charging stage, the charging voltage remains unchanged at 3.65 volts, and the charging current gradually decreases until it drops below 5 amperes. According to these characteristic parameters, the initial sampling period is set. The trickle stage takes 60 seconds, the constant current stage takes 120 seconds, and the constant voltage stage takes 240 seconds. In the actual charging process, the mutation of voltage and current often occurs near the stage conversion point. By calculating the change rate of adjacent sampling points, when the voltage change rate exceeds 0.01 volts per second, or the current change rate exceeds 0.1 amperes per second, it is determined as a mutation point.Increase the sampling frequency in the regions before and after the mutation point. The original sampling period of 60 seconds is shortened to 15 seconds to more densely record the parameter changes during the phase transition. In the data verification process, the effective voltage range is set to 2.5 volts to 3.8 volts, and the effective current range is set to 0 amperes to 60 amperes. When the voltage value of a sampling point is found to be 4.2 volts, which is significantly outside the normal range, this abnormal data needs to be corrected by interpolation. Take 2 valid data points before and after the abnormal point and calculate a reasonable replacement value using linear interpolation. For data smoothing, when the sampling period is 120 seconds during the constant current charging phase, set the sliding window size to 5 sampling points, that is, a time span of 600 seconds. The data points within the window are processed using the weighted average method, where the data points closer to the center of the window have a greater weight, with the maximum weight being 0.4 and the minimum weight being 0.1. This smoothing method not only retains the overall trend of the data but also eliminates the influence of short-term fluctuations. In terms of data storage, each data block corresponds to a complete mutation interval, containing the data of the mutation point and 16 sampling points before and after it. For a data sequence with a 120-second sampling period, the rise and fall range of a data block is approximately 1 hour. The data blocks are associated through timestamps, facilitating subsequent retrieval of charging data for a specific interval according to the time range. This block storage method not only ensures the integrity of the mutation region data but also improves the data retrieval efficiency.

[0013] Step S102, according to the charging pile device model and the charging power range, match the corresponding power reference value from the preset device power performance database to determine the power comparison benchmark.

[0014] Obtain the device nameplate parameters through the charging pile communication interface, retrieve the device parameter record in the device power characteristic database according to the device nameplate parameters to obtain the rated power value, maximum power value, minimum power value, and standard power curve data of the device model; divide the device operation range into intervals according to the power characteristic data. If the device output power is lower than the rated power value threshold, it is divided into the minimum load interval; obtain the ratio of the current output power of the charging pile to the rated power through the load calculation formula to obtain the load level value, and the load level value is used to determine the standard power curve segment within the current load interval; use a curve fitter to perform piecewise linear fitting on the obtained standard power curve segment to calculate the power reference value at the current load level.

[0015] Exemplarily, the equipment nameplate parameters are obtained through the charging pile communication interface, and five basic parameters, namely, equipment manufacturer code, equipment model code, rated power value, output voltage range, and output current range, are extracted from the parameters, and the equipment parameter record is generated according to the extraction result. The equipment parameter record is retrieved from the equipment power characteristic database by using a parameter matcher, and four power characteristic parameters, namely, rated power value, maximum power value, minimum power value, and standard power curve data of the equipment of this model under different load conditions, are obtained from the database. The equipment operating range is divided into intervals according to the power characteristic parameters, and the minimum load interval is set to less than 30% of the rated power value, the normal load interval is set to 30% to 80% of the rated power value, and the maximum load interval is set to more than 80% of the rated power value. The power characteristic parameters are verified by a parameter checker, and abnormal parameter values ​​are eliminated by comparing whether the parameter values ​​are within the value range specified on the equipment nameplate, and the abnormal values ​​are replaced by the corresponding values ​​in the standard power curve data. The load level value is obtained by statistically calculating the ratio of the current output power of the charging pile to the rated power according to the load calculation formula, and the current load interval is determined by the load level value, and the corresponding standard power curve segment is obtained from the load interval. The obtained standard power curve segments are piecewise linearly fitted by a curve fitter to generate a piecewise linear equation group, and the power reference value under the current load level is calculated by the equation group. The power reference value update cycle and trigger conditions are set. When the actual running time reaches the update cycle or the load interval changes, the latest calculated power reference value is updated to the equipment operation parameter table. Taking the AC charging pile as an example, the equipment nameplate parameters include the manufacturer code M001, the equipment model code AC220-40, the rated power value of 40 kilowatts, the output voltage range of 180 volts to 242 volts, and the output current range of 0 to 120 amperes. These basic parameters constitute the core feature identification of the equipment and provide a basis for subsequent power characteristic matching. In the equipment power characteristic database, each model of charging pile has a corresponding power characteristic parameter set. Taking the AC220-40 model as an example, its power characteristic parameters include a rated power of 40 kilowatts, a maximum power of 44 kilowatts, and a minimum power of 8 kilowatts. The standard power curve data records the standard power output value of the model at different load levels, forming a complete power characteristic description. The division of load intervals directly affects the selection of power reference values. When the charging pile operates in the minimum load interval, that is, the output power is less than 12 kilowatts, the charging pile is working in a light load state and the power loss is relatively large. In the normal load interval, that is, when the output power is between 12 kilowatts and 32 kilowatts, the charging pile reaches the best working state. In the maximum load interval, that is, when the output power exceeds 32 kilowatts, the equipment temperature and efficiency need to be closely monitored.During the parameter verification process, if it is found that the power value exceeds the range specified on the equipment nameplate. For example, when the detected output power reaches 50 kW, which significantly exceeds the maximum power limit of 44 kW, the abnormal value needs to be replaced with the standard value corresponding to the load level in the standard power curve. In actual operation, the load level of the charging pile is dynamically changing. Suppose the measured output power at a certain moment is 30 kW, and the ratio to the rated power of 40 kW is 0.75, indicating that the charging pile is operating in the normal load range. The piecewise fitting of the standard power curve is very helpful for improving the calculation accuracy of the reference value. Within the normal load range, the power curve can be divided into multiple linear segments. For example, a set of linear equations is used between the load levels of 0.3 and 0.5, and another set of linear equations is used between 0.5 and 0.8. Through this piecewise fitting method, the power change law can be described more accurately. The update mechanism of the power reference value sets two trigger conditions, the time period and the load change. When the running time reaches the set update period, such as 15 minutes, the reference value is recalculated; or when the load level crosses from one range to another, such as from 0.75 to 0.25, the reference value update is also triggered. Each updated reference value is recorded in the equipment operation parameter table for subsequent power monitoring and management.

[0016] Step S103, for the effective charging data sequence, use the time series segmentation method to divide the complete charging process data segment with the charging start and end times as the boundaries.

[0017] Calculate the difference ratio using a change rate calculator based on the voltage value and the current value. If the difference ratio exceeds the preset change rate threshold, obtain the charging start point and end point through an extreme value detector; perform noise removal on the data between the charging start point and end point using a median filter, and repair the abnormal data points outside the preset range; calculate three characteristic parameters, namely the fluctuation period, the fluctuation amplitude, and the trend slope, from the repaired data using a data feature extractor; calculate the Euclidean distance between the feature vectors using a feature clustering device based on the characteristic parameters. If the Euclidean distance is less than the clustering threshold, merge the repeated charging process data through a data aligner to obtain the complete charging process data.

[0018] Exemplarily, according to the charging data sequence, a rate-of-change calculator is used to calculate the differential ratio of the voltage value and current value of adjacent sampling points. When the differential ratio exceeds a preset rate-of-change threshold, an extreme value detector is used to determine that this sampling point is a mutation point, and the charging start point and end point are marked from the set of mutation points. A median filter is used to remove noise from the charging data sequence. By comparing with a preset standard range of voltage and current, abnormal data points outside the range are excluded, and cubic spline interpolation is used to repair the data at the excluded points to generate a continuous data sequence. The continuous data sequence is segmented according to the charging start point and end point, and a time-span checker is used to verify the data segments to determine whether the duration of the data segments meets the requirement of the minimum charging duration and calculate whether the sampling point interval within the data segments meets the requirement of the minimum sampling density. For the data segments that meet the requirements, a data feature extractor is used to calculate three characteristic parameters, namely, the fluctuation period, fluctuation amplitude, and trend slope within the data segments, and a data segment feature vector is generated. A feature clustering algorithm is used to perform clustering analysis on the data segment feature vectors. By calculating the Euclidean distance between the feature vectors, the data segments with a distance less than the clustering threshold are grouped together to identify repeated charging processes. According to the clustering results, the data segments of the repeated charging processes are merged, and a data aligner is used to align the time axes of the data segments in the same group, and the standard charging process data is generated by weighted average calculation. A data integrity checker is used to verify the standard charging process data. By calculating the data continuity index and data reliability index, it is determined whether the data meets the storage requirements. For the standard charging process data that passes the verification, a block memory is used to write the data into the time-series database in chronological order and establish a data index structure. The segmentation process of the charging data sequence involves multiple key links, among which the calculation and judgment of the rate of change are particularly important. In practical applications, the rate of change of voltage and current during the electric vehicle charging process usually exhibits obvious stage characteristics. Taking the fast charging mode as an example, when charging starts, the voltage will rapidly rise from 0 volts to 250 volts within 1 second, and the rate of change reaches 250 volts per second, while the rate of change of voltage during normal charging is usually within 0.5 volts per second. By setting the rate-of-change threshold to 5 volts per second, the charging start point can be accurately captured. The median filtering process is very effective in removing sudden noise in the charging data. In the collected original data, there may be a situation where the voltage value jumps instantaneously. For example, the voltage value at a certain sampling point is 350 volts, while the adjacent points are all around 250 volts. Such abnormal values can be well eliminated by median filtering. At the same time, for the situation of data loss, cubic spline interpolation can maintain the continuity and smoothness of the data. The time-span verification of data segmentation is an important means to ensure the integrity of the charging process. In the actual charging process, a complete fast charge usually needs to last for more than 30 minutes, so the minimum time span can be set to 15 minutes. In terms of sampling density, considering the characteristics of parameter changes during the charging process, the sampling interval should not exceed 10 seconds, that is, at least 6 sampling points per minute.During the data feature extraction process, the fluctuation period reflects the regulation law of the charging current. Usually, during the constant current charging stage, the current fluctuation period is about 120 seconds. The fluctuation amplitude characterizes the stability of the charging process. Under normal circumstances, the current fluctuation amplitude does not exceed 3% of the rated value. The trend slope reflects the overall change trend of the charging process. For example, during the constant voltage stage, the current decreasing slope usually remains at about 1% per minute. Feature clustering analysis can identify similar charging processes. When the Euclidean distance between the feature vectors of multiple charging processes is less than 0.1, it can be considered that these charging processes are highly similar. By data alignment and weighted average, a more accurate standard charging process curve can be obtained. Data integrity verification includes index evaluation in multiple dimensions. The continuity index mainly examines whether the time interval between data points is uniform, and the allowed time drift does not exceed 1 second. The reliability index focuses on the validity of the data, requiring that the proportion of valid data points in the data segment exceeds 95%. When storing in blocks, each data block corresponds to a complete charging process. The typical data block size is about 500KB, containing about 3600 data points. By establishing an index structure based on timestamps, the charging data for a specific time period can be quickly located and retrieved. The index item contains information such as timestamps, data block positions, and data features, facilitating subsequent data analysis and processing.

[0019] Step S104: For the divided charging process data segments, adopt the adaptive time granularity method to dynamically adjust the time interval according to the data fluctuation degree, and obtain the time scale division scheme.

[0020] Obtain the charging process data segment, calculate the data mean and standard deviation according to the charging process data segment, and obtain the standardized data sequence through a standardization processor; calculate the change amplitude value of adjacent sampling points for the standardized data sequence. If the change amplitude value is greater than the preset threshold, it is marked as a high-fluctuation data segment. If the change amplitude value is less than the preset threshold, it is marked as a low-fluctuation data segment; use a self-organizing mapping network to receive the fluctuation characteristics of the high-fluctuation data segment and the low-fluctuation data segment, and calculate the optimal sampling interval value at the competitive layer nodes according to the fluctuation characteristics to obtain the adaptive sampling parameter table; calculate the correlation coefficient of the time dimension and the numerical dimension for the sampled data, and determine the minimum time granularity according to the size of the correlation coefficient to generate the time scale division scheme.

[0021] Exemplarily, a data preprocessor is used to standardize the charging process data segment. By calculating the mean and standard deviation of the data segment, the voltage value and current value are normalized and transformed to generate a standardized data sequence. A difference calculator is used to calculate the change amplitude between adjacent sampling points of the standardized data sequence. By setting a change amplitude threshold, the data sequence is segmented. When the change amplitude is greater than the threshold, it is marked as a high-fluctuation data segment. When the change amplitude is less than the threshold, it is marked as a low-fluctuation data segment. According to the fluctuation mark of the data segment, a sampling interval calculator is used to set a basic sampling interval for data segments with different fluctuation degrees. The high-fluctuation data segment uses the basic sampling interval T, and the low-fluctuation data segment uses the basic sampling interval 4T. A self-organizing mapping network is used to construct the mapping relationship between data fluctuation and sampling interval. The input layer nodes receive data fluctuation characteristics, and the competitive layer nodes calculate the optimal sampling interval to generate an adaptive sampling parameter table. For each data segment, a density calculator is used to count the number of sampling points per unit time. By comparing with a preset density threshold, data extraction is performed on high-density data segments, and data interpolation is performed on low-density data segments. A spatio-temporal correlation calculator is used to calculate the correlation coefficient in the time dimension and numerical dimension of the sampled data. According to the size of the correlation coefficient, the minimum time granularity is determined to generate a multi-layer time scale division scheme. According to the time scale division scheme, a recursive memory is used to hierarchically organize the data. A time index and a numerical index are established at each level to construct a hierarchical storage structure. A quality validator is used to check the integrity of the data after hierarchical storage. By calculating three indicators: data reduction error, time coverage rate, and spatial distribution uniformity, the data quality is verified. The adaptive time granularity optimization of the charging process data involves multiple levels of processing. In the standardization processing stage, the voltage value range in the original data is usually between 200 and 800 volts, and the current value range is between 0 and 400 amperes. Through normalization transformation, the data is mapped to the range of 0 to 1, making data with different dimensions comparable. The identification of data fluctuation characteristics is the key to time granularity optimization. Taking the fast charging process as an example, in the initial stage of charging when the voltage rises rapidly, the voltage change amplitude between adjacent sampling points can reach 0.1. The normalized value belongs to a high-fluctuation data segment at this time. In the constant voltage charging stage, the voltage change amplitude is usually less than 0.01 and is divided into a low-fluctuation data segment. By setting the change amplitude threshold to 0.05, data segments with different fluctuation characteristics can be accurately distinguished. The dynamic adjustment of the sampling interval directly affects the fineness of the data. When the basic sampling interval T is set to 1 second, the high-fluctuation data segment maintains a sampling interval of 1 second, while the low-fluctuation data segment uses a sampling interval of 4 seconds. Through the self-organizing mapping network, the input layer includes three feature nodes: change amplitude, duration, and data trend. The competitive layer is set with 16 nodes, corresponding to 16 different sampling interval configurations. Data density control ensures the balance between storage efficiency and information fidelity. In the high-fluctuation data segment, the number of sampling points per minute usually reaches 60, while in the low-fluctuation segment, it drops to about 15.The upper limit of the density threshold is set to 50 points per minute, and the lower limit is set to 10 points per minute. The data segments exceeding the upper limit are uniformly extracted, and the data segments below the lower limit are supplemented with data points through cubic spline interpolation. The spatiotemporal correlation analysis reflects the intrinsic structure of the data. The time correlation coefficient of the voltage data is usually above 0.95, indicating that the data at adjacent moments are highly correlated. However, the time correlation of the current data is weak, with a coefficient of about 0.8, requiring more fine-grained sampling. Based on the results of the correlation analysis, the time scale is divided into three layers: second, minute, and hour. The design of the hierarchical storage structure fully considers the data access characteristics. The second layer stores the original sampled data of the last hour, with a capacity of about 2MB. The minute layer stores the aggregated data of the last 24 hours, and each record contains three statistical values: mean, maximum, and minimum. The hour layer stores the statistical information of historical data, which is mainly used for long-term trend analysis. Data quality verification ensures the availability of data from multiple dimensions. The restoration error is measured by comparing the mean square error of the original data with the reconstructed data after hierarchical storage, and the error threshold is set to 0.01. The temporal coverage requires that the time interval between data segments does not exceed 2 times the minimum sampling interval. The spatial distribution uniformity is evaluated by calculating the coefficient of variation of the distance between adjacent sampling points, which is required to be less than 0.2.

[0022] Step S105, calculating the standard deviation of the voltage and current data at the determined time granularity to obtain a data fluctuation index.

[0023] The voltage and current time series data are obtained, and the abnormal point mark is obtained by calculating the median and interquartile range of the data sequence, and the local mean is used to replace the abnormal point to obtain the processed data sequence; the basic window size is set according to the processed data sequence, the window sliding step is obtained by calculating the variance of the data sequence, and the voltage and current standard deviation values ​​are calculated for the sliding window; the standard deviation value is processed by a differential calculator, and the time interval mark with severe fluctuations is obtained by setting the differential threshold; for the time interval mark, the weight coefficient is set according to the influence of the voltage and current in the charging process, and the characteristic vector is weighted summed to obtain the data fluctuation index value.

[0024] Exemplarily, a data preprocessor is used to detect outliers on voltage and current time series data. By calculating the median and interquartile range of the data sequence, a data point is marked as an outlier when it deviates from the median by more than a multiple of the interquartile range, and the outlier is replaced by a local mean. According to the processed data sequence, a normalization calculator is used to standardize the voltage and current data, and the data is mapped to a standard normal distribution by subtracting the mean and dividing by the standard deviation to generate a standardized data sequence. A window generator is used to set the basic window size, and the number of window sampling points is determined according to the spatiotemporal granularity parameters. The window sliding step is determined by calculating the variance of the data sequence to generate a sliding window sequence. For each sliding window, a standard deviation calculator is used to calculate the degree of discreteness of the voltage data and the current data in the window, and the window standard deviation value is obtained by taking the square root of the square difference of all data points. According to the window standard deviation value sequence, a difference calculator is used to generate a first-order difference sequence, and the time interval with drastic data fluctuations is identified by setting a difference threshold, and the fluctuation interval is segmented and marked. For the marked data segments, the feature extractor is used to calculate the mean, peak value, and variance of the voltage standard deviation sequence and the current standard deviation sequence to generate a feature vector. The weight allocator is used to set the weight coefficient according to the degree of influence of voltage and current in the charging process, and the feature vector is weighted and summed to generate the data fluctuation index value. The index verifier is used to verify the validity of the fluctuation index value. By calculating the stability and continuity of the index sequence, the reliability of the calculation result is judged, and the verified fluctuation index is written into the time series database. The fluctuation feature analysis of voltage and current data involves multiple processing links, among which outlier detection is the basis for ensuring data quality. In actual charging data, instantaneous abnormal fluctuations may occur, such as the voltage value suddenly jumping from the normal 350 volts to 500 volts. This abnormal point can be effectively identified by the interquartile range method. When the median of the data is 350 volts and the interquartile range is 20 volts, the data points exceeding 390 volts will be marked as abnormal, and then replaced by the mean of the five normal data points before and after. Standardization makes data of different dimensions comparable. In the original data, voltage values ​​are usually in the range of 200 to 800 volts, while current values ​​are in the range of 0 to 400 amperes. Through standardization transformation, the data is subtracted from the mean and then divided by the standard deviation. The converted data presents the characteristics of a standard normal distribution, with a mean of 0 and a standard deviation of 1. This processing method makes the fluctuation amplitude of voltage and current have the same metric. The setting of the sliding window directly affects the capture accuracy of the fluctuation characteristics. During the fast charging process, the change cycle of the voltage and current parameters is usually between 1 and 2 minutes. If the sampling interval defined by the spatiotemporal granularity parameter is 5 seconds, the window size is set to 24 sampling points, corresponding to 2 minutes of data. The window sliding step size is determined according to the data variance. When the data fluctuates violently, a smaller step size, such as 4 sampling points, is used; when the data is stable, a larger step size, such as 8 sampling points, is used. The standard deviation calculation reflects the degree of discreteness of the data.During the constant-voltage charging stage, the standard deviation of the voltage values within the window is usually less than 2 volts, showing high stability. When transitioning from the constant-current charging stage to the constant-voltage stage, the standard deviation of the current value may reach 20 amperes, showing obvious fluctuation characteristics. By calculating the standard deviation of consecutive windows, a time series reflecting the degree of parameter fluctuation can be obtained. The segmented identification of data fluctuations helps to accurately locate key time points. When the first-order difference value of the standard deviation sequence exceeds a set threshold, such as 0.5, it indicates that a significant change in data fluctuations has occurred. This change usually corresponds to the conversion point of the charging stage. For example, when switching from trickle charging to constant-current charging, the current will rise rapidly within a short period of time. The construction of the feature vector comprehensively considers multiple statistical indicators. The mean value of the voltage standard deviation sequence reflects the overall stability, the peak value represents the maximum degree of fluctuation, and the variance describes the unevenness of the fluctuation. The setting of the weight coefficient needs to consider the impact of voltage and current on the charging quality. Usually, the weight of voltage fluctuation is 0.4, and the weight of current fluctuation is 0.6 because the impact of current fluctuation on charging efficiency and battery life is more direct. The validity verification of the fluctuation index includes multiple dimensions. The stability is evaluated by calculating the coefficient of variation of the index sequence, and the coefficient of variation is required to be less than 0.2. Continuity requires that the change in the index values at adjacent times does not exceed 20%. The verified fluctuation indexes are organized according to the time stamps and stored in the time series database to provide data support for the subsequent optimization of the charging process.

[0025] Step S106, for the charging data that is zero, use a rule-based classification algorithm, combined with the device operation log and environmental parameters, to determine the category of the reason for the zero charging data.

[0026] Use a data pre-processor to obtain the voltage value and current value of the charging device. If it is detected that the voltage value and the current value of several consecutive sampling points are both zero at the same time, then use a zero-value duration calculator to count the start and end times and the duration of the zero-value data; according to the start and end times and the duration of the zero-value data, use a time feature calculator to obtain a zero-value feature vector; for the device operation log corresponding to the time period of the zero-value feature vector, extract the communication status, charging status, and fault status information of the device through a keyword matching method; according to the preset zero-value classification rules, use a decision tree classifier to analyze the zero-value feature vector and the communication status, charging status, and fault status information of the device to obtain the category label of the zero-value data.

[0027] Exemplarily, a data preprocessor is used to detect the charging data. When the voltage values and current values of three consecutive sampling points are both zero, it is marked as a zero-value data segment. The start and end times and the duration of each zero-value data segment are counted by a zero-value duration calculator. A time feature calculator is used to process the zero-value data segment. By calculating three time features, namely the occurrence time, the duration, and the occurrence frequency of the zero value, a zero-value feature vector is generated according to the distribution law of the time features. The device operation logs are extracted by a log parser. For the log content corresponding to the zero-value data segment, a keyword matching method is used to extract the device operation status information, including three types of information: communication status, charging status, and fault status. A running status analyzer is used to classify the information extracted from the logs. The status information is converted into status codes through a device status coding table, and a device status feature vector is generated according to the combination relationship of the status codes. The operating environment parameters of the charging device, including temperature value, humidity value, and grid voltage value, are obtained from the environmental monitoring database. A parameter threshold judge is used to verify the validity of the environmental parameters and generate an environmental status feature vector. A rule library manager is used to extract a zero-value classification rule set from a preset rule library. The rule set includes three types of rules: time rule, status rule, and environmental rule. The rules are encoded and converted by a rule parser. According to the zero-value classification rules, a decision tree classifier is used to comprehensively analyze the zero-value feature vector, the status feature vector, and the environmental status feature vector to generate a class label for the zero-value data. A classification result validator is used to evaluate the reliability of the class label. By calculating two indicators, namely the rule matching degree and the data integrity, the classification result is scored, and the classification result that passes the verification is written into the classification result database. The classification and judgment of the zero-value data of the charging pile involve the feature analysis of multiple dimensions. When the voltage and current values of three consecutive sampling points are both zero, it can be marked as a zero-value data segment. Taking a fast charging pile as an example, the normal sampling interval is 10 seconds. Therefore, 30 seconds of consecutive zero-value records constitute a complete zero-value data segment. In actual operation, the duration of the zero-value data segment may vary, ranging from several minutes to several hours. The time features of the zero-value data have typical patterns. Through statistics, it is found that the zero values caused by device failures usually occur suddenly during working hours and last for more than 2 hours; while the zero values caused by daily maintenance mostly appear during the night low period and last between 30 minutes and 1 hour; the zero values caused by communication interruptions occur randomly and have a short duration, usually within 10 minutes. The device operation logs contain rich status information. The status information extracted by keyword matching can be divided into three categories: communication status such as "communication interruption", "data timeout"; charging status such as "standby", "charging", "charging completed"; fault status such as "overtemperature protection", "overcurrent protection". These status information are converted into corresponding status codes. For example, communication interruption corresponds to status code 101, and overtemperature protection corresponds to status code 301.Environmental parameters have a significant impact on the operation of charging devices. The effective operating range of the temperature parameter is from -20 degrees Celsius to 55 degrees Celsius, the humidity parameter should not exceed 95%, and the grid voltage fluctuation should be within plus or minus 10% of the rated value. When the environmental parameters exceed the threshold range, the device may trigger a protection mechanism resulting in a charging interruption, and the zero-value data at this time should be classified as caused by environmental factors. The zero-value classification rule set contains multi-level judgment logic. The time rule mainly focuses on the time point and duration of the zero value occurrence. For example, "if the zero value occurs between 2 am and 4 am and the duration is less than 1 hour, it is classified as planned maintenance". The status rule focuses on the combination of device status codes. For example, "if the communication status code is 101 and there is no fault status code, it is classified as a communication fault". The environmental rule comprehensively considers environmental parameters and device status. For example, "if the temperature exceeds 55 degrees Celsius and the status code is 301, it is classified as over-temperature protection". The reliability assessment of the classification results uses a dual standard. The rule matching degree reflects the degree of compliance between data characteristics and classification rules, with a value range of 0 to 1. Usually, a value greater than 0.8 is considered a qualified matching degree. Data integrity measures the completeness of information related to the zero-value data segment, including three aspects: time characteristics, status information, and environmental parameters, and requires the information completeness rate to exceed 90%. Through a multi-dimensional classification method, the zero-value data of charging piles can be accurately classified into specific cause categories, such as device failures, planned maintenance, communication interruptions, environmental impacts, etc. This classification result is recorded in the classification result database for subsequent device operation and maintenance and optimization management.

[0028] Step S107, input the power comparison benchmark, data fluctuation index, and fault judgment result into a preset comprehensive evaluation model to calculate the operation status score of the charging pile.

[0029] Perform data cleaning based on the power comparison benchmark data, fluctuation index data, and fault judgment data, and obtain a normalized data sequence through the maximum-minimum value method; use a historical data analyzer to perform statistical operations on the normalized data sequence, and obtain the statistical characteristic values of the scoring indicators by calculating the historical mean and standard deviation of each indicator; use a multi-layer perceptron to perform non-linear transformation on the statistical characteristic values, and obtain the scoring result value through the input layer nodes receiving characteristic data, the hidden layer nodes processing, and the output layer nodes operating; set the scoring threshold range according to the statistical characteristic values of the scoring indicators, and correct the scoring values exceeding the threshold range by calculating the change rate of adjacent moments of the scoring result value.

[0030] Exemplarily, a data pre-processor is used to clean the data of the power comparison benchmark, data fluctuation index, and fault judgment result, and a pre-processed data sequence is generated by removing outliers and filling in missing values. According to the pre-processed data sequence, a standardization converter is used to map each index value to the range from zero to one, and the maximum-minimum method is used to perform normalization calculations on the power benchmark data, fluctuation index data, and fault judgment data. A historical data analyzer is used to statistically analyze the existing operation scoring records, and statistical characteristic values of the scoring indexes are generated by calculating the historical mean and standard deviation of each index. According to the hierarchical analysis structure, the weights of each layer of indexes are extracted from the weight database. A basic weight is set for the power benchmark data, a fluctuation weight is set for the fluctuation index data, and a fault weight is set for the fault judgment data. A multi-layer perceptron is used to perform combined calculations on the normalized data. The input layer nodes receive the normalized index data, the hidden layer nodes perform non-linear transformations, and the output layer nodes generate the scoring result. The scoring result is verified by a scoring validity checker. The upper and lower limits of the scoring threshold are set according to the statistical characteristic values of the historical data, and the scoring values outside the threshold range are corrected. A trend analyzer is used to perform continuity analysis on the scoring result. By calculating the change rate of the scoring values at adjacent moments, the scoring results with a change rate exceeding the preset range are smoothed. The corrected scoring result is written into the scoring database through a result memory, and four pieces of information, namely the scoring time, original scoring value, corrected scoring value, and scoring parameters, are recorded at the same time. The data pre-processing stage mainly deals with abnormal situations in the original data. For example, data points where the power comparison benchmark shows negative values or exceeds 150% of the equipment rated power need to be removed, and data breakpoints in the data fluctuation index need to be filled in by linear interpolation. The standardization conversion uses the maximum-minimum method to map indexes with different dimensions to a unified interval. Taking the power benchmark as an example, the original data range is from 0 to 60 kilowatts, and it is converted to a standardized value from 0 to 1 by subtracting the minimum value and then dividing by the value range. The original range of the fluctuation index is between 0 and 0.5, and the fault judgment result is an integer value from 0 to 5, both of which need to be subjected to similar standardization processing. The statistical characteristics of the historical data reflect the overall distribution law of the scoring. By analyzing the scoring records in the recent 30 days, the calculated mean of the power benchmark scoring is 0.85, and the standard deviation is 0.08; the mean of the fluctuation index scoring is 0.92, and the standard deviation is 0.05; the mean of the fault judgment scoring is 0.78, and the standard deviation is 0.12. These statistical values provide a basis for the validity verification of the scoring result. The weight allocation adopts the hierarchical analysis method, and the scoring indexes are divided into three layers. The first layer is the power benchmark, with a weight set to 0.4, reflecting the basic operation state of the equipment. The second layer is the fluctuation index, with a weight of 0.3, reflecting the stability of the operation. The third layer is the fault judgment, with a weight of 0.3, indicating the health status of the equipment. The hierarchical structure ensures a reasonable proportion of each index.The multi-layer perceptron structure consists of 3 nodes in the input layer, 8 nodes in the hidden layer, and 1 node in the output layer. After the input data is normalized, it enters the network and undergoes non-linear transformation through the hidden layer, finally generating a score value ranging from 0 to 100. The training samples are from the typical operating condition data in the historical operation records to ensure the accuracy of the scoring results. The validation of scoring effectiveness is based on threshold constraints in multiple dimensions. According to the statistical characteristic values, the upper limit of the score is set at 95 points and the lower limit at 60 points. If the scoring result exceeds this range, it needs to be corrected in combination with the historical mean. The continuity requirement of the score is that the change rate of the score at adjacent times does not exceed 20%. For mutation points, the moving average method is used for smoothing. The result storage adopts a structured data format. Each scoring record contains three basic pieces of information: scoring time, original score value, and corrected score value. At the same time, the key parameters in the calculation process are recorded, including normalization coefficients, weight values, threshold settings, etc.

[0031] Step S108, according to the scoring result of the operating state, divide the charging piles into three levels: normal, warning, and fault, to obtain the final data quality assessment result. Associate and store the assessment result with the device operating parameters, and update it to the charging pile operating state database.

[0032] Use a scoring interval divider to divide the scoring value of the charging pile operating state to obtain the charging pile operating state label. The operating state label includes a normal state label, a warning state label, and a fault state label; according to the operating state label, obtain the operating voltage value, operating current value, output power value, and device temperature value from the charging pile to generate an operating parameter vector, which is used to characterize the operating state of the charging pile; perform feature mapping on the operating parameter vector through a support vector machine to obtain a state prediction result; use a data associator to combine the operating state label, the operating parameter vector, and the state prediction result to generate a state description record, which is used to update the charging pile operating state database.

[0033] Exemplarily, a scoring range divider is used to divide the scoring range of the charging pile operation status into three intervals. When the scoring value is higher than the normal threshold, it is marked as the normal state. When the scoring value is between the warning threshold and the normal threshold, it is marked as the warning state. When the scoring value is lower than the warning threshold, it is marked as the fault state. A data validator is used to verify the validity of the scoring value. By calculating the standard deviation and mean of the scoring sequence, the fluctuation range of the scoring data is judged, and the scoring data outside the reasonable fluctuation range is marked. According to the status mark of the scoring data, a parameter extractor is used to obtain four real-time parameters, namely the operating voltage value, operating current value, output power value, and device temperature value, from the charging pile to generate an operating parameter vector. A parameter recorder is used to conduct a historical comparison of the operating parameter vector. By calculating the deviation value between the current parameter and the historical parameter, the parameter change trend is judged, and a parameter change feature sequence is generated. A support vector machine is used to predict the status of the parameter change feature sequence. The feature data is transformed into a high-dimensional space through kernel function mapping to generate a status prediction result. According to the status prediction result, a continuous monitor is used to continuously judge the status of the charging pile. By counting the continuous duration and conversion frequency of the status mark, a status persistence index is generated. A data associator is used to combine the status mark, operating parameter vector, status prediction result, and status persistence index to generate a complete status description record. The status description record is written into the operation status database through a database updater, and a time index structure is used to establish the mapping relationship between the status record and the timestamp to achieve real-time update of the status data. The hierarchical judgment of the charging pile operation status scoring involves the setting of multiple thresholds. The division of the scoring interval is based on a large number of historical data statistics. The normal threshold is set at 85 points, and the warning threshold is set at 70 points. When the scoring value is higher than 85 points, it indicates that all indicators of the charging pile are operating well. When the scoring value is between 70 and 85 points, it indicates that the device performance has a slight decline. When the scoring value is lower than 70 points, it indicates that the device has obvious faults. The validity verification of the scoring data is judged through statistical features. Under normal operating conditions, the standard deviation of the scoring sequence usually remains within 3 points, and the mean is around 90 points. If the scoring data fluctuates violently during a certain period, such as the standard deviation suddenly rising to 10 points, the data for that period needs to be focused on and verified. The acquisition of operating parameters reflects the real-time working status of the charging pile. The voltage parameter fluctuates within no more than 5% above and below the rated value of 350 volts. The current parameter varies between 0 and 120 amperes with the charging stage. The output power is determined by the product of voltage and current. The device temperature does not exceed 55 degrees Celsius under normal operating conditions. These parameters together constitute the basic characteristics of the device operation status. The analysis of the parameter change trend can detect potential problems early. By comparing the deviation between the current parameter and the historical data, the deviation threshold is set at 10% of the historical mean. Taking the temperature parameter as an example, if the current temperature is 6 degrees higher than the historical mean and shows a continuous upward trend, this abnormal fluctuation needs to be focused on.In state prediction, the support vector machine adopts a Gaussian kernel function, and the input features include parameter deviation values and change rates. The training samples are from the typical operating condition data in the historical operation records, including various states such as normal operation, performance decay, and fault occurrence. The kernel function maps low-dimensional features to a high-dimensional space to improve the accuracy of state classification. When the charging pile is in a warning state for 3 consecutive time points, or the number of warning states exceeds 5 times within 24 hours, it indicates that the device performance has a continuous decline. The fault state usually shows suddenness and needs to be processed immediately. The data associated storage adopts a multi-dimensional index structure, with the timestamp as the main index, supporting fast retrieval according to the time range. The state description record contains information in four dimensions: score value, state flag, parameter vector, and prediction result. The secondary index includes device number, state category, and parameter type, facilitating data analysis from different dimensions.

[0034] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the present invention to only the specific implementation manners. Obviously, according to the content of this specification, many modifications and changes can be made. These embodiments are selected and specifically described in this specification to better explain the principle and practical application of the present invention, so that those skilled in the art in the relevant technical field can well understand and utilize the present invention. The present invention is only limited by the claims and their full scope and equivalents.

Claims

1. A method for calculating the accuracy of charging voltage, current and frequency data of a charging pile by standard deviation, characterized in that: The method comprises: From the original time series data collected from the charging pile, extract the continuous non-zero voltage and current data segments to obtain the effective charging data sequence; According to the charging pile equipment model and charging power range, the corresponding power reference value is matched from the preset equipment power performance database to determine the power comparison benchmark; For the effective charging data sequence, the time series segmentation method is used to divide the complete charging process data segment with the charging start and end time as the boundary; For the divided charging process data segments, an adaptive time granularity method is used to dynamically adjust the time interval according to the degree of data fluctuation to obtain a time scale division scheme; At a certain time granularity, the standard deviation of voltage and current data is calculated to obtain the data fluctuation index; For zero charging data, a rule-based classification algorithm is used to determine the cause of zero charging data by combining equipment operation logs and environmental parameters. Input the power comparison benchmark, data fluctuation index and fault judgment results into the preset comprehensive evaluation model to calculate the charging pile operation status score; According to the operation status scoring results, the charging piles are divided into three levels: normal, warning and fault, and the final data quality assessment results are obtained. The assessment results are associated with the equipment operation parameters and stored, and updated to the charging pile operation status database.

2. The method according to claim 1, characterized in that The method of collecting the original time series data from the charging pile and extracting the continuous non-zero voltage and current data segments to obtain the effective charging data sequence includes: obtaining the original data set of the charging pile with timestamp identification from the time series database, and obtaining the preprocessed data set after the original data set is sorted by timestamp and missing value is filled by the preprocessor; Based on the preprocessed data set, a minimum-maximum normalization method is used to transform the data with abnormal markers removed to obtain a normalized data set; The normalized data set is grouped using a density-based spatial clustering algorithm, the Euclidean distance between the data point and the cluster center is calculated by using a set cluster radius parameter, the data points beyond the cluster radius are marked as noise data, and a sliding window is used to extract data segments whose voltage and current values ​​are not zero; the data segments are smoothed using a Kalman filter, and the state vector of the Kalman filter includes two components, the voltage value and the current value, to obtain a valid charging data sequence; and the method also includes: determining the sampling frequency of different stages according to the battery type and the charging mode, and judging the key change points in the charging process. If a sudden change in voltage or current is detected, the sampling frequency is increased to form original time series data.

3. The method according to claim 2, characterized in that The sampling frequency of different stages is determined according to the battery type and charging mode, and the key change point in the charging process is judged. If a sudden change in voltage or current is detected, the sampling frequency is increased to form original time series data, including: according to the battery model identifier in the battery management unit, the rated capacity and charging curve parameters corresponding to the battery model identifier are obtained from the charging parameter database; A basic sampler is used to divide the charging stage according to the charging curve parameters, and a sampling period value is set for the trickle charging stage, the constant current charging stage, and the constant voltage charging stage respectively to obtain an initial sampling sequence; The voltage change rate and the current change rate of adjacent sampling points in the initial sampling sequence are calculated by a differential calculator, and the mutation point is determined by comparison with the standard change rate threshold in the charging parameter database, and the sampling period value is adjusted according to the mutation point; the validity of the sampled sequence is verified by a data verifier, and the verified sampling data forms the original time series data.

4. The method according to claim 1, characterized in that: The method of matching the corresponding power reference value from a preset equipment power performance database according to the charging pile equipment model and the charging power range to determine the power comparison benchmark includes: obtaining equipment nameplate parameters through the charging pile communication interface, retrieving equipment parameter records in the equipment power characteristic database according to the equipment nameplate parameters, and obtaining the rated power value, maximum power value, minimum power value and standard power curve data of the equipment model; dividing the equipment operating range into intervals according to the power characteristic data, and if the equipment output power is lower than the rated power value threshold, dividing it into the minimum load interval; The ratio of the current output power of the charging pile to the rated power is obtained through the load calculation formula to obtain the load level value, which is used to determine the standard power curve segment within the current load range; a curve fitter is used to perform piecewise linear fitting on the obtained standard power curve segment to calculate the power reference value under the current load level.

5. The method according to claim 1, characterized in that The effective charging data sequence is segmented by a time series segmentation method, with the charging start and end time as the boundary, to divide the complete charging process data segment, including: using a rate of change calculator to calculate the differential ratio according to the voltage value and the current value, if the differential ratio exceeds the preset rate of change threshold, obtaining the charging start point and the end point through an extreme value detector; using a median filter to remove noise from the data between the charging start point and the end point, and repairing abnormal data points that exceed the preset range; using a data feature extractor to calculate three feature parameters of fluctuation period, fluctuation amplitude and trend slope from the repaired data; using a feature clusterer to calculate the Euclidean distance between feature vectors according to the feature parameters, if the Euclidean distance is less than the clustering threshold, merging the repeated charging process data through a data aligner to obtain complete charging process data.

6. The method according to claim 1, characterized in that The charging process data segments after division are subjected to an adaptive time granularity method, and the time interval is dynamically adjusted according to the degree of data fluctuation to obtain a time scale division scheme, including: obtaining the charging process data segments, calculating the data mean and standard deviation according to the charging process data segments, and obtaining a standardized data sequence through a standardized processor; calculating the change amplitude value of adjacent sampling points for the standardized data sequence, if the change amplitude value is greater than a preset threshold, marking it as a high-fluctuation data segment, and if the change amplitude value is less than the preset threshold, marking it as a low-fluctuation data segment; Using a self-organizing map network to receive the fluctuation characteristics of the high-fluctuation data segment and the low-fluctuation data segment, and calculating the optimal sampling interval value at the competition layer node according to the fluctuation characteristics to obtain an adaptive sampling parameter table; The correlation coefficient between the time dimension and the numerical dimension of the sampled data is calculated, the minimum time granularity is determined according to the size of the correlation coefficient, and a time scale division scheme is generated.

7. The method according to claim 1, characterized in that The method of calculating the standard deviation of voltage and current data at a determined time granularity to obtain a data fluctuation index includes: obtaining voltage and current time series data, obtaining abnormal point marks by calculating the median and interquartile range of the data sequence, and replacing the abnormal points with local mean values ​​to obtain a processed data sequence; setting a basic window size according to the processed data sequence, obtaining a window sliding step size by calculating the variance of the data sequence, and calculating the voltage and current standard deviation value for the sliding window; A differential calculator is used to process the standard deviation value, and a time interval marker with severe fluctuation is obtained by setting a differential threshold; For the time interval mark, a weight coefficient is set according to the influence of the voltage and current in the charging process, and the characteristic vector is weighted and summed to obtain the data fluctuation index value.

8. The method according to claim 1, characterized in that The method comprises: using a data preprocessor to obtain the voltage value and current value of the charging device, and if it is detected that the voltage value and the current value of a plurality of consecutive sampling points are zero at the same time, using a zero value duration calculator to count the start and end time and duration of the zero value data; According to the start and end time and duration of the zero-value data, a zero-value feature vector is obtained through a time feature calculator; for the device operation log of the time period corresponding to the zero-value feature vector, the communication status, charging status and fault status information of the device are extracted through a keyword matching method; according to the preset zero-value classification rules, a decision tree classifier is used to analyze the zero-value feature vector and the communication status, charging status and fault status information of the device to obtain a category label for the zero-value data.

9. The method according to claim 1, characterized in that: The power comparison benchmark, data fluctuation index and fault judgment result are input into a preset comprehensive evaluation model to calculate the charging pile operation status score, including: performing data cleaning according to the power comparison benchmark data, fluctuation index data and fault judgment data, and obtaining a normalized data sequence by the maximum and minimum method; using a historical data analyzer to perform statistical operations on the normalized data sequence, and obtaining a statistical characteristic value of the scoring index by calculating the historical mean and standard deviation of each indicator; using a multi-layer perceptron to perform nonlinear transformation on the statistical characteristic value, and obtaining a scoring result value by receiving characteristic data through input layer nodes, processing through hidden layer nodes and calculating through output layer nodes; setting a scoring threshold range according to the statistical characteristic value of the scoring index, and correcting the scoring value that exceeds the threshold range by calculating the adjacent moment change rate of the scoring result value.

10. The method according to claim 1, characterized in that According to the operation status scoring result, the charging pile is divided into three levels: normal, warning and fault, to obtain the final data quality evaluation result, the evaluation result is associated with the equipment operation parameter and stored, and updated to the charging pile operation status database, including: using a scoring interval divider to divide the charging pile operation status score value into intervals to obtain the charging pile operation status mark, the operation status mark includes a normal state mark, a warning state mark and a fault state mark; According to the operation status mark, the operation voltage value, the operation current value, the output power value and the equipment temperature value are obtained from the charging pile to generate an operation parameter vector, and the operation parameter vector is used to characterize the operation status of the charging pile; the operation parameter vector is feature mapped by a support vector machine to obtain a state prediction result; the operation status mark, the operation parameter vector and the state prediction result are combined by a data associator to generate a state description record, and the state description record is used to update the charging pile operation status database.

Images