Vehicle pile interaction safety online early warning method based on charging risk assessment

Abstract

The invention discloses a vehicle pile interaction safety online early warning method based on charging risk assessment, and the method comprises the steps: starting from the practical application of a conventional early warning method, considering the particularity of the operation state of a charging pile for the sensitivity of each index, constructing a charging risk assessment mechanism for an electric vehicle charging pile, and carrying out the early warning. According to the method, whether intervention control needs to be carried out at a higher degree or not is judged by considering follow-up load fluctuation and variability of actual operation conditions, meanwhile, a judgment result is checked by analyzing similar fragments of historical fault data features, and the judgment reliability is better guaranteed. According to the method, the situation that the charging pile cannot obtain efficient and accurate early warning in the operation process due to the fact that the operation characteristics of the charging pile are not fully utilized, and safety accidents are caused can be well solved, the control precision and reliability are improved, and a more refined control method is provided for online intelligent early warning control of the charging pile.

Description

Technical Field

[0001] The present invention belongs to the field of electric vehicle charging pile regulation and control, and in particular relates to a vehicle-pile interactive safety online early warning method based on charging risk assessment. Background Art

[0002] With the maturity of Internet of Things technology, charging piles can be connected to cloud platforms through sensors to realize the collection and transmission of real-time data of key indicators such as current, voltage, and temperature, saving the cost of manual patrol and monitoring and facilitating the issuance of accurate commands from the background port.

[0003] In recent years, electric vehicles have become increasingly popular, and the number and frequency of use of charging piles have increased significantly, while the risk of accidents during the use of charging piles by users has increased simultaneously. Traditional early warning methods are mainly about monitoring and identifying potential risks, including regular maintenance and inspection, manual inspections, threshold monitoring and other methods. However, charging piles themselves are devices that are highly sensitive to changes in indicators such as current. Traditional early warning methods cannot quickly and accurately detect potential fault risks and locate dangerous indicators. At the same time, even if the risk of a fault is confirmed through traditional early warning methods, it is impossible to provide timely feedback, which may even lead to accidents that endanger personal safety. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a vehicle-pile interactive safety online early warning method based on charging risk assessment.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] A method for online early warning of vehicle-pile interaction safety based on charging risk assessment includes the following steps:

[0007] S1. Obtain historical data of all indicators of charging status of the charging pile;

[0008] S2. Divide the historical data of all indicators into a normal set and a fault set, perform correlation analysis on the historical data of all indicators, and select the top five evaluation indicators with the highest correlation with the health status of the charging pile as the first-level indicators;

[0009] S3. Set two secondary indicators, amplitude and volatility, for all primary indicators;

[0010] S4. Construct a judgment matrix of the first-level indicators based on the correlation coefficients of the first-level indicators, and determine the weights of each first-level indicator based on the judgment matrix of the first-level indicators;

[0011] S5. Calculate the health value of each fault data in the fault set using the weight of the first-level indicator; and use the minimum value of the health value of each fault data as the critical health value of the charging state of the charging pile;

[0012] S6. Construct a fault gene set using secondary indicators;

[0013] S7. Obtain the real-time data of each primary indicator under the charging state of the charging pile, and use the real-time data of each primary indicator to predict the predicted values of each primary indicator in the future;

[0014] S8. Conduct a correlation analysis by combining the real-time data of each primary indicator and the predicted values of each primary indicator, and update the correlation coefficient of the primary indicators; adjust the judgment matrix of the primary indicators, update the weights of each primary indicator, and calculate the predicted health value based on the updated weights;

[0015] S9. If the critical health value is less than the predicted health value, it is considered that the charging pile is in normal operation and no intervention is required; otherwise, it is considered that the charging pile is in the marginal operation state, and the predicted values of each primary indicator are matched and verified with the fault gene set. If the verification passes, go to S10; otherwise, return to S8;

[0016] S10. Query the fault data corresponding to the health value with the smallest difference between the health value of each fault data in the fault set and the predicted health value; and calculate the similarity between the primary indicators of this fault data and the predicted values of each primary indicator in the future;

[0017] S11. If the similarity is greater than the similarity threshold, it is determined that the charging pile status is a high-risk level, and a "power-off warning" command is sent back to the background; otherwise, the charging pile status is a low-risk level, and a "manual intervention" command is sent back to the background.

[0018] To optimize the above technical solution, the specific measures taken also include:

[0019] Further, in S2, the specific process of dividing the historical data of all indicators into a normal set and a fault set is as follows:

[0020] Perform fuzzy clustering on the historical data of all indicators, calculate and record the clustering centers, and continuously update the membership degrees of the clustering centers by calculating the Euclidean distances between each historical data and each clustering center until the maximum number of iterations is reached;

[0021] Compare all the indicators of each cluster of historical data with the rated operating state, and divide the historical data of all indicators into a normal set and a fault set.

[0022] Further, in S2, the correlation analysis of the historical data of all indicators is specifically a Pearson correlation analysis, and in S4, the correlation coefficient is the Pearson correlation coefficient.

[0023] Further, S4 is specifically:

[0024] The correlation coefficients of the first-level indicators are divided into four levels of importance in ascending order of absolute value. The levels of importance include: equally important, slightly important, significantly important, and strongly important. The judgment matrix of the first-level indicators is constructed as follows:

[0025]

[0026] In the formula, d ii represents the importance of indicator i with respect to indicator i, equal to 1, indicating equally important; d ij represents the importance of indicator j with respect to indicator i; d ji represents the importance of indicator i with respect to indicator j;

[0027] Based on the judgment matrix of the first-level indicators, the weights w = [w 1 , w 2 , w 3 , w 4 , w 5 of each first-level indicator are determined through the analytic hierarchy process. T , where w represents the weight vector composed of the weights of the five first-level indicators, and w i (i = 1, 2,..., 5) represents the weight of the i-th first-level indicator.

[0028] Furthermore, in S5, the specific method for calculating the health values of each fault data in the fault set using the weights of the first-level indicators is as follows:

[0029] Regarding the rated operating state of the charging pile as the standard operating state A = [a 1 , a 2 , a 3 , a 4 , a 5 . T , where a i (i = 1, 2,..., 5) represents the rated value of the i-th first-level indicator when the charging pile is operating normally;

[0030] The health values of each fault data in the fault set are calculated using the weights of the first-level indicators and the rated values of the first-level indicators. The formula is as follows:

[0031]

[0032] In the formula, bw k represents the health value of the k-th group of fault data, w i represents the weight of the i-th first-level indicator, b ki (i = 1, 2,..., 5) represents the fault data of the i-th first-level indicator in the k-th group of fault data, a i represents the rated value of the i-th first-level indicator when the charging pile is operating normally; the k-th group of fault data is denoted as B k= [b k1 , b k2 , b k3 , b k4 , b k5 T .

[0033] Furthermore, S6 is specifically as follows:

[0034] S61. Extract the fault data with the largest volatility difference of the first-level indicators corresponding to the fault set and the normal set;

[0035] S62. Convert the fault data into waveforms for visual analysis, specifically including: dividing the fault data into bands of three different waveforms of linear increase, sine, and linear decrease, recording the volatility and amplitude corresponding to different waveform bands, and calculating the time from the starting point of each waveform band to the fault occurrence point as the fault occurrence time;

[0036] The fault gene fragment consists of amplitude, volatility, and fault occurrence time, and is expressed by the formula as follows:

[0037] wave u = [A u , fault_t u , f u

[0038] In the formula, wave u represents the fault gene fragment of waveform u, where u = 1 indicates that the waveform is linearly increasing, u = 0 indicates sine, u = -1 indicates linearly decreasing, fault_t u represents the fault occurrence time of waveform u, and f u represents the volatility of waveform u;

[0039] f u = c u / fault_t u

[0040] In the formula, c u represents the number of zero-crossing points of waveform u within the fault occurrence time;

[0041] S63. Compose the fault gene fragments into a fault gene set.

[0042] Furthermore, in S7, the method for predicting the predicted values of future first-level indicators using the real-time data of each first-level indicator is the extreme gradient boosting tree algorithm.

[0043] Furthermore, in S8, calculating the predicted health value based on the updated weights is specifically as follows:

[0044] ​​

[0045] In the formula, β represents the predicted health value, and w' i represents the updated weight of the i-th primary indicator, and p i represents the predicted value of the i-th primary indicator in the future, and a i represents the rated value of the i-th primary indicator when the charging pile operates normally.

[0046] Furthermore, in S9, the specific process of matching and verifying the predicted values of each primary indicator with the fault gene set is as follows:

[0047] Analyze the fluctuation characteristics of the predicted values of each primary indicator. The fluctuation characteristics include waveform, amplitude, fault occurrence time, and volatility, and match the fluctuation characteristics with the fault gene fragments in the fault gene set;

[0048] The standard for passing the verification is that a matching error within ±5% is regarded as passing the inspection.

[0049] Furthermore, in S10, the calculation formula for the similarity is as follows:

[0050]

[0051] In the formula, ρ represents the similarity between the primary indicators of the fault data and the predicted values of the future primary indicators, P = [p 1 , p 2 , p 3 , p 4 , p 5 T , P represents the predicted future operating state of the charging pile, and p i (i = 1, 2,..., 5) represents the predicted value of the i-th primary indicator in the future, B represents the fault data, Cov(P, B) represents the covariance between the future operating state of the charging pile and the fault data, DP represents the variance of the future operating state P of the charging pile, and DB represents the variance of the fault data B.

[0052] The beneficial effects of the present invention are as follows: The present invention can improve the prediction accuracy, judgment reliability, and reduce the risk of safety accidents; aiming at the personalized characteristics of the charging pile's own structure and operating state and the operability of the charging state evaluation, the present invention proposes a charging risk assessment mechanism based on key indicators of the charging state; aiming at the special characteristics of the charging pile station area environment and instructions, the present invention proposes a multi-level sub-control early warning model, and introduces a fault gene fragment data set as the verification of the judgment result, and selects different control commands through multi-level analysis. Description of the Drawings

[0053] Figure 1 is the overall flowchart of the present invention.

[0054] ​Figure 2 It is the flowchart of the method for constructing the fault gene set of the present invention.

[0055] Figure 3 It is the flowchart of the intelligent early warning model of the present invention. Specific embodiments

[0056] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0057] Embodiment 1

[0058] The present invention proposes an online safety early warning method for vehicle-pile interaction based on charging risk assessment. The flowchart of this method is as Figure 1 shown, and it includes the following steps:

[0059] S1. Obtain the historical data of all indicators of the charging status of the charging pile; for example, the temperature, voltage, current, insulation resistance, etc. of the charging pile.

[0060] S2. Divide the historical data of all indicators into a normal set and a fault set, conduct a correlation analysis on the historical data of all indicators, and screen out the top five evaluation indicators with the highest correlation with the health status of the charging pile as the primary indicators; specifically as follows:

[0061] Conduct fuzzy clustering on the historical data of all indicators, calculate and record the clustering center:

[0062] centre(q,m)=∑(d(q,c) n f(q,m)) / ∑d(q,c) n

[0063] Among them, centre(q,m) represents the clustering center position of the data point q in the feature dimension m, d(q,c) n is the membership degree of the data point q belonging to the cluster k, f(q,m) is the value of the data point q in the feature dimension m, n is the fuzzy factor, and the value of n = 1.8.

[0064] Continuously update the membership degree of the clustering center by calculating the Euclidean distance between each historical data and each clustering center until the maximum number of iterations is reached;

[0065] Compare all the metrics of each cluster of historical data with the rated operating status, divide the historical data of all metrics into a normal set and a fault set, assign a value of 1 to the fault set and a value of 0 to the normal set. Add the previously recorded fault status data to the divided fault set. Perform a Pearson correlation analysis on the historical data of all metrics, and select the top five evaluation metrics with the highest correlation with the charging pile health status as the primary metrics.

[0066] S3. Set two secondary metrics, namely amplitude and volatility, for all primary metrics;

[0067] The definition of amplitude is as follows:

[0068] Define the maximum value of the charging pile operating metric within a unit time as the amplitude A of this metric within this unit time i ;

[0069] The definition of volatility is as follows:

[0070] Set the fluctuation state of a certain metric within a unit time

[0071]

[0072] Use the rated operating status as the judgment benchmark, where is the fluctuation state of metric i at time t, 0 means not crossing the zero point, 1 means crossing the zero point, is the difference between the real-time value and the rated operating value of metric i at time t. Record the number of zero-crossing times c of the operating metric value within a unit time i

[0073]

[0074] Define c i The quotient of c and the unit time is defined as the volatility of this metric within this unit time.

[0075] S4. Construct a judgment matrix for the primary metrics based on the Pearson correlation coefficients of the primary metrics, and determine the weights of each primary metric based on the judgment matrix of the primary metrics; specifically:

[0076] Divide the correlation coefficients of the primary metrics into four levels (1, 3, 5, 7) of importance in ascending order of absolute value. The importance levels include: equally important, slightly important, significantly important, and strongly important, where 1 is equally important, 3 is slightly important, 5 is significantly important, and 7 is strongly important. Construct the judgment matrix of the primary metrics as follows:

[0077]

[0078] In the formula, d iirepresents the diagonal element of the matrix, indicating the importance of index i with respect to index i, equal to 1, indicating equal importance; d ij represents the importance of index j with respect to index i; d ji represents the importance of index i with respect to index j;

[0079] Based on the judgment matrix of the first-level indicators, the weights of each first-level indicator are determined by the analytic hierarchy process, w = [w 1 , w 2 , w 3 , w 4 , w 5 T , w represents the weight vector composed of the weights of the five first-level indicators, w i (i = 1, 2,..., 5) represents the weight of the i-th first-level indicator, reflecting the influence degree of this indicator on the health state of the charging pile.

[0080] Let the weight values of the secondary indicators be the same as those of the first-level indicators to ensure the consistency of the scoring criteria for the health state of the charging pile.

[0081] Analyze and label the fault characteristics of each indicator data in the fault set, compare it with the normal set, find the critical safety state in the normal set with an error within ±5% of the fault characteristic value, include it in the fault set, and re-divide the data set to avoid missing the critical health state and reduce the accident risk.

[0082] S5. Calculate the health values of each fault data in the fault set using the weights of the first-level indicators; specifically:

[0083] Regard the rated operating state of the charging pile as the standard operating state A = [a 1 , a 2 , a 3 , a 4 , a 5 T , a i (i = 1, 2,..., 5) represents the rated value of the i-th first-level indicator when the charging pile is operating normally;

[0084] Calculate the health values of each fault data in the fault set using the weights of the first-level indicators and the rated values of the first-level indicators. The formula is as follows:

[0085]

[0086] In the formula, bw k represents the health value of the k-th group of fault data, w i represents the weight of the i-th first-level indicator, b ki (i = 1, 2,..., 5) represents the fault data of the i-th first-level indicator in the k-th group of fault data, a i ​​Denote the rated value of the \(i\)-th primary indicator when the charging pile is operating normally; the \(k\)-th group of fault data is denoted as \(B\). k = [b k1 , b k2 , b k3 , b k4 , b k5 T .

[0087] Take the minimum value of the health values of each fault data as the critical health value of the charging state of the charging pile; set the critical health value \(\alpha\) of the charging state of the charging pile as \(\min\{BW\}\); \(BW\) represents the set of health values of the data set.

[0088] BW = [bw 1 , bw 2 , …, bw k T

[0089] bw k represents the health value of the \(k\)-th group of fault data.

[0090] S6. Use secondary indicators to construct a fault gene set; as Figure 2 shown, specifically:

[0091] S61. Extract the fault data with the largest difference in volatility of the primary indicators corresponding to the normal set in the fault set.

[0092] S62. Convert the fault data into a waveform for visual analysis, specifically including: splitting the fault data into bands of three different waveforms: linearly increasing, sinusoidal, and linearly decreasing, recording the volatility and amplitude corresponding to different waveform bands, and calculating the time from the starting point of each waveform band to the fault occurrence point as the fault occurrence time.

[0093] The fault gene fragment consists of amplitude, volatility, and fault occurrence time, and the formula is as follows:

[0094] wave u = [A u , fault_t u , f u

[0095] In the formula, wave u represents the fault gene fragment of waveform \(u\), where \(u = 1\) indicates that the waveform is linearly increasing, \(u = 0\) indicates sinusoidal, \(u = -1\) indicates linearly decreasing, fault_t u represents the fault occurrence time of waveform \(u\), and f u represents the volatility of waveform \(u\).

[0096] f u = c u ​​​ / fault_t u

[0097] Wherein, c u represents the number of zero-crossing points of the waveform u within the fault occurrence time;

[0098] S63. Combine the fault gene fragments to form a fault gene set.

[0099] S7. Obtain the real-time data of each first-level index under the charging state of the charging pile, and perform median filtering on the real-time data to reduce data noise; use the real-time data of each first-level index and predict the predicted values P = [p 1 , p 2 , p 3 , p 4 , p 5 of each future first-level index based on the extreme gradient boosting tree algorithm. T , p i represents the predicted value of the i-th future first-level index, and P is regarded as the future operating state of the charging pile. In this embodiment, the future refers to the next 2 minutes.

[0100] S8. Perform Pearson correlation analysis by combining the real-time data of each first-level index and the predicted values of each first-level index, and update the Pearson correlation coefficient of the first-level index; adjust the judgment matrix of the first-level index, update the weights of each first-level index, and calculate the predicted health value based on the updated weights; The specific calculation of the predicted health value based on the updated weights is as follows:

[0101]

[0102] Wherein, β represents the predicted health value, w′ i represents the updated weight of the i-th first-level index, p i represents the predicted value of the i-th future first-level index, and a i represents the rated value of the i-th first-level index when the charging pile is operating normally.

[0103] S9. If the critical health value α is less than the predicted health value β, it is considered that the charging pile is in a normal operating state and no intervention is required; otherwise, it is considered that the charging pile is in a marginal operating state, and the predicted values of each first-level index are matched and verified with the fault gene set. If the verification passes, enter S10; otherwise, return to S8; The specific matching and verification of the predicted values of each first-level index with the fault gene set is as follows:

[0104] Analyze the fluctuation characteristics of the predicted values of each first-level index. The fluctuation characteristics include waveform, amplitude, fault occurrence time, and volatility, and match the fluctuation characteristics with the fault gene fragments in the fault gene set; verify the accuracy and reliability of the prediction result, thereby improving the charging pile management efficiency and reducing the risk of charging accidents of the charging pile;

[0105] The standard for passing the verification is that a matching error within ±5% is regarded as passing the inspection.

[0106] S10. Query the fault data corresponding to the health value with the smallest difference between the health value and the predicted health value among the fault data in the fault set; and calculate the similarity between each first-level index of this fault data and the predicted values of the future first-level indexes. The calculation formula for the similarity is as follows:

[0107]

[0108] In the formula, ρ represents the similarity between each first-level index of the fault data and the predicted values of the future first-level indexes, P = [p 1 , p 2 , p 3 , p 4 , p 5 T , P represents the predicted future operating state of the charging pile, p i (i = 1, 2,..., 5) represents the predicted value of the i-th future first-level index, B represents the fault data, Cov(P, B) represents the covariance between the future operating state of the charging pile and the fault data, DP represents the variance of the future operating state P of the charging pile, and DB represents the variance of the fault data B.

[0109] S11. If the similarity is greater than the similarity threshold, and in this embodiment, the similarity threshold is 0.8, then determine that the charging pile state is a high-risk level and feedback the "power-off alarm" command to the background; otherwise, the charging pile state is a low-risk level and feedback the "manual intervention" command to the background.

[0110] The method of this embodiment further includes: obtaining the current operating environment conditions of the charging pile, such as the temperature inside the pile, the hard instructions from the background, the large-scale parallel situation of the station area equipment, etc.;

[0111] Predict the load fluctuation situation through the prophet algorithm. If there is a large fluctuation or increase in the load, or there are extremely harsh situations such as hard instructions from the substation area, high-temperature alarms, policy instructions, etc. in the charging environment at this moment, then upgrade the risk level.

[0112] ​Based on the practical application of traditional warning methods, considering the particularity of the sensitivity of various indicators to the operating state of charging piles, this invention constructs a charging risk assessment mechanism for electric vehicle charging piles. By considering the subsequent load fluctuations and the variability of actual operating conditions, it determines whether higher-level intervention control is required. At the same time, it verifies the judgment results by analyzing similar segments of the characteristics of historical fault data, better ensuring the reliability of the judgment. This invention can well solve the situation where the charging piles cannot obtain efficient and accurate early warnings during operation due to the insufficient utilization of the operating characteristics of charging piles, resulting in safety accidents, improving the accuracy and reliability of control, and providing a more refined control method for the online intelligent early warning control of charging piles.

[0113] The above are only the preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions within the idea of the present invention belong to the protection scope of the present invention. It should be pointed out that for those of ordinary skill in the art in this technical field, several improvements and refinements made without departing from the principle of the present invention should be regarded as within the protection scope of the present invention.

Claims

1. A vehicle-pile interaction safety online early warning method based on charging risk assessment, characterized in that: The following steps are involved: S1. Obtain historical data of all indicators of charging status of the charging pile; S2. Divide the historical data of all indicators into a normal set and a fault set, perform correlation analysis on the historical data of all indicators, and select the top five evaluation indicators with the highest correlation with the health status of the charging pile as the first-level indicators; S3. Set two secondary indicators, amplitude and volatility, for all primary indicators; S4. Construct a judgment matrix of the first-level indicators based on the correlation coefficients of the first-level indicators, and determine the weights of each first-level indicator based on the judgment matrix of the first-level indicators; S5. Calculate the health value of each fault data in the fault set using the weight of the first-level indicator; The minimum value of the health value of each fault data is used as the critical health value of the charging state of the charging pile; S6. Use the secondary indicators to construct the fault gene set; S7, obtaining real-time data of various first-level indicators under the charging state of the charging pile, and using the real-time data of various first-level indicators to predict the predicted values ​​of various first-level indicators in the future; S8. Perform correlation analysis based on the real-time data of each first-level indicator and the predicted value of each first-level indicator, and update the correlation coefficient of the first-level indicator; adjust the judgment matrix of the first-level indicator, update the weight of each first-level indicator, and calculate the predicted health value based on the updated weight; S9. If the critical health value is less than the predicted health value, the charging pile is considered to be in normal operation and no intervention is required; otherwise, the charging pile is considered to be in marginal operation, and the predicted values ​​of each primary indicator are matched with the fault gene set for verification. If the verification passes, enter S10, otherwise return to S8; S10, querying the fault data corresponding to the health value with the smallest difference between the health value of each fault data in the fault set and the predicted health value; And calculate the similarity between each first-level index of the fault data and the predicted value of each first-level index in the future; S11. If the similarity is greater than the similarity threshold, the charging pile status is judged to be at a high risk level, and a "power failure alarm" command is fed back to the background; otherwise, the charging pile status is at a low risk level, and a "manual intervention" command is fed back to the background.

2. The vehicle-pile interactive safety online early warning method based on charging risk assessment as claimed in claim 1 is characterized in that: In S2, the specific process of dividing the historical data of all indicators into a normal set and a fault set is: Perform fuzzy clustering on the historical data of all indicators, calculate and record the cluster centers, and continuously update the membership of the cluster centers by calculating the Euclidean distance between each historical data and each cluster center until the maximum number of iterations is reached; All indicators of each cluster of historical data are compared with the rated operating state, and the historical data of all indicators are divided into a normal set and a fault set.

3. The vehicle-pile interactive safety online early warning method based on charging risk assessment as claimed in claim 1 is characterized in that: In S2, the correlation analysis of the historical data of all indicators is specifically a Pearson correlation analysis, and in S4, the correlation coefficient is a Pearson correlation coefficient.

4. The vehicle-pile interactive safety online early warning method based on charging risk assessment as claimed in claim 1 is characterized in that: S4 is specifically: The correlation coefficients of the first-level indicators are divided into four levels of importance in ascending order of absolute value. The importance levels include: equally important, slightly important, obviously important, and strongly important. The judgment matrix of the first-level indicators is constructed as follows: Where, d ii Indicates the importance of indicator i with respect to indicator i, and 1 indicates that it is equally important; d ij Indicates the importance of indicator j to indicator i; d ji Indicates the importance of indicator i with respect to indicator j; Based on the judgment matrix of the first-level indicators, the weights of each first-level indicator w=[w1,w2,w3,w4,w5] are determined by hierarchical analysis. T , w represents the weight vector composed of the weights of the five first-level indicators, w i (i=1, 2, ..., 5) represents the weight of the i-th primary indicator.

5. The vehicle-pile interactive safety online early warning method based on charging risk assessment as claimed in claim 1 is characterized in that: In S5, the health value of each fault data in the fault set is calculated by using the weight of the first-level index as follows: The rated operating state of the charging pile is regarded as the standard operating state A = [a1, a2, a3, a4, a5] T , a i (i=1, 2, ..., 5) represents the rated value of the i-th primary indicator when the charging pile is operating normally; The health value of each fault data in the fault set is calculated using the weight of the first-level indicator and the rated value of the first-level indicator. The formula is as follows: Where bw k represents the health value of the kth group of fault data, w i represents the weight of the i-th primary indicator, b ki (i=1,2,……,5) represents the fault data of the i-th primary indicator in the k-th group of fault data, a i It represents the rated value of the first-level indicator of the i-th item when the charging pile is operating normally; the k-th group of fault data is represented by B k =[b k1 , b k2 , b k3 , b k4 , b k5 ] T .

6. The vehicle-pile interactive safety online early warning method based on charging risk assessment as claimed in claim 1 is characterized in that: S6 is specifically: S61, extracting the fault data with the largest volatility difference between the primary index corresponding to the fault set and the normal set; S62, converting the fault data into a waveform for visual analysis, specifically comprising: dividing the fault data into three different waveform bands: linear increasing, sinusoidal, and linear decreasing; recording the fluctuation rate and amplitude corresponding to different waveform bands; and calculating the time from the starting point of each waveform band to the fault occurrence point as the fault occurrence time; The fault gene fragment consists of amplitude, volatility and fault occurrence time, and the formula is as follows: wave u =[A u ,fault_t u ,f u ] In the formula, wave u The fault gene fragment represents the waveform u, where u=1 means the waveform is linearly increasing, u=0 means sine, and u=-1 means linearly decreasing. u represents the fault occurrence time of waveform u, f u represents the volatility of waveform u; f u =c u / fault_t u In the formula, c u Indicates the number of zero crossings of waveform u during the fault occurrence time; S63. The faulty gene fragments are combined into a faulty gene set.

7. The vehicle-pile interactive safety online early warning method based on charging risk assessment as claimed in claim 1 is characterized in that: In S7, the method of using the real-time data of each primary indicator to predict the predicted value of each primary indicator in the future is an extreme gradient boosting tree algorithm.

8. The vehicle-pile interactive safety online early warning method based on charging risk assessment as claimed in claim 1 is characterized in that: In S8, the predicted health value is calculated based on the updated weights as follows: In the formula, β represents the predicted health value, w′ i represents the updated weight of the i-th primary indicator, p i represents the predicted value of the first-level indicator of the future, a i It indicates the rated value of the i-th primary indicator when the charging pile operates normally.

9. The vehicle-pile interactive safety online early warning method based on charging risk assessment as claimed in claim 1, characterized in that: In S9, the matching and verification of the predicted values ​​of each primary index with the fault gene set is specifically as follows: Analyze the fluctuation characteristics of the predicted values ​​of each primary indicator, wherein the fluctuation characteristics include waveform, amplitude, fault occurrence time and fluctuation rate, and match the fluctuation characteristics with the fault gene fragments in the fault gene set; The standard for passing the verification is that the matching error is within ±5%, which is considered to be passed.

10. The vehicle-pile interactive safety online early warning method based on charging risk assessment as claimed in claim 1, characterized in that: In S10, the similarity calculation formula is as follows: Where ρ represents the similarity between the first-level indicators of the fault data and the predicted values ​​of the first-level indicators in the future, P = [p1, p2, p3, p4, p5] R , P represents the predicted future charging pile operation status, p i (i=1, 2, ..., 5) represents the predicted value of the i-th primary indicator in the future, B represents the fault data, Cov(P, B) represents the covariance of the future charging pile operating state and the fault data, DP represents the variance of the future charging pile operating state P, and DB represents the variance of the fault data B.

Images