Charging pile fault diagnosis method and diagnosis interval regulation and control method
By combining a charging pile fault type database and a predictive model with temperature, current, and voltage data for fault diagnosis, the problem of decreased accuracy under low and high temperature environments has been solved, achieving automated and precise charging pile fault diagnosis and improving efficiency and accuracy.
Patent Information
- Application Number
- CN202511082840.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-04
- Publication Date
- 2025-10-28
AI Technical Summary
Existing charging pile fault diagnosis methods have reduced accuracy in low or high temperature environments, and manual testing is inefficient, making it difficult to meet the fault diagnosis needs of large-scale charging piles.
By acquiring a database of charging pile fault types and a prediction model, and combining ambient temperature, three-phase current, and switching voltage, fault types are predicted. The diagnostic interval is adjusted by dividing temperature ranges based on historical data, and the weight factors are updated using an optimized width learning model and elastic network regression to achieve automated and accurate fault diagnosis.
The accuracy and efficiency of charging pile fault diagnosis have been improved under different temperature conditions, unnecessary fault diagnosis costs have been reduced, faults have been detected and dealt with in a timely manner, and economic losses have been reduced.
Smart Images
Figure CN120847518A_ABST
Abstract
Description
Technical Field
[0001] This invention generally relates to the field of charging pile fault diagnosis technology, and specifically to a charging pile fault diagnosis method and a method for adjusting the diagnosis interval. Background Technology
[0002] Existing charging stations require regular fault diagnosis. Traditionally, this involves manual on-site inspections using testing equipment to determine if a fault has occurred. However, with the increasing number of electric vehicles, the number of charging stations is also increasing. This traditional method of manual fault detection is inefficient and extremely labor-intensive, making it difficult to meet the current needs for charging station fault diagnosis.
[0003] For the reasons mentioned above, existing technologies have introduced more intelligent fault diagnosis methods for charging piles during the fault diagnosis process. These include multiple sensors, artificial intelligence algorithms (machine learning / deep learning), reliable communication, and other key technologies. The aim is to achieve remote, automated, intelligent, and precise fault diagnosis, providing core technical support for intelligent operation and maintenance.
[0004] However, existing intelligent diagnostic methods only consider the current and voltage parameters of the charging pile circuit, without taking into account the actual temperature of the charging pile and its impact on the fault. As a result, the accuracy of existing diagnostic methods is relatively accurate in normal temperature environments, but the accuracy decreases significantly in low or high temperature environments. Summary of the Invention
[0005] In view of the above-mentioned defects or deficiencies in the prior art, it is desirable to provide a method for diagnosing charging pile faults and a method for adjusting the diagnostic interval.
[0006] On one hand, the present invention provides a method for diagnosing charging pile faults, including: Obtain a fault type database for charging piles; the fault type database contains fault type datasets for various models of charging piles; the fault type dataset contains all fault types for the same model of charging pile, as well as the faulty components corresponding to each fault type; Obtain a prediction model; the prediction model is used to predict the fault type of the charging pile by taking into account the ambient temperature, three-phase current and switching voltage of the charging pile. The actual three-phase current at the input terminal of the charging pile, the actual switching voltage at the output terminal, and the actual ambient temperature of the charging pile when a fault occurs are detected. Obtain the model number of the charging station to be diagnosed; Based on the model of the charging pile to be diagnosed, match the corresponding fault type dataset in the fault type database; If the match is successful, the actual ambient temperature, actual three-phase current and actual switching voltage are input into the prediction model to obtain the first actual fault type of the charging pile. The first actual fault type is matched with multiple fault types in the fault type dataset to obtain the faulty component corresponding to the first actual fault type.
[0007] According to the technical solution provided by the present invention, the prediction model includes weighting factors; The method for diagnosing charging pile faults also includes: If the matching fails, a fault type dataset corresponding to the model of the charging pile to be diagnosed will be added to the fault type database. Update the weight factors of the prediction model to obtain the updated weight factors; The updated weighting factor is used to replace the weighting factor in the prediction model to obtain the updated prediction model. The actual ambient temperature, actual three-phase current and actual switching voltage are input into the updated prediction model to obtain the second actual fault type of the charging pile. The second actual fault type is matched with multiple fault types in the fault type dataset corresponding to the model of the charging pile to be diagnosed to obtain the faulty component corresponding to the second actual fault type.
[0008] According to the technical solution provided by the present invention, the prediction model is an optimized width learning model, which includes a weight factor between feature nodes and enhancement nodes; The updated weight factors of the prediction model are obtained by updating the weight factors, including: Obtain the weighting factors between the feature nodes and enhancement nodes of the prediction model; The weight factors are updated using an elastic network regression method to obtain updated weight factors.
[0009] On the other hand, the present invention also provides a method for adjusting and diagnosing the interval of a charging pile, characterized in that it includes: S1: Obtain historical data on multiple charging pile malfunctions; the historical data includes: the time of the malfunction and the ambient temperature at the time of the malfunction. S2: Based on the time of the fault and the ambient temperature at the time of the fault, the temperature is divided into low temperature range, normal temperature range and high temperature range; S3: Real-time detection of actual ambient temperature; S4: If the actual ambient temperature is in the low temperature range, the charging pile is subjected to the above-described charging pile fault diagnosis method at a first interval; the first interval is the shortest interval between the times when the charging pile malfunctions when the ambient temperature is in the low temperature range. If the actual ambient temperature is within the normal temperature range, the charging pile is subjected to the above-described charging pile fault diagnosis method at a second interval; the second interval is the shortest interval between the times when the charging pile malfunctions when the ambient temperature is within the normal temperature range. If the actual ambient temperature is within the high temperature range, the charging pile is subjected to the above-described charging pile fault diagnosis method at a third interval; the third interval is the shortest interval between the times when the charging pile malfunctions when the ambient temperature is within the high temperature range.
[0010] According to the technical solution provided by the present invention, the historical data further includes: the temperature of the charging pile when the fault occurred; The method for diagnosing charging pile faults also includes: The maximum temperature of the charging pile in the low-temperature range from the historical data is obtained to obtain the first reference temperature; The minimum temperature of the charging pile under high-temperature conditions is obtained from the historical data to obtain the second reference temperature; Real-time monitoring of the actual temperature of the charging pile; If the actual temperature of the charging pile is equal to the first reference temperature, an additional fault diagnosis is immediately performed, and the charging pile is subjected to the above-described charging pile fault diagnosis method according to the first interval. If the actual temperature of the charging pile is equal to the second reference temperature, an additional fault diagnosis is immediately performed, and the charging pile is subjected to the aforementioned charging pile fault diagnosis method according to the third interval.
[0011] According to the technical solution provided by the present invention, the above-described method for diagnosing charging pile faults based on the first interval includes: After the first interval, the actual temperature of the charging pile is detected to obtain the first charging pile temperature fluctuation. Perform a fault diagnosis on the charging station; If the fluctuating temperature of the first charging pile is greater than or equal to the actual temperature of the charging pile detected during the last fault diagnosis, then the above-described charging pile fault diagnosis method is performed at a second interval. If the fluctuating temperature of the first charging pile is less than the actual temperature of the charging pile detected during the previous fault diagnosis, then the above-described charging pile fault diagnosis method is performed at the first interval.
[0012] According to the technical solution provided by the present invention, S2: based on the time of the fault and the ambient temperature at the time of the fault, the temperature is divided into a low temperature range, a normal temperature range, and a high temperature range, including: Based on the time of the failure, the time interval between each failure and the previous failure is calculated to obtain multiple failure intervals; Obtain the median of the multiple fault intervals to get the median fault interval; Calculate the low-temperature critical temperature and high-temperature critical temperature corresponding to the median fault interval; The temperature range below the critical low-temperature temperature is defined as the low-temperature range; The temperature range that is greater than or equal to the low-temperature critical temperature and less than or equal to the high-temperature reference range is set as the normal temperature range; The temperature range above the high-temperature critical temperature is defined as the high-temperature range.
[0013] According to the technical solution provided by the present invention, calculating the low-temperature critical temperature and high-temperature critical temperature corresponding to the median fault interval includes: The maximum value among the multiple fault intervals is obtained to obtain the maximum fault interval; Obtain the ambient temperature of the maximum fault interval to obtain the reference ambient temperature; The ambient temperature corresponding to the largest fault interval among the multiple fault intervals that meets the first set condition is taken as the low temperature critical temperature; the first set condition is: less than or equal to the median fault interval, and the corresponding ambient temperature is less than the reference ambient temperature. The ambient temperature corresponding to the largest fault interval among the multiple fault intervals that meets the second set condition is taken as the high-temperature critical temperature; the second set condition is: less than or equal to the median fault interval, and the corresponding ambient temperature is greater than the reference ambient temperature.
[0014] The beneficial effects of this invention are as follows: The charging pile fault diagnosis method includes: matching the model of the charging pile to be diagnosed with the corresponding models in multiple fault type databases; then using a prediction model to predict the first actual fault type based on ambient temperature, three-phase current, and switching voltage; matching the first actual fault type with multiple fault types in the matched fault type dataset to obtain the faulty component corresponding to the first actual fault type, thus completing the diagnosis. Based on this method, the influence of the actual temperature of the charging pile on the fault is considered, and the prediction of charging pile faults has high accuracy regardless of low, normal, or high temperature conditions. Furthermore, it can perform fault type matching separately for different charging pile models, avoiding comparisons with fault types of different models during the matching process, thus improving the efficiency of fault diagnosis. Attached Figure Description
[0015] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a flowchart illustrating a method for adjusting and diagnosing the interval of a charging pile. Figure 2 This is a graph showing the change in fault intervals as a function of ambient temperature. Detailed Implementation
[0016] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.
[0017] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.
[0018] In actual working environments, charging piles may operate in high or low temperature environments. The various components in the charging pile have different tolerances to high and low temperature environments, which leads to a higher failure rate of charging piles in high and low temperature environments compared to normal temperature environments.
[0019] Since charging pile failures are not frequent occurrences, current technologies, in order to reduce the cost of fault diagnosis, still rely on manually set methods to perform fault diagnosis on charging piles at a low and fixed frequency. This results in the frequency of charging pile failures being significantly higher than the actual fault diagnosis frequency in some high or low temperature environments. Consequently, fault diagnosis becomes lagging, and a small fault may trigger a series of component failures, ultimately making it impossible to compensate for the economic losses caused by the failure through the cost savings of low-frequency fault diagnosis. Simply increasing the fault diagnosis frequency would add unnecessary costs even when no fault occurs.
[0020] refer to Figure 1 To address the above problems, this invention provides a method for adjusting the diagnostic interval of a charging pile, comprising: S1: Obtain historical data of multiple charging pile failures; the historical data includes: the time when multiple failures of the same model of the charging pile to be diagnosed occurred, the ambient temperature when the failure occurred, and the temperature of the charging pile when the failure occurred. It should be noted that if the charging station to be diagnosed is a new model and is being used for the first time, historical data obtained from the manufacturer's pre-shipment testing should be selected.
[0021] S2: Based on the time of the fault and the ambient temperature at the time of the fault, the temperature is divided into low temperature range, normal temperature range and high temperature range; Specifically, in this embodiment, the temperature range is divided according to historical data of the same model of charging pile failures. This makes the division of the temperature range closer to the failure situation of this model, and thus the design of subsequent fault diagnosis intervals can be more accurate.
[0022] Meanwhile, the solution in this embodiment can provide different temperature range division schemes when facing different models of charging piles, and has good versatility.
[0023] In this embodiment, it is assumed that repairs will be carried out immediately upon the discovery of a fault.
[0024] Specifically, S2 includes: S2-1: Calculate the time interval between each fault occurrence and the previous fault based on the time of the fault occurrence, and obtain multiple fault intervals; Specifically, the interval between the current fault and the previous fault is calculated by successively subtracting the timestamps of multiple faults. Then, the fault interval is plotted as a function of ambient temperature in a two-dimensional Cartesian coordinate system. (See reference for details.) Figure 2 .
[0025] Figure 2 This is a graph showing the fault interval as a function of ambient temperature. Here, T represents the ambient temperature, t represents the fault interval, T1 represents the first reference temperature, and T2 represents the second reference temperature. The first and second reference temperatures are the condition temperatures that trigger immediate fault diagnosis.
[0026] The first reference temperature is the maximum temperature of the charging pile in the historical data when it is in the low-temperature range; The second reference temperature is the minimum temperature of the charging pile when it is in the high-temperature range, based on the historical data.
[0027] Figure 2 The dots in the data represent multiple charging pile failures recorded in historical data, including the ambient temperature and the interval between each failure and the previous failure.
[0028] It is evident that in the normal temperature range, the fault points are sparse, the fault intervals are relatively large, and the difference in fault intervals between adjacent points is small; in the low temperature and high temperature ranges, the fault points are dense, the fault intervals are small, and the difference in fault intervals between adjacent points is large. This indicates that when charging piles operate in the normal temperature range, faults occur less frequently, and the intervals between faults are also longer; when charging piles operate in the low temperature and high temperature ranges, faults occur more frequently, and the intervals between faults are shorter.
[0029] Figure 2All data are based on real historical data calculated using a specific method, without any manually input parameters. Therefore, they can more realistically reflect the fault conditions of charging piles under different temperature environments. Furthermore, it is convenient to set the frequency of fault diagnosis of charging piles (hereinafter, fault diagnosis refers to a charging pile fault diagnosis method provided by this invention) based on the actual fault conditions.
[0030] S2-2: Obtain the median of the multiple fault intervals to get the median fault interval; It should be noted that, in actual circumstances, Figure 2 There may be an odd number or an even number of points.
[0031] When there is an even number of points, the median fault interval is not equal to the fault interval calculated from any historical data, but is calculated as the average of two fault intervals that are close to the median fault interval.
[0032] When there is an odd number of points, the median fault interval is equal to the fault interval corresponding to one of the points.
[0033] S2-3: Calculate the low-temperature critical temperature and high-temperature critical temperature corresponding to the median fault interval (the low-temperature critical temperature and high-temperature critical temperature are the critical temperatures for dividing the low-temperature interval, normal temperature interval, and high-temperature interval), including: S2-3-1: Obtain the maximum value among the multiple fault intervals to obtain the maximum fault interval; S2-3-2: Obtain the ambient temperature of the maximum fault interval to obtain the reference ambient temperature; according to Figure 2 The maximum fault interval must be within the normal temperature range, therefore the reference ambient temperature must be between the low temperature critical temperature and the high temperature critical temperature.
[0034] S2-3-3: Among the multiple fault intervals, the ambient temperature corresponding to the largest fault interval that meets the first set condition is taken as the low temperature critical temperature; the first set condition is: less than or equal to the median fault interval, and the corresponding ambient temperature is less than the reference ambient temperature. by Figure 2 For example, the points that satisfy the first set condition are the point corresponding to T1, the first point to its right, and the two points to its left. Among these four points, the first point to the right of the point corresponding to T1 has the largest fault interval (equal to the median fault interval); therefore, the ambient temperature corresponding to this point is taken as the low-temperature critical temperature.
[0035] The ambient temperature corresponding to the largest fault interval among the multiple fault intervals that meets the second set condition is taken as the high-temperature critical temperature; the second set condition is: less than or equal to the median fault interval, and the corresponding ambient temperature is greater than the reference ambient temperature.
[0036] by Figure 2 For example, the point that satisfies the second set condition is the point corresponding to T2, and the three points to its right. The point corresponding to T2 has the largest fault interval; therefore, T2 is taken as the high-temperature critical temperature.
[0037] S2-4: The temperature range below the critical low-temperature temperature is set as the low-temperature range; The temperature range that is greater than or equal to the low-temperature critical temperature and less than or equal to the high-temperature reference range is set as the normal temperature range; The temperature range above the high-temperature critical temperature is defined as the high-temperature range.
[0038] In this embodiment, considering the time required for temperature to cause component failure, the median was used to sample the low-temperature and high-temperature critical temperatures. Because the data points in the low-temperature and high-temperature ranges are more densely packed, the median can be lowered.
[0039] It should be noted that choosing the average value may result in the following: For example, the fault interval of a certain model of charging pile is much larger than that of the fault interval under normal temperature conditions (e.g., more than 10 times) than that under low temperature and high temperature conditions; therefore, when averaged, the average fault interval is higher, which in turn makes the normal temperature range narrower and the low temperature and high temperature range wider.
[0040] For example, if the median fault interval is taken, a temperature of -5 degrees Celsius is required to enter the low-temperature range; if the average value is taken, it enters the low-temperature range at 15 degrees Celsius. It is clear that the two methods have significantly different criteria for determining whether a temperature range has been entered. During the gradual cooling process, fault diagnosis may be performed prematurely at a higher frequency, while the actual fault interval of the charging pile will not be significantly shortened, thus increasing the cost of fault diagnosis.
[0041] Therefore, this embodiment does not select the average value, which can avoid the impact of individual interfering data or extreme data on the overall calculation.
[0042] Because the median value is relatively low in this embodiment, the triggering time of the additional fault diagnosis during the transition from a normal temperature environment to a low or high temperature environment will be slightly delayed. This allows sufficient time for the temperature effect to cause component failure and enables timely detection of whether temperature changes will cause component failure. This reduces the probability of a failure occurring shortly after fault diagnosis has been completed and no fault has been confirmed.
[0043] At the same time, a lower median value allows for a wider normal temperature range. Because the second interval within the normal temperature range is relatively long, fault diagnosis can be performed at a lower frequency, saving some of the cost of fault diagnosis.
[0044] In this embodiment, no manual input of any parameters is required. Only one data filtering method is set, and the calculation is based entirely on real historical data. The temperature range is divided in a more objective way, thereby completing the fully automatic fault diagnosis of the charging pile.
[0045] Because it is not affected by human factors, the timing of fault diagnosis can be closer to the actual time of fault occurrence, thus having a shorter lag. At the same time, under different ambient temperatures, the charging pile is diagnosed according to the shortest historical interval of fault occurrence under the current ambient temperature, ensuring timely detection of faults and avoiding unnecessary costs when the temperature is in other ranges.
[0046] S3: Real-time detection of actual ambient temperature; specifically, this is achieved through a temperature sensor installed on the charging station.
[0047] S4: If the actual ambient temperature is within the low-temperature range, then fault diagnosis of the charging pile is performed at a first interval; the first interval is the shortest interval between the times when the charging pile malfunctions when the ambient temperature is within the low-temperature range; that is... Figure 2 The fault interval at the leftmost point.
[0048] If the actual ambient temperature is within the normal temperature range, then fault diagnosis of the charging pile is performed at a second interval; the second interval is the shortest interval between the times when the charging pile malfunctions when the ambient temperature is within the normal temperature range; that is... Figure 2 Among them, the fault interval is greater than or equal to the median fault interval and the smallest fault interval.
[0049] If the actual ambient temperature is within the high-temperature range, then fault diagnosis of the charging pile is performed at a third interval; the third interval is the shortest interval between the times when the charging pile malfunctions when the ambient temperature is within the high-temperature range. Figure 2 The fault interval at the rightmost point.
[0050] Specifically, by using the shortest fault interval for each temperature range to diagnose charging piles, the lag time between the occurrence of a fault and the discovery of the fault can be reduced; this allows for timely detection and resolution of faults, reducing property damage caused by faults.
[0051] Furthermore, a method for adjusting the diagnostic interval of a charging pile also includes: The maximum temperature of the charging pile under low-temperature conditions is obtained from the historical data to obtain the first reference temperature T1; The minimum temperature of the charging pile under high-temperature conditions is obtained from the historical data to obtain the second reference temperature T2; Real-time monitoring of the actual temperature of the charging pile; If the actual temperature of the charging pile is equal to the first reference temperature, an additional fault diagnosis is immediately performed, and the charging pile is diagnosed according to the first interval. If the actual temperature of the charging pile is equal to the second reference temperature, an additional fault diagnosis is immediately performed, and the charging pile is also diagnosed according to the third interval.
[0052] Specifically, charging stations require energy conversion, and according to the laws of thermodynamics, energy conversion cannot reach 100%; therefore, charging stations generate heat during charging. The temperature of a charging station during charging is significantly higher than the ambient temperature; especially during daytime charging in summer, the temperature inside the charging station may exceed 70 degrees Celsius.
[0053] If charging continues when the internal temperature of the charging station reaches 70 degrees Celsius, the internal temperature may continue to rise, significantly increasing the likelihood of damage to internal components. Assuming the second reference temperature is set to 78 degrees Celsius, if the charging station's temperature continues to rise to 78 degrees Celsius, it is necessary to immediately perform a fault diagnosis.
[0054] It is important to note that the first and second reference temperatures are the conditions for triggering immediate fault diagnosis, and are not equal to the temperature of the charging pile corresponding to the low-temperature critical temperature or high-temperature critical temperature at the edge of the normal temperature range. The reason is as follows: If the temperature of the charging pile corresponding to the low-temperature critical temperature or high-temperature critical temperature in the normal temperature range is used as the first or second reference temperature for additional fault diagnosis, it may lead to fault diagnosis before the temperature causes the component to fail, thus increasing the probability that a fault will occur shortly after a fault diagnosis is completed.
[0055] This method allows the first reference temperature to be lower than the minimum value of the normal temperature range and the second reference temperature to be higher than the maximum value of the normal temperature range. This allows the temperature to have a certain impact on the component before fault diagnosis is performed. As a result, faulty components in the low temperature range or high temperature range can be detected in a timely manner.
[0056] Specifically, when the temperature fluctuates between normal, low, and high temperature ranges, the time interval for fault diagnosis needs to be switched. Specific methods include: Situation near the low temperature range: Fault diagnosis of the charging pile based on the first interval includes: After the first interval, the actual temperature of the charging pile is detected to obtain the first charging pile temperature fluctuation. Perform a fault diagnosis on the charging station; If the fluctuating temperature of the first charging pile is greater than or equal to the actual temperature of the charging pile detected during the last fault diagnosis, then fault diagnosis will be performed at the second interval. If the fluctuating temperature of the first charging pile is less than the actual temperature of the charging pile detected during the last fault diagnosis, then the fault diagnosis is performed at the first interval.
[0057] Situation near the high temperature range: Fault diagnosis of the charging pile based on the third interval includes: After the third interval, the actual temperature of the charging pile is detected to obtain the second charging pile fluctuation temperature. Perform a fault diagnosis on the charging station; If the fluctuating temperature of the second charging pile is greater than the actual temperature of the charging pile detected during the previous fault diagnosis, then fault diagnosis will be performed at the third interval. If the fluctuating temperature of the second charging pile is less than or equal to the actual temperature of the charging pile detected during the last fault diagnosis, then the fault diagnosis is performed at the second interval.
[0058] Based on the above, the time interval for fault diagnosis can be freely switched as the temperature fluctuates between the three ranges.
[0059] In reality, the time interval between charging pile malfunctions is quite long, generally at least several tens of days. For example, in winter, the ambient temperature remains at a low level of several degrees below zero to over ten degrees Celsius, both day and night. The ambient temperature may remain within the low-temperature range of the charging pile calculated according to the scheme of this invention for several consecutive days.
[0060] Therefore, fault diagnosis of the charging pile is continuously performed at the first interval to detect faults in the charging pile in a timely manner.
[0061] For example, in summer, the temperature difference between day and night is quite significant. Since charging stations are generally installed in the open air, the ambient temperature at the charging station during the day in summer may exceed 38 degrees Celsius, while the ambient temperature at night in summer will drop below 30 degrees Celsius, to around 27-29 degrees Celsius.
[0062] Since the diurnal temperature range is approximately 10 degrees Celsius, it typically spans the critical high-temperature zone between the high-temperature and normal-temperature zones. Therefore, during the warmer daytime hours, a third interval is needed for fault diagnosis to promptly detect component burnout due to overheating; while during summer nights, a second interval within the normal-temperature zone is used for fault diagnosis.
[0063] In summer, rainy days can cause a drop in temperature, and the fault diagnosis interval will switch repeatedly between the second and third intervals.
[0064] Specifically, when the ambient temperature first decreases and then increases, or first increases and then decreases, fault diagnosis can be avoided in the low temperature range and high temperature range with a longer second interval; or fault diagnosis can be performed in the normal temperature range with a first interval or a third interval.
[0065] This invention also provides a method for diagnosing charging pile faults (i.e., the fault diagnosis mentioned above), comprising: Q1: Obtain the fault type database for charging piles; if the fault type database has been updated multiple times, the most recently updated fault type database shall prevail. The fault type database contains fault type datasets for various models of charging piles; the fault type dataset contains all fault types (actually fault type numbers) for the same model of charging pile, and the fault components (actually fault component numbers) corresponding to each fault type.
[0066] The fault type dataset contains the fault types corresponding to the permutations and combinations of all component failures for a certain model of charging pile. The number of fault types is equal to the total number of permutations and combinations of one, multiple, or even all components of that model of charging pile failing, plus 1; the additional fault type is the case where the charging pile does not fail, in which case the faulty component is numbered "None" or assigned a fixed number when there is no fault.
[0067] Because there are many different models of charging piles, it is generally difficult to include fault type datasets for all models of charging piles in the fault type database at once. Therefore, initially, the fault type database only contains fault type datasets for a few commonly used models of charging piles.
[0068] Before troubleshooting a new model of charging station, staff can manually update the fault type database in advance by adding the fault type database for the new model of charging station.
[0069] Therefore, as the number of uses increases, the fault type database will gradually include fault type datasets for more charging pile models; when the manufacturer produces the latest model of charging pile, the solution in this embodiment can also be adapted to the new model of charging pile, greatly improving the applicability of the solution.
[0070] The model of the charging piles put into use is known, and all the components of a certain model of charging pile, as well as the arrangement of all component failures, are also known (obtained by the manufacturer through actual testing during the production of this model of charging pile).
[0071] Q2: Obtain the prediction model; the prediction model contains weighting factors, which are used to predict the fault type of the charging pile based on the ambient temperature, three-phase current and switching voltage of the charging pile. Specifically, the prediction model is trained by taking the three-phase current, switching voltage and ambient temperature from the known data of the charging pile as input, and the data in the fault type database as label data, and the fault type of the charging pile as output.
[0072] The prediction model is an Optimized Wide Learning Model (IBLS), which includes weight factors between feature nodes and reinforcement nodes. It is an efficient planar neural network improved on the basis of a stochastic vector function chain neural network, and it has reinforcement learning capabilities. That is, after training is completed, in the actual application process, the weights of the neural network are changed according to whether the prediction results match the fault type, so that it can more effectively adapt to the problem of fault categories accumulating gradually over time in actual working conditions.
[0073] The optimized width learning model is a special type of one-dimensional convolutional neural network model (1DCNN-Transformer), and its specific structure includes: The layers are connected in sequence: input layer, convolutional layer, pooling layer, flattening layer, Transformer layer, fully connected layer, and output layer.
[0074] Q3: Detect the actual three-phase current at the input end of the charging pile (including the three current values of the three-phase circuit), the actual switching voltage at the output end (including the two voltage values at the output end), and the actual ambient temperature of the charging pile when the fault occurs; Q4: Obtain the model number of the charging pile to be diagnosed; specifically, the model number of the charging pile. Specifically, the fault type database contains multiple fault type datasets, each containing the corresponding model number of the charging pile. By querying, the fault type dataset corresponding to the model of the charging pile to be diagnosed can be obtained.
[0075] Q5: Match the corresponding fault type dataset in the fault type database according to the model of the charging pile to be diagnosed; Specifically, before fault diagnosis, the corresponding fault type database is first selected from the fault type database according to the model of the charging pile to be diagnosed, and then the fault type is matched in the corresponding fault type database.
[0076] The matching process avoids comparing fault types with different models of charging piles, greatly reducing the total amount of data to be matched and the time required to match fault types, thus enabling faster fault diagnosis and timely fault detection.
[0077] Step Q5 specifically includes the following situations: Case 1: If the match is successful, it includes: Q5-1-1: Input the actual ambient temperature, actual three-phase current and actual switching voltage into the prediction model to obtain the first actual fault type of the charging pile; Based on the structure of the prediction model mentioned above, step Q5-1-1 includes the following: First, the actual three-phase current and the actual switching voltage are mapped to different feature node groups according to weights. Then, these feature nodes are used to build enhancement nodes to achieve the purpose of generalizing and expanding the network structure. Finally, all the above nodes are sent to the input layer; then the input layer is fed into the convolutional layer.
[0078] The main function of convolutional layers is to efficiently extract representative feature information from the original input signal. The ReLU function is used as the non-linear activation function for convolutional layers. The processed data is then passed to pooling layers.
[0079] To account for the impact of noise, the pooling layer uses the Mish activation function, which is continuously differentiable and can effectively mitigate gradient abrupt changes caused by impulse noise. The pooling layer reduces the spatial dimensionality of the feature map through downsampling operations, thereby reducing the number of network parameters and computational complexity.
[0080] In this invention, the three-phase current and switching voltage are detected by multiple different sensors. The number of terms in the input parameters of the neural network is called the channel. Therefore, different sensors can be regarded as different channels, resulting in the input data in this embodiment being multiple one-dimensional multi-channel wave signals, with both ends filled with zeros.
[0081] The sliding window of the convolution kernel is also one-dimensional and multi-channel. Considering the different sampling frequencies of different sensors, different convolution kernels are used for each channel, and convolution operations are performed on the data of each channel separately. Subsequently, the obtained feature maps are processed by max pooling or average pooling, then flattened by a flattening layer and stitched together, and finally projected into the Transformer layer.
[0082] The Transformer layer incorporates a multi-head attention mechanism. The Transformer layer performs computation on each head and utilizes this mechanism to analyze the correlation patterns of signals from multiple sensors. Compared to other models, the multi-head attention mechanism can process multiple sets of data in parallel, thereby comprehensively extracting the overall features of the data, making feature extraction more refined and efficient, and significantly improving computational efficiency.
[0083] This invention uses a 1DCNN-Transformer feature extraction model, placing the Transformer layer after the convolutional layer to combine the advantages of both, achieving feature extraction from local to global, while reducing computational cost and improving computational efficiency.
[0084] The fully connected layer receives the output of the pooling layer, flattens it into a one-dimensional vector, and integrates the local feature information extracted by the convolutional or pooling layers to achieve global information integration. The output layer is usually equipped with a Softmax classifier to calculate the probability value of each category, and the final classification result is determined based on the maximum probability value.
[0085] Finally, the output layer completes the output, and the final prediction result, namely the first actual fault type of the charging pile, can be obtained.
[0086] Q5-1-2: Match the first actual fault type with multiple fault types in the fault type dataset to obtain the faulty component corresponding to the first actual fault type; Since the predicted fault types are also in the form of numbers, the corresponding fault component number can be obtained by searching for the number corresponding to the first actual fault type in the fault type database.
[0087] The system then displays the faulty component's number and the charging station that experienced the malfunction on the charging station's monitoring screen. Based on this data, staff can then proceed to the corresponding charging station for repairs.
[0088] Because the fault type database includes fault types for cases without faults, a successful match also includes cases where no component is faulty. If there is no fault, the faulty component is displayed as "None," or a fixed number for cases without faults.
[0089] Case 2: If the match fails, it includes: Q5-2-1: Add a fault type dataset corresponding to the model of the charging pile to be diagnosed to the fault type database; Specifically, if a match is not found, it means that the models of multiple charging piles corresponding to multiple fault type databases currently stored in the fault type database are different from the model of the charging pile that needs to be diagnosed (the staff did not update the fault type database in advance).
[0090] At this point, the fault type database needs to be updated automatically. This involves adding all fault types and corresponding faulty components corresponding to the current charging pile model to the fault type database, combining them to form a new fault type database, and finally adding the current model charging pile's number label to the new fault type database.
[0091] As mentioned above, new fault type databases can be added manually by staff; if staff do not update them in advance, automatic updates can be performed according to the scheme in this embodiment before fault diagnosis can begin.
[0092] Therefore, by adding a new fault type database each time, the fault type database can be continuously expanded as the charging pile models are iterated; thus, this embodiment can be used as a technical solution for long-term application.
[0093] Q5-2-2: Update the weight factors of the prediction model to obtain the updated weight factors; Obtain the weighting factors between the feature nodes and enhancement nodes of the prediction model; The weight factors are updated using an elastic network regression method to obtain updated weight factors.
[0094] Specifically, conventional width learning models (BLS) have the following drawbacks: The weights generated between feature nodes and augmentation nodes are random. Even if the weights are calculated using pseudo-inverse, the random mapping of weights in nonlinear problems will lead to a decrease in classification accuracy.
[0095] Therefore, this invention considers using an elastic network regression method to update the weight factors in order to reduce the randomness of the weights and ensure classification accuracy.
[0096] The following describes the process of applying the elastic network regression method to the BLS model to form IBLS: In traditional BLS, the final output of BLS can be expressed as: Weighting factor W m The pseudo-inverse (least squares method) is usually used to solve this problem, which has two main issues: overfitting risk, when there are many feature nodes and augmentation nodes (high dimension), the pseudo-inverse solution is prone to overfitting; and multicollinearity, the randomly generated feature nodes may have linear correlations, leading to unstable weights.
[0097] Elastic net regression optimizes the weight matrix by combining L1 and L2 regularization: L1 regularization (Lasso regression model): Lasso regression models promote weight sparsity and automatically select feature nodes.
[0098] L2 regularization (Ridge regression model): Ridge regression models handle collinearity issues and stabilize weight values.
[0099] Elastic network regression applies a penalty regularization factor to limit the weights as follows: Formula 1; Among them, W m Indicates the weighting factor. This indicates an update to the weighting factor. This represents the regularization factor in the lasso regression model. This represents the regularization factor of the ridge regression model. The Manhattan norm in L1 regularization is represented. Let represent the Euclidean norm in L2 regularization.
[0100] Regularization is expressed as: Formula 2; in, This represents finding the vector that minimizes the latter expression (in this embodiment, the vector is the weight factor W). m A represents the feature matrix, and Y represents the target output matrix. This represents the mean square error term. n represents the number of samples, t represents the number of iterations, λ is the regularization coefficient, ρ is the hybrid parameter of the elastic network, when ρ=0, the regularization degenerates into pure L2 regularization (Ridge regression model); when ρ=1, the regularization degenerates into pure L1 regularization (Lasso regression model), and Q represents the dimension of the weights, i.e., the total number of weight parameters that need to be regularized. This represents the j-th weight factor.
[0101] By using L1 regularization, the weights of unimportant nodes are compressed to 0, enabling the automatic selection of key nodes while retaining feature nodes and enhancement nodes that are truly useful for prediction.
[0102] Randomly generated feature nodes and augmentation nodes may be highly correlated (e.g., two feature nodes are linearly correlated). The optimization of elastic networks uses L2 regularization to penalize large weights, reducing sensitivity to correlated nodes and avoiding the influence of the weight factor W. m Extreme values appeared in the data.
[0103] It should be noted that: The shortcomings of traditional BLS include: the pseudo-inverse solution may have excessively high variance in the case of small samples (i.e., "overfitting" occurs); due to the continuous advancement of charging pile related technologies, their failure is a low-probability event, and the total number of statistically recorded charging pile failure events is not very large.
[0104] Using the BLS model directly can easily lead to overfitting. When applied to scenarios such as charging pile fault diagnosis, although it can still make predictions, the accuracy of the predictions is slightly lower than the ideal state.
[0105] Therefore, this invention chooses to optimize BLS using elastic network regression, and uses regularization terms to adjust λ1 and λ2 to control model complexity and achieve a balance between training error and generalization performance. Ultimately, this makes the prediction accuracy of the prediction model approach the ideal state.
[0106] In some implementations, the moment of the fault is extracted based on the waveform changes of the three-phase current and switching voltage in the circuit under real-time monitoring; thus, the moment of the fault can be obtained more accurately.
[0107] For example, before a fault occurs, the three-phase current is a sinusoidal curve with three phases of equal phase difference. After a fault occurs, the three-phase current becomes an irregular curve or still remains a sinusoidal curve, but its amplitude changes.
[0108] For example, before the fault occurred, the switching voltage was a stable square wave, but after the fault occurred, it became a spike wave.
[0109] Based on the two examples above, the moment when the current or voltage waveform changes can be extracted (the extraction process is existing technology), thus obtaining the moment when the fault occurs.
[0110] Each time a fault occurs, the above steps are performed to obtain the time of the fault. The ambient temperature and the temperature of the charging pile at the time of the fault are then combined to form new data and added to the historical data. This can increase the total amount of historical data, reduce the overfitting degree of the prediction model training process, and improve the prediction accuracy of the prediction model.
[0111] Q5-2-3: Replace the weight factors in the prediction model with the updated weight factors to obtain the updated prediction model; Q5-2-4: Input the actual ambient temperature, actual three-phase current and actual switching voltage into the updated prediction model to obtain the second actual fault type of the charging pile; the specific method is the same as in case 1.
[0112] Q5-2-5: Match the second actual fault type with multiple fault types in the fault type dataset corresponding to the model of the charging pile to be diagnosed to obtain the faulty component corresponding to the second actual fault type; the specific method is the same as in case 1.
[0113] Based on the above design of the present invention, an IBLS-type prediction model is used, leveraging its reinforcement learning capabilities. When encountering a new model of charging pile during fault diagnosis, the system can adjust the weights to integrate all fault types of the new model with those of existing models, ultimately predicting the fault type.
[0114] In this process, there is no need to retrain the prediction model, which greatly improves the applicability of the prediction model and prevents it from being eliminated due to the iteration of charging pile models.
[0115] The above description is merely a preferred embodiment of the present invention and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention is not limited to the specific combination of the above-described technical features, but also includes other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in this invention.
Claims
1. A method for diagnosing faults in charging piles, characterized in that, include: Obtain the database of charging pile fault types; The fault type database contains fault type datasets for various models of charging piles; the fault type dataset contains all fault types for the same model of charging pile, as well as the faulty components corresponding to each fault type. Obtain a prediction model; the prediction model is used to predict the fault type of the charging pile by taking into account the ambient temperature, three-phase current and switching voltage of the charging pile. The actual three-phase current at the input terminal of the charging pile, the actual switching voltage at the output terminal, and the actual ambient temperature of the charging pile when a fault occurs are detected. Obtain the model number of the charging station to be diagnosed; Based on the model of the charging pile to be diagnosed, match the corresponding fault type dataset in the fault type database; If the match is successful, the actual ambient temperature, actual three-phase current and actual switching voltage are input into the prediction model to obtain the first actual fault type of the charging pile. The first actual fault type is matched with multiple fault types in the fault type dataset to obtain the faulty component corresponding to the first actual fault type.
2. The charging pile fault diagnosis method according to claim 1, characterized in that, The prediction model includes weighting factors; The method for diagnosing charging pile faults also includes: If the matching fails, a fault type dataset corresponding to the model of the charging pile to be diagnosed will be added to the fault type database. Update the weight factors of the prediction model to obtain the updated weight factors; The updated weighting factor is used to replace the weighting factor in the prediction model to obtain the updated prediction model. The actual ambient temperature, actual three-phase current and actual switching voltage are input into the updated prediction model to obtain the second actual fault type of the charging pile. The second actual fault type is matched with multiple fault types in the fault type dataset corresponding to the model of the charging pile to be diagnosed to obtain the faulty component corresponding to the second actual fault type.
3. The charging pile fault diagnosis method according to claim 2, characterized in that, The prediction model is an optimized width learning model. This includes the weighting factor between feature nodes and enhancement nodes; The updated weight factors of the prediction model are obtained by updating the weight factors, including: Obtain the weighting factors between the feature nodes and enhancement nodes of the prediction model; The weight factors are updated using an elastic network regression method to obtain updated weight factors.
4. A method for adjusting and diagnosing the interval of a charging pile, characterized in that, include: S1: Obtain historical data on multiple charging pile malfunctions; the historical data includes: the time of the malfunction and the ambient temperature at the time of the malfunction. S2: Based on the time of the fault and the ambient temperature at the time of the fault, the temperature is divided into low temperature range, normal temperature range and high temperature range; S3: Real-time detection of actual ambient temperature; S4: If the actual ambient temperature is in the low temperature range, then the charging pile is subjected to a charging pile fault diagnosis method as described in any one of claims 1-3 at a first interval; the first interval is the shortest interval between the times when the charging pile malfunctions when the ambient temperature is in the low temperature range. If the actual ambient temperature is within the normal temperature range, then the charging pile is subjected to a charging pile fault diagnosis method as described in any one of claims 1-3 at a second interval; the second interval is the shortest interval between the times when the charging pile malfunctions when the ambient temperature is within the normal temperature range. If the actual ambient temperature is within the high temperature range, then the charging pile is subjected to a charging pile fault diagnosis method as described in any one of claims 1-3 at a third interval; the third interval is the shortest interval between the times when the charging pile malfunctions when the ambient temperature is within the high temperature range.
5. The method for adjusting and diagnosing the interval of a charging pile according to claim 4, characterized in that, The historical data also includes: the temperature of the charging station when the fault occurred; The method for diagnosing charging pile faults also includes: The maximum temperature of the charging pile in the low-temperature range from the historical data is obtained to obtain the first reference temperature; The minimum temperature of the charging pile under high-temperature conditions is obtained from the historical data to obtain the second reference temperature; Real-time monitoring of the actual temperature of the charging pile; If the actual temperature of the charging pile is equal to the first reference temperature, an additional fault diagnosis is immediately performed, and the charging pile is subjected to a charging pile fault diagnosis method as described in any one of claims 1-3 according to the first interval. If the actual temperature of the charging pile is equal to the second reference temperature, an additional fault diagnosis is immediately performed, and the charging pile is subjected to a charging pile fault diagnosis method as described in any one of claims 1-3 according to the third interval.
6. The method for adjusting and diagnosing the interval of a charging pile according to claim 5, characterized in that, A method for diagnosing charging pile faults according to any one of claims 1-3, based on the first interval, includes: After the first interval, the actual temperature of the charging pile is detected to obtain the first charging pile temperature fluctuation. Perform a fault diagnosis on the charging station; If the fluctuating temperature of the first charging pile is greater than or equal to the actual temperature of the charging pile detected during the last fault diagnosis, then the charging pile fault diagnosis method as described in any one of claims 1-3 is performed at a second interval. If the fluctuating temperature of the first charging pile is less than the actual temperature of the charging pile detected during the last fault diagnosis, then the charging pile fault diagnosis method as described in any one of claims 1-3 is performed at the first interval.
7. The method for adjusting and diagnosing the interval of a charging pile according to claim 4, characterized in that, S2: Based on the time of the fault and the ambient temperature at the time of the fault, the temperature is divided into a low-temperature range, a normal-temperature range, and a high-temperature range, including: Based on the time of the failure, the time interval between each failure and the previous failure is calculated to obtain multiple failure intervals; Obtain the median of the multiple fault intervals to get the median fault interval; Calculate the low-temperature critical temperature and high-temperature critical temperature corresponding to the median fault interval; The temperature range below the critical low-temperature temperature is defined as the low-temperature range; The temperature range that is greater than or equal to the low-temperature critical temperature and less than or equal to the high-temperature reference range is set as the normal temperature range; The temperature range above the high-temperature critical temperature is defined as the high-temperature range.
8. The method for adjusting and diagnosing the interval of a charging pile according to claim 7, characterized in that, The calculation of the low-temperature critical temperature and high-temperature critical temperature corresponding to the median fault interval includes: The maximum value among the multiple fault intervals is obtained to obtain the maximum fault interval; Obtain the ambient temperature of the maximum fault interval to obtain the reference ambient temperature; The ambient temperature corresponding to the largest fault interval among the multiple fault intervals that meets the first set condition is taken as the low temperature critical temperature; the first set condition is: less than or equal to the median fault interval, and the corresponding ambient temperature is less than the reference ambient temperature. The ambient temperature corresponding to the largest fault interval among the multiple fault intervals that meets the second set condition is taken as the high-temperature critical temperature; the second set condition is: less than or equal to the median fault interval, and the corresponding ambient temperature is greater than the reference ambient temperature.