Charging pile part life cycle prediction and management method

Abstract

The invention discloses a charging pile part life cycle prediction and management method, and relates to the technical field of charging piles, and the method comprises the steps: constructing a multi-source data collection system, and obtaining the operation state, environment parameters and user behavior data; key characteristic variables are extracted through principal component analysis, and health state evaluation indexes are generated through a coupling aging model; an improved Transform prediction network is designed to realize accurate prediction of the residual life; and making a dynamic operation and maintenance strategy based on a prediction result and optimizing a maintenance period. The prediction precision can be improved through the environment adaptation factor and the nonlinear mapping layer, the maintenance cost is reduced, and the equipment availability rate is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of charging piles, in particular to a charging pile component life cycle prediction and management method. BACKGROUND

[0002] With the popularization of electric vehicles, the reliability and operation and maintenance efficiency of charging piles as core infrastructure have become the focus of the industry. The existing maintenance and management of charging piles mainly rely on periodic maintenance or post-failure maintenance mode, which has significant limitations: the traditional maintenance strategy based on time or fixed threshold is not adaptive enough to the effects of different environmental factors (such as temperature and humidity, salt fog concentration, altitude, etc.) on component aging, which may cause deviation in component replacement timing, increase costs and possibly cause safety hazards; in addition, current life prediction models mainly rely on a single data source (such as charging times and power), and fail to effectively integrate environmental parameters, multi-component collaborative aging effects and user behavior characteristics, resulting in limited prediction accuracy and weak generalization ability; at the same time, there is a lack of dynamic coupling of the whole life cycle management system, which makes it difficult to real-time perceive the health status of components and link to adjust the operation and maintenance strategy, resulting in low resource allocation efficiency.

[0003] Specifically, the existing technology has room for improvement in the following aspects: the influence of environmental factors on the life of charging piles has not been fully modeled, such as the embrittlement of plastic parts in low-temperature environments in the north and the corrosion of metal components in coastal areas by salt mist, which may cause the actual life of the same model of charging pile in different environments to deviate from the predicted value by more than 30%; single component prediction methods (such as finite element simulation of charging gun lines) do not consider the interaction between multiple components, such as the coupling effect of power module aging and the performance degradation of the cooling system, which may increase the risk of cascading failures; in addition, there is a data island phenomenon between charging pile operation data, environmental data and maintenance records, which makes it difficult to provide comprehensive support for the life characteristic index system; the existing maintenance plan also lacks a dynamic optimization mechanism, which cannot flexibly adjust the maintenance period according to real-time load, environmental changes and component health, resulting in high operation and maintenance costs and limited equipment availability. Therefore, we propose a charging pile component life cycle prediction and management method to solve the technical problems raised in the above. SUMMARY

[0004] The present application aims to provide a charging pile component life cycle prediction and management method to solve the problems raised in the background art.

[0005] To achieve the above-mentioned purpose, the present application adopts the following technical solution: comprising the following steps: Step S1, a multi-source data acquisition system is constructed to obtain real-time charging pile operation state data, environmental parameter data and user behavior data, the charging pile operation state data includes charging power, current fluctuation value, heat dissipation module temperature distribution and charging gun line stress distribution, the environmental parameter data includes temperature and humidity, salt fog concentration, altitude and ultraviolet intensity, and the user behavior data includes charging frequency, charging time and abnormal operation record; Step S2, a component life characteristic index system is constructed based on multi-source data, principal component analysis algorithm is used to reduce the dimension of the collected multi-dimensional data, key characteristic variables are extracted, the interaction between different components is modeled by coupling the aging model, a multi-component collaborative aging relationship matrix is formed, and the characteristic variables and the aging relationship matrix are fused to generate a component health state evaluation index; Step S3, a dynamic life prediction model is designed, the component health state evaluation index is input into the improved prediction network based on the Transformer architecture to obtain the component remaining life prediction result; Step S4, a dynamic operation and maintenance strategy is formulated based on the component remaining life prediction result, and a hierarchical maintenance plan is generated combining real-time load changes and environmental conditions.

[0006] As a preferred scheme, in step S1, the environmental parameter data is measured by a micro sensor array, the micro sensor array includes a temperature and humidity sensor, a salt fog concentration detector and an ultraviolet sensor, and each sensor is connected to a central processing unit through a wireless communication module.

[0007] As a preferred scheme, in step S2, the coupling aging model combines the aging rate of the power module, the heat conduction efficiency decay curve of the heat dissipation system and the embrittlement coefficient of the plastic parts to form a multi-component collaborative aging relationship matrix.

[0008] As a preferred scheme, in step S3, the improved prediction network based on the Transformer architecture adds an adaptive weight adjustment layer after the multi-head attention mechanism module, the adaptive weight adjustment layer dynamically adjusts the importance weight of different features by introducing an environmental adaptation factor.

[0009] As a preferred scheme, the environmental adaptation factor in the adaptive weight adjustment layer is calculated in real time by the current environmental parameter data.

[0010] As a preferred scheme, in step S3, the improved prediction network based on the Transformer architecture adds a nonlinear mapping layer before the feedforward neural network module, the nonlinear mapping layer is composed of multiple activation function units and is used to capture complex nonlinear relationships.

[0011] As a preferred solution, the activation function unit in the nonlinear mapping layer comprises a ReLU function, a Sigmoid function and a Tanh function.

[0012] As a preferred solution, in step S3, the residual connection part of the improved prediction network based on the Transformer architecture realizes the cross-time step transmission of feature information by using a bidirectional gated recurrent unit.

[0013] As a preferred solution, in step S4, the dynamic operation and maintenance strategy optimizes the maintenance period through a reinforcement learning module, and the reinforcement learning module takes the maintenance cost, device availability rate and user satisfaction as a reward function.

[0014] As a preferred solution, in step S4, when the component health state evaluation index is lower than a preset threshold, a warning mechanism is triggered and a maintenance task is pushed to an operation and maintenance personnel through a mobile terminal.

[0015] The present application has obvious advantages and beneficial effects compared with the prior art, specifically, from the above technical solution, it mainly has the following advantages: The present application constructs a multi-source data acquisition system to obtain real-time charging pile operation state data, environmental parameter data and user behavior data, extracts key feature variables using a principal component analysis algorithm, and models the multi-component collaborative aging effect through a coupling aging model to generate a component health state evaluation index. The evaluation index is input into an improved prediction network based on the Transformer architecture, and the self-adaptive weight adjustment layer and the nonlinear mapping layer enhance the environmental adaptability and the nonlinear relationship capturing ability, thereby improving the component residual life prediction accuracy. In addition, a dynamic operation and maintenance strategy is formulated based on the prediction results, a hierarchical maintenance plan is generated in combination with real-time load changes and environmental conditions, and a reinforcement learning algorithm is used to optimize the maintenance period, thereby significantly reducing the maintenance cost and improving the device availability rate. The self-adaptive weight adjustment layer introduced in the present application can dynamically adjust the importance weight of different features, enhancing the role of environmental adaptation factors; the nonlinear mapping layer effectively captures complex nonlinear relationships through multiple activation function units, further improving the prediction accuracy.

[0016] To more clearly illustrate the structural features and effects of the present application, the present application will be described in detail below with reference to the accompanying drawings and specific embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0017] Fig. 1 The figure is a structural schematic diagram of the multi-source data acquisition system in the embodiment of the present application. Fig. 2 The figure is an architecture diagram of the dynamic life prediction model in the embodiment of the present application. Fig. 3 A flow chart for generating a dynamic operation and maintenance strategy in an embodiment of the present application is shown.

[0018] Reference signs: 1, micro sensor array; 2, wireless communication module; 3, central processing unit; 4, principal component analysis module; 5, coupled aging model; 6, adaptive weight adjustment layer; 7, nonlinear mapping layer; 8, bidirectional gated recurrent unit; 9, reinforcement learning module; 10, mobile terminal. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application will be further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the present application and do not limit the present application.

[0020] It should be noted that when an element is referred to as being "fixed" to another element, it can be directly on the other element or there can be an intervening element. When an element is referred to as being "connected" to another element, it can be directly connected to the other element or intervening elements can be present. The terms "vertical", "horizontal", "left", "right", and the like as used herein are for purposes of explanation only and are not intended to be limiting.

[0021] Referring to Figs. 1 to 3 The embodiment of the present application provides a charging pile component life cycle prediction and management method, and a structural schematic diagram of a multi-source data acquisition system shows the arrangement mode of a distributed sensor network and the connection relationship with a central processing unit 3. The micro sensor array 1 is arranged at different positions of the external shell of the charging pile, and is used for measuring environmental parameter data in real time, including temperature and humidity, salt fog concentration and ultraviolet intensity. The micro sensor array 1 is connected with the central processing unit 3 through the wireless communication module 2, and the wireless communication module 2 adopts a low-power Bluetooth protocol to ensure the stability and real-time performance of data transmission. The central processing unit 3 is responsible for receiving and preliminarily processing the data from the micro sensor array 1, and simultaneously integrating charging pile operation state data and user behavior data to form a multi-source data set.

[0022] The charging pile operating state data is collected by built-in current sensors, temperature sensors, and stress sensors, which are installed on the key components inside the charging pile, such as power modules, heat dissipation systems, and charging gun wire assemblies. The current sensor is fixed at the input end of the power module to monitor the charging power and current fluctuation value; the temperature sensor is distributed on the heat conduction path of the heat dissipation module to record the temperature distribution of the heat dissipation module; the stress sensor is embedded in the stress concentration area of the charging gun wire to detect the stress distribution of the charging gun wire. These sensors are connected to the central processing unit 3 through wired mode to ensure high reliability of data transmission. User behavior data is sourced from the interactive terminal of the charging pile, which is connected to the central processing unit 3 through an Ethernet interface to transmit information such as charging frequency, charging duration, and abnormal operation records.

[0023] The principal component analysis module 4 performs dimensionality reduction processing on the above-mentioned multi-source data set to extract key feature variables. The principal component analysis module 4 is integrated in the central processing unit 3, and its working principle is to calculate the covariance matrix of the original high-dimensional data and extract principal components that can reflect the main trend of data change based on eigenvalue decomposition. The extracted key feature variables include charging power fluctuation coefficient, heat dissipation module temperature gradient, charging gun wire stress peak value, environmental temperature and humidity change rate, salt fog concentration cumulative amount, and ultraviolet radiation dose. These feature variables constitute the basic data of the component life characteristic index system.

[0024] The coupling aging model 5 is used to model the interaction between different components, and its input is the above-mentioned key feature variables. The core of the coupling aging model 5 is to combine the aging rate of the power module, the heat conduction efficiency decay curve of the heat dissipation system, and the embrittlement coefficient of the plastic parts to form a multi-component collaborative aging relationship matrix. The aging rate of the power module is obtained by experimentally determining the performance degradation law of the power module under different current fluctuation conditions; the heat conduction efficiency decay curve of the heat dissipation system is generated by simulating the influence of long-term running process of the surface oxidation and dust accumulation of the heat dissipation fins; the embrittlement coefficient of the plastic parts is determined based on the material weather resistance test results. The multi-component collaborative aging relationship matrix is generated by weighting and superimposing the aging characteristics of each component, and the weight coefficients are distributed according to the importance of the components in the overall performance of the charging pile. The output results of the coupling aging model 5 are fused with the key feature variables extracted by the principal component analysis module 4 to generate component health status evaluation indexes.

[0025] The architecture of the dynamic life prediction model is as follows Fig. 2As shown, the model is based on an improved Transformer architecture design. The adaptive weight adjustment layer 6 is located after the multi-head attention mechanism module, which dynamically adjusts the importance weight of different features by introducing an environmental adaptation factor. The environmental adaptation factor is calculated in real time from the current environmental parameter data, and its value is dynamically adjusted with the change of environmental conditions. The nonlinear mapping layer 7 is located in front of the feedforward neural network module, which is composed of multiple activation function units, used to capture complex nonlinear relationships. The activation function unit includes ReLU function, Sigmoid function and Tanh function, which are used to process different types of nonlinear features. The bidirectional gated recurrent unit 8 is located in the residual connection part, which is used to realize the cross-time step transmission of feature information. The bidirectional gated recurrent unit 8 ensures the effective use of long time series data through the information flow mechanism in the forward and reverse directions.

[0026] The training process of the dynamic life prediction model is as follows: First, the historical data set is divided into training set and validation set, the training set is used for optimization of model parameters, and the validation set is used for evaluation of generalization ability of the model. The goal of model training is to minimize the prediction error, and the error calculation uses the root mean square error index. During the training process, the weight coefficient of the adaptive weight adjustment layer 6 is gradually optimized by the gradient descent algorithm until convergence. After the model training is completed, the component health status evaluation index is sent into the prediction network as input data, which is processed layer by layer through the multi-head attention mechanism module, the adaptive weight adjustment layer 6, the nonlinear mapping layer 7 and the bidirectional gated recurrent unit 8, and finally outputs the component remaining life prediction result.

[0027] The process of formulating dynamic operation and maintenance strategy based on the component remaining life prediction result is as shown in Fig. 3 When the component health status evaluation index is lower than the preset threshold, the early warning mechanism is triggered, and the early warning information is sent to the mobile terminal 10 through the wireless communication module 2. The mobile terminal 10 displays the specific maintenance task content, including the component name that needs to be maintained, the expected maintenance time and the maintenance priority. For the scenario of multiple components aging together, the key components that have the greatest impact on the overall performance are preferentially maintained. The reinforcement learning module 9 is used to optimize the maintenance period, and its reward function considers three factors: maintenance cost, equipment availability and user satisfaction. The reinforcement learning module 9 continuously adjusts the maintenance priority and resource allocation through the Q-learning algorithm, ensuring that the equipment availability is improved while the maintenance cost is reduced.

[0028] The workflow of the whole system is as follows: the micro sensor array 1 collects environmental parameter data in real time and transmits them to the central processing unit 3 through the wireless communication module 2, the central processing unit 3 receives the operating state data of the charging pile and the user behavior data at the same time. The principal component analysis module 4 processes the multi-source data for dimension reduction, extracts the key characteristic variables and inputs them into the coupled aging model 5 to generate the component health status evaluation index. The evaluation index is then sent to the dynamic life prediction model, which processes it through the adaptive weight adjustment layer 6, the nonlinear mapping layer 7 and the bidirectional gate recurrent unit 8, and outputs the component remaining life prediction result. The prediction result enters the reinforcement learning module 9, generates a hierarchical maintenance plan and pushes the maintenance task to the operation and maintenance personnel through the mobile terminal 10. Through the cooperative work of the distributed sensor network, the central processing unit 3 and the mobile terminal 10, the whole system realizes the accurate prediction and efficient management of the charging pile component life cycle.

[0029] In order to better enable relevant persons in the art to fully understand and implement the present application, the specific implementation principles of the present application are supplemented below in conjunction with a specific application scenario.

[0030] In the electric vehicle charging station in the coastal area, the charging pile is exposed to the environment with high salt mist concentration and high humidity for a long time. Through the method provided by the present application, first, a micro sensor array 1 is arranged at different positions of the outer shell of the charging pile for real-time measurement of environmental parameter data such as temperature and humidity, salt mist concentration and ultraviolet intensity. These data are transmitted to the central processing unit 3 through the wireless communication module 2, and the central processing unit 3 receives the operating state data collected from the current sensor, temperature sensor and stress sensor inside the charging pile, as well as the user behavior data transmitted through the Ethernet interface. After dimension reduction processing by the principal component analysis module 4, the key characteristic variables reflecting the aging trend of the components are extracted, such as the charging power fluctuation coefficient, the temperature gradient of the heat dissipation module, the stress peak value of the charging gun wire, etc.

[0031] Subsequently, the coupled aging model 5 generates a multi-component cooperative aging relationship matrix based on the above-mentioned key characteristic variables, combined with the aging rate of the power module, the thermal conductivity efficiency decay curve of the heat dissipation system and the embrittlement coefficient of the plastic parts. Specifically, the aging rate of the power module is determined by experimentally measuring the performance degradation law of the power module under different current fluctuation conditions; the thermal conductivity efficiency decay curve of the heat dissipation system is generated by simulating the influence of long-term running on the surface oxidation and dust accumulation of the heat dissipation fins; the embrittlement coefficient of the plastic parts is determined based on the material weather resistance test results. These aging characteristics are superimposed by weighting to generate the cooperative aging relationship matrix, and the weight coefficients are distributed according to the importance of each component in the overall performance of the charging pile, and finally the component health status evaluation index is generated.

[0032] The dynamic life prediction model is as follows: Fig. 2As shown, the core improvement lies in the design of the adaptive weight adjustment layer 6 and the nonlinear mapping layer 7. When the environmental parameter data changes significantly, for example, a sudden heavy rain on a certain day causes the environmental humidity to increase sharply, the adaptive weight adjustment layer 6 will dynamically calculate the environmental adaptation factor according to the current humidity value, and adjust the importance weight of the humidity-related features accordingly. The nonlinear mapping layer 7 captures the nonlinear relationship between charging power fluctuations and component aging through the ReLU function, processes the impact of humidity on the embrittlement coefficient of plastic parts through the Sigmoid function, and analyzes the effect of salt spray concentration on the corrosion rate of metal components through the Tanh function. The bidirectional gated recurrent unit 8 ensures the effective use of long time series data, such as the cumulative impact of continuous high humidity environment on component health status in the past week.

[0033] After completing the component remaining life prediction, the system enters the dynamic operation and maintenance strategy generation phase. As shown, Fig. 3 When the health status evaluation index of a certain key component is lower than the preset threshold, the early warning mechanism is triggered, and the relevant information is sent to the mobile terminal 10 through the wireless communication module 2. For example, when the heat conduction efficiency of the cooling system decreases to the critical value, the mobile terminal 10 displays the specific maintenance task content, including the operation steps required to clean the surface oxide and dust of the cooling fins, the estimated maintenance time, and the priority level. For the scenario of multiple components aging together, such as the simultaneous performance degradation of the power module and the cooling system, the reinforcement learning module 9 optimizes the maintenance period through the Q-learning algorithm, considering the maintenance cost, device availability, and user satisfaction, and dynamically adjusts resource allocation. For example, during the summer peak period, the device availability is prioritized, and the maintenance frequency is appropriately increased; while in the winter valley period, the maintenance frequency is reduced to save costs.

[0034] The workflow of the entire system runs through the collaborative work of the distributed sensor network, the central processing unit 3, and the mobile terminal 10. The micro sensor array 1 collects environmental parameter data in real time and transmits it to the central processing unit 3 through the wireless communication module 2, and the central processing unit 3 integrates multi-source data and generates component health status evaluation indexes. The index is input into the dynamic life prediction model, which is processed layer by layer through the adaptive weight adjustment layer 6, the nonlinear mapping layer 7, and the bidirectional gated recurrent unit 8, and outputs the component remaining life prediction result. The prediction result enters the reinforcement learning module 9, generates a hierarchical maintenance plan, and pushes the maintenance task to the operation and maintenance personnel through the mobile terminal 10.

[0035] Through the implementation of the above specific application scenarios, the present application realizes precise prediction and efficient management of the charging pile component life cycle. For example, in a high-salt mist concentration environment, by accurately modeling the corrosion rate of metal components and dynamically adjusting the maintenance cycle, the component replacement cost is significantly reduced, and the overall availability of the equipment is improved. In addition, the system optimizes resource allocation through the reinforcement learning module 9 to ensure that the balance between the lowest maintenance cost and the highest equipment performance can be achieved under different environmental conditions.

[0036] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement and improvement within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for predicting and managing the lifecycle of charging pile components, characterized in that, The method comprises the following steps: Step S1, a multi-source data acquisition system is constructed to obtain charging pile operation state data, environmental parameter data and user behavior data in real time, the charging pile operation state data includes charging power, current fluctuation value, heat dissipation module temperature distribution and charging gun line stress distribution, the environmental parameter data includes temperature and humidity, salt mist concentration, altitude and ultraviolet intensity, and the user behavior data includes charging frequency, charging time length and abnormal operation record; Step S2, a component life characteristic index system is constructed based on multi-source data, principal component analysis algorithm is used to perform dimension reduction processing on the collected multi-dimensional data, key characteristic variables are extracted, the interaction between different components is modeled by coupling an aging model, a multi-component collaborative aging relationship matrix is formed, and the characteristic variables and the aging relationship matrix are fused to generate a component health state evaluation index; Step S3, a dynamic life prediction model is designed, the component health state evaluation index is input into an improved prediction network based on a Transformer architecture to obtain a component residual life prediction result; Step S4, a dynamic operation and maintenance strategy is formulated based on the component residual life prediction result, and a hierarchical maintenance plan is generated in combination with real-time load changes and environmental conditions.

2. The charging pile component life cycle prediction and management method according to claim 1, characterized in that: In step S1, the environmental parameter data is measured by a micro sensor array (1), the micro sensor array (1) includes a temperature and humidity sensor, a salt mist concentration detector and an ultraviolet sensor, and each sensor is connected to a central processing unit (3) through a wireless communication module (2).

3. The charging pile component life cycle prediction and management method according to claim 1, characterized in that: In step S2, the coupling aging model (5) combines the aging rate of the power module, the heat conduction efficiency decay curve of the heat dissipation system and the embrittlement coefficient of the plastic parts to form a multi-component collaborative aging relationship matrix.

4. The charging pile component life cycle prediction and management method of claim 1, wherein: In step S3, the improved prediction network based on the Transformer architecture adds an adaptive weight adjustment layer (6) after the multi-head attention mechanism module, the adaptive weight adjustment layer (6) dynamically adjusts the importance weight of different features by introducing an environmental adaptation factor.

5. The charging pile component life cycle prediction and management method according to claim 4, characterized in that: The environmental adaptation factor in the adaptive weight adjustment layer (6) is calculated in real time based on the current environmental parameter data.

6. The charging pile component life cycle prediction and management method according to claim 1, characterized in that: In step S3, the improved prediction network based on the Transformer architecture adds a nonlinear mapping layer (7) before the feedforward neural network module, the nonlinear mapping layer (7) is composed of multiple activation function units and is used to capture complex nonlinear relationships.

7. The charging pile component life cycle prediction and management method according to claim 6, characterized in that: The activation function units in the nonlinear mapping layer (7) include ReLU function, Sigmoid function and Tanh function.

8. The charging pile component life cycle prediction and management method according to claim 1, characterized in that: In step S3, the residual connection part of the improved prediction network based on the Transformer architecture uses a bidirectional gated recurrent unit (8) to realize cross-time step transmission of feature information.

9. The charging pile component life cycle prediction and management method according to claim 1, characterized in that: In step S4, the dynamic operation and maintenance strategy optimizes the maintenance period through a reinforcement learning module (9), the reinforcement learning module (9) takes maintenance cost, device availability and user satisfaction as a reward function.

10. The charging pile component life cycle prediction and management method of claim 1, wherein: In step S4, when the component health state evaluation index is lower than a preset threshold, a warning mechanism is triggered and a maintenance task is pushed to the operation and maintenance personnel through a mobile terminal (10).

Images