A charger life prediction system based on big data analysis

Abstract

This invention discloses a charger life prediction system based on big data analysis, belonging to the field of computer-aided design technology. It includes a data fusion unit, a benchmark modeling unit, a drift monitoring unit, a risk assessment unit, and a strategy generation unit. The data fusion unit generates structured feature vectors and sends them to the benchmark modeling unit. The benchmark modeling unit establishes a benchmark life prediction model based on the structured feature vectors and sends the benchmark life prediction model to the drift monitoring unit. The drift monitoring unit generates a conceptual drift index and sends it to the risk assessment unit. This invention profoundly reflects the inherent attributes and operating conditions of the equipment, and its accuracy and reliability far exceed those of traditional models relying on a single data source.

Description

Technical Field

[0001] This invention relates to the field of computer-aided design, specifically to a charger life prediction system based on big data analysis. Background Technology

[0002] With the development of new energy vehicles and energy storage technologies, virtual power plants, as an emerging energy management model, have become a key technology for achieving efficient energy utilization and grid stability by aggregating a large number of distributed energy sources to participate in peak shaving and frequency regulation of the power grid. In virtual power plants, the stable operation of a large number of chargers is its core foundation. Therefore, accurate prediction of charger lifespan and early identification of potential cluster failure risks are crucial to ensuring the overall safety and economic benefits of virtual power plants.

[0003] Existing charger life prediction methods are mainly divided into two categories: physical model-based and data-driven. However, in high-risk scenarios such as national-level virtual power plants, these methods are difficult to cope with the concept drift problem caused by factors such as hidden defects in the supply chain and changes in the working environment. That is, the data distribution changes over time, causing the prediction model to fail. Furthermore, traditional methods pursue a single prediction accuracy and lack a mechanism to quantify the safe distance between the system and cluster failure boundaries. They are also unable to proactively adjust operating strategies to balance short-term economic benefits and long-term system survival when risks increase dramatically.

[0004] Therefore, there is an urgent need in this field for a novel technical solution, which is not just a prediction algorithm, but a computer-aided design system capable of risk perception, self-reflection, and dynamic evolution, in order to design and manage the health status and risks of the entire charger cluster. Summary of the Invention

[0005] The purpose of this invention is to provide a charger life prediction system based on big data analysis to solve the problems mentioned in the background art.

[0006] The technical solution of the present invention includes a data fusion unit, a benchmark modeling unit, a drift monitoring unit, a risk assessment unit, and a strategy generation unit;

[0007] The data fusion unit is used to generate structured feature vectors and send the structured feature vectors to the benchmark modeling unit;

[0008] The benchmark modeling unit is used to establish a benchmark lifetime prediction model based on the structured feature vector, and send the benchmark lifetime prediction model to the drift monitoring unit;

[0009] The drift monitoring unit is used to generate a concept drift index and send the concept drift index to the risk assessment unit;

[0010] The risk assessment unit is used to generate a safe distance based on the concept drift index and send the safe distance to the strategy generation unit;

[0011] The strategy generation unit is used to generate an optimal operation strategy based on the safe distance.

[0012] Preferably, the process by which the data fusion unit generates the structured feature vector is as follows:

[0013] Collect high-frequency time-series operational data, static supply chain data, and operation and maintenance event data of the charger cluster;

[0014] The collected data is processed into the structured feature vector.

[0015] Preferably, the process by which the benchmark modeling unit establishes the benchmark lifetime prediction model is as follows:

[0016] The shape and scale parameters are determined by performing maximum likelihood estimation on the structured feature vectors.

[0017] Based on the shape parameters and the scale parameters, a baseline lifetime prediction model using the Weibull distribution is established.

[0018] Preferably, the process by which the drift monitoring unit generates the concept drift index is as follows:

[0019] Fault data within a preset time window is collected, and the observed distribution is generated by fitting the data using a censored data processing method.

[0020] Calculate the KL divergence of the observed distribution relative to the baseline lifetime prediction model, and set the KL divergence as the concept drift index.

[0021] Preferably, the process by which the risk assessment unit generates the safe distance is as follows:

[0022] The difference is obtained by comparing the preset concept drift threshold with the concept drift index.

[0023] If the difference is positive, then the difference is set as the safe distance;

[0024] If the difference is negative or zero, the safety distance is set to zero.

[0025] Preferably, the process by which the strategy generation unit generates the optimal operating strategy first executes:

[0026] Based on the stated safety distance, a risk response adjustment factor is calculated using a preset Sigmoid function.

[0027] Preferably, the process of the strategy generation unit generating the optimal operating strategy further includes:

[0028] Based on the observed distribution, calculate the probability of conditional failure during the future operating cycle;

[0029] Calculate the expected risk cost based on the failure probability of the conditions and the preset cost per failure.

[0030] Preferably, the process by which the strategy generation unit generates the optimal operating strategy is ultimately executed as follows:

[0031] Based on the risk response adjustment factor, the expected risk cost, and the preset expected economic return function, a strategy utility function is constructed.

[0032] The optimal operating strategy is generated by maximizing the strategy utility function.

[0033] This invention provides an improved charger life prediction system based on big data analysis, which has the following improvements and advantages compared with the prior art:

[0034] 1. This system establishes a more accurate and reliable lifespan prediction benchmark. The data fusion unit can integrate high-frequency time-series operational data, static supply chain data, and maintenance event data from the charger cluster to generate high-dimensional structured feature vectors. This multi-source data fusion mechanism enables the benchmark modeling unit to establish a benchmark lifespan prediction model using the Weibull distribution through maximum likelihood estimation. This model is rooted in comprehensive operational realities, profoundly reflecting the inherent attributes and operating conditions of the equipment, and its accuracy and reliability far exceed those of traditional models that rely on a single data source.

[0035] 2. This system achieves proactive monitoring and quantitative assessment of model failure risk. The design of the drift monitoring unit is the core advancement of this invention. Instead of passively relying on the initial model, it collects fault data within a preset time window and uses censored data processing methods to fit and generate an observation distribution, thereby keenly capturing the deviation between the actual operating conditions and the benchmark model. The system calculates this deviation as the KL divergence and innovatively sets it as the concept drift index. The risk assessment unit further compares this index with a preset critical threshold to generate an intuitive safety distance. This design transforms the ambiguous process of system health degradation into a clear and quantifiable risk metric, achieving a fundamental leap from post-analysis to pre-emptive warning.

[0036] 3. This system can generate dynamically optimized operational strategies that balance economy and security. The strategy generation unit calculates a risk response adjustment factor based on a safe distance using a preset sigmoid function, achieving a smooth mapping of risk level to decision weights. Simultaneously, this unit calculates the conditional failure probability based on the latest observed distribution and quantifies the expected risk cost by combining it with the cost of a single failure. Based on the risk response adjustment factor, expected risk cost, and expected economic benefit function, the system constructs a strategy utility function and generates the optimal operational strategy by maximizing it. This approach overcomes the drawback of the separation between prediction and decision-making in traditional technologies, enabling the system not only to identify risks but also to avoid them, dynamically seeking the optimal balance between economic benefits and system survival, and greatly improving the robustness and resilience of critical infrastructure. Attached Figure Description

[0037] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0038] Figure 1 This is a flowchart of a charger life prediction system based on big data analysis according to the present invention. Detailed Implementation

[0039] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0040] Example 1

[0041] Please see Figure 1 This invention provides a charger life prediction system based on big data analysis, including a data fusion unit, a benchmark modeling unit, a drift monitoring unit, a risk assessment unit, and a strategy generation unit;

[0042] The data fusion unit is used to generate structured feature vectors and send the structured feature vectors to the baseline modeling unit;

[0043] The baseline modeling unit is used to build a baseline lifetime prediction model based on structured feature vectors and send the baseline lifetime prediction model to the drift monitoring unit.

[0044] The drift monitoring unit is used to generate a concept drift index and send the concept drift index to the risk assessment unit;

[0045] The risk assessment unit is used to generate a safe distance based on the concept drift index and send the safe distance to the strategy generation unit;

[0046] The strategy generation unit is used to generate the optimal operational strategy based on the safe distance.

[0047] This embodiment provides a computer-aided design system for the health status of a charger cluster in a virtual power plant scenario, capable of avoiding concept drift. It aims to move from passive prediction to proactive risk perception, self-reflection, and dynamic evolutionary strategy design. The system, through logically interconnected data fusion, benchmark modeling, drift monitoring, risk assessment, and strategy generation units, transforms multi-source heterogeneous data into dynamic optimal operating strategies. The data fusion unit collects and processes raw data, generating structured feature vectors. The benchmark modeling unit receives these vectors and establishes a parameterized benchmark lifetime prediction model. The drift monitoring unit compares real-world data with the benchmark model in real time, quantifying and generating a concept drift index. The risk assessment unit transforms this index into an intuitive safety distance indicator. Based on this safety distance, the strategy generation unit designs and outputs optimal operating strategies that balance economic benefits and system survival, thereby ensuring the overall safety and economic efficiency of the virtual power plant.

[0048] Example 2

[0049] The process by which the data fusion unit generates structured feature vectors is as follows:

[0050] Collect high-frequency time-series operational data, static supply chain data, and operation and maintenance event data of the charger cluster;

[0051] The collected data is processed into structured feature vectors.

[0052] The process of establishing a baseline lifetime prediction model using a baseline modeling unit is as follows:

[0053] The shape and scale parameters are determined by performing maximum likelihood estimation on the structured feature vectors.

[0054] Based on shape and scale parameters, a baseline lifetime prediction model using the Weibull distribution is established.

[0055] In this embodiment, the data fusion unit serves as the system's input layer, responsible for collecting various types of data from the charger cluster. The collected data includes high-frequency timing data, such as the voltage of a single charger. Current and internal temperature Static supply chain data, such as charger unique identifier ID, manufacturer M, production batch B, etc.; and operation and maintenance event data, such as historical fault records. The data fusion unit transforms this raw data into a structured feature vector through feature extraction methods. This vector contains information such as the number of charging cycles and average operating temperature, and is then sent to the baseline modeling unit.

[0056] High-frequency timing data, such as voltage Current and internal temperature By calculating its time-series integral, mean, variance, and other statistical characteristics, and combining this with the sliding time window method, information such as the number of charging cycles and average operating temperature can be extracted. Static supply chain data, such as manufacturer, production batch, and maintenance event data, such as historical fault records, are incorporated into the vector as classification features or discrete features. This information is integrated into a structured feature vector containing both continuous and discrete features.

[0057] The benchmark modeling unit receives the structured feature vector and establishes a benchmark lifetime prediction model for a specific charger cluster, serving as a subsequent comparison benchmark. The benchmark modeling unit determines the shape and scale parameters of the Weibull distribution by applying maximum likelihood estimation to the structured feature vector. Based on the determined shape and scale parameters, a benchmark lifetime prediction model using the Weibull distribution is established. Its failure probability density function is defined by the following equation:

[0058]

[0059] in, The charger's operating time is measured in hours. The dimensionless shape parameter reflects the failure mode; As a scale parameter, its dimension is also in hours, representing a typical value of the lifespan of this batch of chargers: e: the base of the natural logarithm;

[0060] Although the Weibull distribution can effectively fit the lifetime data of many products, in practical applications, if there are multiple independent failure modes, a hybrid Weibull distribution or a more complex lifetime model can be considered to more accurately characterize the physical failure mechanism.

[0061] The process by which the drift monitoring unit generates the concept drift index is as follows:

[0062] Fault data within a preset time window is collected, and the observed distribution is generated by fitting the data using a censored data processing method.

[0063] Calculate the KL divergence of the observed distribution relative to the baseline lifetime prediction model, and set the KL divergence as the concept drift index.

[0064] The process of generating a safe distance using a risk assessment unit is as follows:

[0065] The difference is obtained by comparing the preset concept drift threshold with the concept drift index.

[0066] If the difference is positive, then the difference is set as the safe distance;

[0067] If the difference is negative or zero, the safety distance is set to zero.

[0068] In this embodiment, the drift monitoring unit receives a baseline lifetime prediction model sent by the baseline modeling unit. To accurately quantify the deviation between the model and reality, the drift monitoring unit collects data within a preset time window. The actual operating data within the window was used. To address the issue of sparse fault data, a censoring method was employed. This method incorporates chargers that are still operating normally within the window as censored data points into the calculation, thereby fitting and generating an observation distribution that represents the current true situation. The drift monitoring unit calculates the KL divergence of the observed distribution relative to the baseline lifetime prediction model and sets this KL divergence as the concept drift index. The calculation formula is:

[0069]

[0070] in, It is determined by the time window The parameters are obtained by fitting real-time observation data within the system. These are the baseline parameters provided by the baseline lifetime prediction and distribution modeling module; It is the Euler-Macheroni constant; The gamma function is used. The generated concept drift index is sent to the risk assessment unit.

[0071] To enhance the robustness of the model under anomalous conditions, the system should have built-in data validation and outlier handling modules. For example, when the amount of observed data is lower than a preset threshold, historical data or simulation data can be used to supplement it; when outliers appear in the input data, filtering or smoothing algorithms can be used for preprocessing to ensure that the model's behavior under all possible inputs conforms to physical common sense.

[0072] For chargers that are still operating normally within a preset time window, their operating time up to the end of the window is used as censored data points. Together with the data of chargers that have failed, the observation distribution that can represent the current state of the group is fitted by methods such as maximum likelihood estimation.

[0073] The risk assessment unit receives the concept drift index and transforms it into a risk metric to guide decision-making. The risk assessment unit then sets a preset concept drift threshold. Concept drift index compared to real-time computing The difference is obtained through comparison. Here, the concept drift threshold is obtained based on backtesting analysis or simulation calibration of historical cluster failure events, and represents the maximum risk level acceptable to the operator. If the difference is positive, it is set as the safe distance. If the difference is negative or zero, the safety distance is set to zero. The formula for generating the safety distance is as follows:

[0074]

[0075] in, It is a preset concept drift threshold; It is a real-time conceptual drift index calculated by the drift monitoring unit. The final generated safe distance transforms the ambiguous system risk into a clear quantitative indicator, which is then sent to the strategy generation unit as the core basis for decision-making.

[0076] Example 3

[0077] The process of generating the optimal operational strategy by the strategy generation unit also includes:

[0078] Based on the observed distribution, calculate the probability of conditional failure during the future operating cycle;

[0079] Calculate the expected risk cost based on the probability of conditional failure and the preset cost per failure.

[0080] The process by which the strategy generation unit generates the optimal operational strategy, and finally executes it:

[0081] Based on the risk response adjustment factor, expected risk cost, and the pre-defined expected economic return function, a strategy utility function is constructed.

[0082] The optimal operating strategy is generated by maximizing the strategy utility function.

[0083] In this embodiment, the strategy generation unit is the core of the CAD system's decision-making and output. (Received safe distance) Subsequently, the strategy generation unit calculates and generates a risk response adjustment factor based on the safety distance using a preset Sigmoid function. This is used to smoothly map the safety distance to decision weights;

[0084]

[0085] in, and These are preset constants that control the steepness and center point of the function, respectively; e: the base of the natural logarithm;

[0086] constant and The determination of this usually relies on historical fault backtesting data or expert experience, for example, It can be set to the average safe distance value observed in historical backtesting where intervention is initially required; a constant. Then, by adjusting, it can be made to approach a safe distance. At that time, risk response adjustment factor The changes are steeper, ensuring the system can respond quickly to changes in risk; these preset values ​​are not set arbitrarily, but are obtained through regression analysis and simulation of historical operational data. For example, It can be set as the average safe distance threshold that leads to cluster failures as observed in historical data; The value can be calibrated through multiple iterative simulations with the goal of minimizing the actual failure rate and maximizing system benefits;

[0087] It is a safe distance generated by the risk assessment unit;

[0088] The strategy generation unit is based on the latest observation distribution generated by the drift monitoring unit. Calculate the pre-set operating cycle in the future Within, the conditional failure probability of each charger. ;

[0089]

[0090] in, This is a reliability function based on the observed distribution; This represents the current running time of the charger; For a pre-set future operating cycle, such as 24 hours; where, subscript Used to distinguish different individual chargers;

[0091] The strategy generation unit calculates the conditional failure probability and the preset single-failure cost. Calculate the expected risk cost ;

[0092] All chargers

[0093]

[0094] in, Operational strategies that need optimization; It is a strategy for chargers The call intensity factor is dimensionless and its value ranges from [0, 1], representing the degree to which the charger is called; The economic loss from a single failure is determined by historical data analysis. : The conditional failure probability of charger i; : Summation symbol, sums all chargers;

[0095] This cost can take into account a variety of factors, including the economic losses caused by the failure, repair costs, and the potential impact on grid stability and the operator's reputation.

[0096] The cost of a single failure can take into account a variety of factors, including the economic losses caused by the failure, maintenance costs, and potential impact on grid stability and the operator's reputation.

[0097] The strategy generation unit is based on risk response adjustment factors. Expected risk costs and the pre-defined expected economic return function Construct the policy utility function By solving the problem of maximizing the strategy utility function, an optimal operating strategy that can dynamically balance economic benefits and risk costs is generated. :

[0098] Optimal operating strategy The content can be power scheduling instructions or maintenance scheduling priorities for the charger cluster. For example, at a certain moment, the optimal strategy might instruct the system to: limit the power output of charger ID-101 to 80% and issue a maintenance alarm for charger ID-202, where the scheduling intensity factor is... This is a quantitative representation of the power output limit of the charger;

[0099]

[0100] in, S represents the expected economic return function related to the strategy; S: represents the variable of the operational strategy to be optimized. Risk response adjustment factor; : Expected risk costs associated with the strategy;

[0101] Expected economic return function This can be achieved through the charger scheduling strategy. This can be constructed by mapping it to the economic value it generates; for example, it can be defined as a weighted sum of the number of chargers scheduled, the charging duration, or the charging power, and can be expressed in the form of:

[0102]

[0103] in It is the first The economic benefit weight of each charger, and That is the strategy Next Power or service provided by a charger; S: Charger scheduling strategy; In Interpreted as a strategy The selected set of chargers;

[0104] This process has greatly enhanced the robustness and resilience of critical infrastructure such as virtual power plants when faced with existential threats;

[0105] This model achieves dynamic optimization of the core problem within the current framework. However, in practical applications, the comprehensiveness and prediction accuracy of the model can be enhanced by further introducing external environmental factors, such as the physical coupling relationship between temperature, humidity and charger.

[0106] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A charger life prediction system based on big data analysis, characterized by, The data fusion unit, the benchmark modeling unit, the drift monitoring unit, the risk assessment unit and the strategy generation unit are included. The data fusion unit is configured to generate a structured feature vector and send the structured feature vector to the benchmark modeling unit. The benchmark modeling unit is configured to establish a benchmark life prediction model based on the structured feature vector and send the benchmark life prediction model to the drift monitoring unit. The drift monitoring unit is configured to generate a concept drift index and send the concept drift index to the risk assessment unit. The risk assessment unit is configured to generate a safety distance based on the concept drift index and send the safety distance to the strategy generation unit. The strategy generation unit is configured to generate an optimal operation strategy based on the safety distance. The data fusion unit generates the structured feature vector in the following process: High-frequency time-series operation data, static supply chain data and operation and maintenance event data of a charger cluster are collected. The collected data is processed into the structured feature vector. The benchmark modeling unit establishes the benchmark life prediction model in the following process: Shape parameters and scale parameters are determined by maximum likelihood estimation of the structured feature vector. The benchmark life prediction model using Weibull distribution is established according to the shape parameters and the scale parameters. The drift monitoring unit generates the concept drift index in the following process: Fault data within a preset time window are collected, and an observation distribution is generated by fitting using a censored data processing method. The KL divergence of the observation distribution relative to the benchmark life prediction model is calculated, and the KL divergence is set as the concept drift index.

2. The charger life prediction system based on big data analytics as claimed in claim 1 wherein, The risk assessment unit generates the safety distance in the following process: A preset concept drift critical threshold is compared with the concept drift index to obtain a difference value. If the difference value is positive, the difference value is set as the safety distance. If the difference value is negative or zero, the safety distance is set to zero.

3. The charger life prediction system based on big data analytics as claimed in claim 2, wherein, The strategy generation unit generates the optimal operation strategy in the following process: A risk response adjustment factor is calculated by a preset Sigmoid function according to the safety distance.

4. The charger life prediction system based on big data analytics as claimed in claim 3, wherein, The strategy generation unit generates the optimal operation strategy in the following process: A conditional failure probability in a future operation period is calculated according to the observation distribution. An expected risk cost is calculated according to the conditional failure probability and a preset single fault cost.

5. The charger life prediction system based on big data analytics as claimed in claim 4, wherein, The strategy generation unit generates the optimal operation strategy in the following process: A strategy utility function is constructed based on the risk response adjustment factor, the expected risk cost and a preset expected economic benefit function. The optimal operation strategy is generated by solving the maximization of the strategy utility function.

Images