A charging station operation decision system and an operation decision method
By using the charging station operation management module, the operation decision-making big data model module, and the comprehensive dataset of charging stations, autonomous root cause diagnosis and decision generation for charging station operation have been achieved, solving the problems of judgment errors and response delays caused by reliance on manual labor, and improving operational efficiency and safety.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHUHAI TITANS TECH
- Filing Date
- 2026-03-06
- Publication Date
- 2026-06-05
AI Technical Summary
The current operation and management of charging stations relies on human experience and knowledge base Q&A, which leads to problems such as judgment errors, delayed response and high maintenance costs, making it difficult to achieve efficient and automated operation decisions.
By employing a charging station operation management module, an operation decision-making big data model module, and a comprehensive charging station dataset, and through multimodal information data processing, training, and analysis, autonomous root cause diagnosis and decision generation are achieved, reducing manual intervention and improving the automation and accuracy of operation strategies.
It improves the safety and reliability of charging station operation, shortens the time cycle from problem discovery to handling, reduces system maintenance costs, and improves operational service efficiency.
Smart Images
Figure CN122155266A_ABST
Abstract
Description
Technical Field
[0001] This invention applies to the technical field of charging station systems, and particularly relates to a charging station operation decision-making system and operation decision-making method. Background Technology
[0002] In recent years, with the rapid development of the new energy vehicle industry, charging stations, as energy replenishment units for electric vehicles, have become very important infrastructure and operating sites. In order to ensure business revenue and improve service efficiency, operators have invested resources and technology in equipment operation and maintenance, site management, charging services and energy dispatch of charging stations, so as to ensure the stable operation of the charging station system and carry out intelligent and refined management, thereby ensuring safety and reliability.
[0003] Currently, intelligent operation and management is mainly achieved through data analysis to assist human decision-making in charging systems. This is primarily achieved through three methods: data visualization analysis, automatic system alarms, and a knowledge base Q&A system. Data visualization analysis refers to the information system visualizing station operation data, allowing the operations team to analyze data charts and perform manual diagnostics to take appropriate measures. Automatic system alarms involve the business system monitoring equipment status and charging processes in real time. In the event of a fault or anomaly, the system automatically sends notifications to promptly notify human intervention. The knowledge base Q&A system compiles common operational issues and historical events into an online knowledge base, allowing operations personnel to obtain solutions to relevant problems through a question-and-answer format, assisting in making more reasonable judgments and decisions.
[0004] The aforementioned methods rely on operators' experience in analyzing displayed data and their timely response to notifications. This is susceptible to the influence of their own expertise and subjective biases, leading to oversights or misjudgments. Furthermore, personnel changes can cause inconsistencies and discrepancies in decision-making standards. The process from viewing data to analysis and judgment, from receiving alerts to responding, all require a certain time frame and are sometimes constrained by processes and operational management systems. This often results in delays in the decision-making and implementation of operational strategies, reducing the actual effectiveness of strategy responses. In contrast, knowledge-based question-and-answer decision-making methods are limited by the completeness of the knowledge base and the capabilities of the questioners. The ability to find the most appropriate measures for timely and accurate implementation is uncertain, and the continuous improvement of the question-and-answer system consumes significant human resources for maintenance.
[0005] Therefore, there is a need for a charging station operation decision-making system and method that can autonomously analyze and diagnose abnormal situations in charging station operation and generate decision-making suggestions, thereby significantly reducing reliance on human experience, improving the automation of operational decision-making, reducing the impact of subjective judgment errors and personnel changes, and improving operational service efficiency. Summary of the Invention
[0006] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a charging station operation decision system and operation decision method that reduces the degree of reliance on manual labor, reduces the probability of judgment errors, and can improve the efficiency of operation services.
[0007] The technical solution adopted in this invention is as follows: the charging station operation decision system includes a charging station operation management module, an operation decision large model module, and a comprehensive charging station dataset; The charging station operation management module is used to control the charging station and collect charging station operation data. The large-scale operation decision model module is used to understand and train the multimodal information data of the comprehensive dataset of the charging station, and to perform fusion reasoning and continuous analysis. It automatically diagnoses the root causes of business anomalies and generates operation strategies, which are then output to the charging station operation management module and the manual control terminal. The large-scale operation decision model module includes a multimodal recognition encoder, a business root cause reasoning model, a self-supervised training module, and a planning decision orchestrator. The charging station omnidirectional dataset is used to store multimodal information data and communicate with the operation decision-making big model module, which then performs understanding and training based on the multimodal information data.
[0008] As can be seen from the above scheme, the charging station operation management module executes the operation control of charging equipment within the charging station according to the operation strategy. Simultaneously, the module collects various parameters through a data acquisition device and stores them in the comprehensive dataset of the charging station, providing sufficient data samples for the inference and analysis of the large-scale operation decision-making model module. The large-scale operation decision-making model module, through a multimodal recognition encoder, a business root cause reasoning model, a self-supervised training module, and a planning and decision orchestrator, sequentially processes, trains, infers, analyzes, and generates operation strategies from multimodal information data. The model's inference of multimodal information data enables rapid processing of operation decisions, thereby reducing reliance on manual intervention and increasing the risk of operational strategy errors, thus improving the safety and reliability of charging station operation. Furthermore, by reducing manual intervention, the large-scale operation decision-making model module issues commands to the charging station operation management module, avoiding the slow response speed and low processing efficiency problems caused by traditional manual processing. Through the fusion reasoning and continuous analysis of large models, the system can quickly identify operational problems and verify handling strategies in a simulated environment, shortening the time cycle from problem discovery to action and improving operational service efficiency. By integrating multimodal information with a business knowledge base, it enables proactive root cause reasoning and decision orchestration, changing the dependence of traditional question-answering systems on the completeness of the knowledge base and the ability of the questioner, and also greatly reducing the system's own maintenance costs.
[0009] A preferred embodiment is that the multimodal recognition encoder is used to parse and classify multimodal information data; The business root cause reasoning model is used to analyze the parsed and categorized data information in conjunction with business knowledge base rules; The self-supervised training module is used to perform self-supervised learning and training based on the analysis results, and to optimize the logical sequence of the operational strategies. The planning decision orchestrator is used to match logical order according to preset rules, and generate a corrected operation strategy and the control mode of the charging station operation management module.
[0010] A further preferred embodiment is that the self-supervised training module includes a digital twin environment unit, which is used to simulate and predict the effect of the strategy based on the modified operation strategy and the control mode of the charging station operation management module.
[0011] A preferred embodiment is that the multimodal information data includes text logs, time-series data, and images and videos related to the operation of the charging station. The text logs include system logs and operation reports, the time-series data includes charging current, charging voltage, and device temperature, and the images and videos include site photos and site monitoring videos.
[0012] A preferred approach is that, after generating the operational strategy, the operational decision-making big data model module sends it to the charging station operation management module and the manual control terminal via instruction issuance, strategy push, or operational report.
[0013] An operational decision-making method that includes the following specific steps: Step S1: The charging station operation and management module collects and summarizes data information of the charging station and its surrounding environment through the acquisition module to form multimodal information data, and stores the multimodal information data in the charging station all-round data center; Step S2: The operation decision-making big model module retrieves multimodal information data from the charging station's all-round dataset and preprocesses the multimodal information data to form an analyzable dataset; Step S3: The multimodal recognition encoder performs multimodal encoding on different types of data and transforms them into vectorized representations that can be understood by large models; Step S4: Construct a business knowledge base containing industry knowledge, operation and maintenance procedures and practical results as the basis for reasoning, and use charging station data to pre-train and fine-tune the business root cause reasoning model. Step S5: The business root cause reasoning model combines multimodal information data and business knowledge base analysis to generate preliminary decision recommendations; Step S6: The planning decision orchestrator transforms the decision recommendations generated by the business root cause reasoning model into an executable sequence of instructions. Step S7: Before issuing the instruction, the strategy effect prediction is performed through the digital twin environment unit in the self-supervised training module; Step S8: Output the verified final operation strategy to the charging station operation management module; Step S9: The charging station operation management module executes instructions and collects the effect data after execution to continuously optimize the business root cause reasoning model. Attached Figure Description
[0014] Figure 1 This is a system block diagram of the charging station operation decision-making system; Figure 2 This is a flowchart of the operational decision-making method. Detailed Implementation
[0015] like Figure 1 As shown, in this embodiment, the charging station operation decision system includes a charging station operation management module 1, an operation decision large model module 2, and a comprehensive charging station dataset 3. The charging station operation management module 1 is used to control the charging station and collect charging station operation data; The large-scale operation decision model module 2 is used to understand and train the multimodal information data of the comprehensive dataset 3 of the charging station, and to perform fusion reasoning and continuous analysis. It automatically diagnoses the root causes of business anomalies and generates operation strategies. The operation strategies are output to the charging station operation management module 1 and the manual control terminal. The large-scale operation decision model module 2 includes a multimodal recognition encoder, a business root cause reasoning model, a self-supervised training module, and a planning decision orchestrator. The charging station all-around dataset 3 is used to store multimodal information data and communicate with the operation decision big model module 2, which then performs understanding and training based on the multimodal information data.
[0016] In this embodiment, the multimodal recognition encoder serves as a front-end component of the operational decision-making big model module 2. It preprocesses multimodal information data, cleans and processes various multimodal information input to the model before recognition, serializes and encodes various data, and implements preliminary business identification and association fusion. The business root cause reasoning model is used to understand and reason about various multimodal vector information, and to continuously analyze and synthesize reasoning based on a pre-set business knowledge base to retrieve business rules and causal data, thereby generating root cause determination and preliminary operational decisions. The self-supervised training module drives a large model to perform extensive unsupervised learning. This module includes a digital twin environment unit, which performs simulation predictions and feedback comparisons to evaluate and verify the effectiveness of the strategy. Machine learning algorithms are then used for iterative deduction and fine-tuning to improve the accuracy of business understanding. The digital twin environment unit simulates and predicts the effect of the revised operational strategy based on the control mode of the charging station operation management module 1. The machine learning algorithm can be a decision tree algorithm, a neural network algorithm, etc.
[0017] The planning and decision orchestrator will match and compare the sequence of operational decision suggestions obtained through training with a pre-set standardized decision rule model to obtain various output forms of decision suggestions at different levels, such as direct execution, suggestion recommendation and report export.
[0018] In this embodiment, the multimodal information data includes all collectable text logs, time-series data, and multidimensional information such as images and videos related to the charging station. It includes static and dynamic data, which are structured, continuous, and multimedia data. It is continuously acquired and collected in real time through a series of physical devices, information systems, and data carriers related to the charging station business.
[0019] In this embodiment, after the operation decision-making big model module 2 generates the operation strategy, it sends it to the charging station operation management module 1 and the manual control terminal through instruction issuance, strategy push or operation report.
[0020] like Figure 2 As shown, the operational decision-making method includes the following specific steps: Step S1: The charging station operation management module 1 collects multimodal information data through the sensors built into the charging pile, the station camera, the environmental monitoring equipment, the service station operation monitoring system, the payment system, and the external data interface. This data includes charging time-series data, text data, image and video data, and business data, which are then collected and aggregated to form multimodal information data. This multimodal information data is then stored in the charging station's comprehensive dataset 3. The charging time-series data includes voltage, current, power, and temperature; the text data includes system logs, maintenance reports, and user feedback; the image and video data includes station monitoring video streams and equipment images; and the business data includes charging orders, electricity prices, and user information. Step S2: The operation decision big model module 2 retrieves multimodal information data from the charging station all-round dataset 3, preprocesses the multimodal information data, uses automated scripts or tools to process missing values and remove outliers, standardizes the data, unifies data from different sources and of different types into a standardized format, and then performs data association and fusion. Step S3: The multimodal recognition encoder performs multimodal encoding on different types of data, transforming them into vectorized representations that can be understood by large models. Specifically, text data is converted into text vectors using the BERT (Bidirectional Encoder Representations from Transformers) architecture, features are extracted from time-series data and converted into time-series vectors using the Transformer encoder, and spatial and visual features are extracted from image and video data and converted into visual vectors using CNN (Convolutional Neural Network) encoders, etc. The multimodal encoder projects different modal vectors onto a unified semantic space, enabling the business root cause reasoning model to understand the inherent relationships between various vectors, such as the correlation between images of "voltage rise" and "device overheating" and "overload warning" logs. Step S4: Construct a business knowledge base containing industry knowledge, operation and maintenance procedures, and practical results as the basis for reasoning. This allows the business root cause reasoning model to quickly retrieve problem-related information from the business knowledge base, ensuring the professionalism and accuracy of reasoning and decision-making. The business knowledge base is constructed by vectorizing and storing unstructured and structured documents such as equipment-related parameters, standard operating procedures, fault code libraries, historical work order records, and industry reports, forming a business knowledge base that can be retrieved in real time. The self-supervised training module continuously and extensively applies a large amount of unsupervised learning to the business root cause reasoning model using a large amount of charging station data, enabling the model to grasp the inherent patterns and relationships within the data, such as the common occurrence of "abnormal current fluctuations" accompanied by "temperature increases." Finally, supervised fine-tuning of the model is performed using labeled data, enabling it to both reproduce data flow and understand business logic. Step S5: The business root cause reasoning model combines multimodal information data and business knowledge base analysis to generate preliminary decision suggestions. Specifically, when the system detects an anomaly, such as "the utilization rate of a certain charging pile drops sharply in the afternoon," the business root cause reasoning model will combine the multimodal vector information generated in step S3 to form a comprehensive perception of the current event. Then, it will retrieve relevant cases and rules from the business knowledge base in step S4 and perform reasoning chain analysis in combination with historical data, such as "non-charging vehicles are identified through monitoring video -> logs show no charging requests -> it is determined to be caused by parking space occupation." Finally, the root cause of the problem is located, and preliminary decision suggestions are generated. Step S6: The planning and decision orchestrator transforms the decision recommendations generated by the business root cause reasoning model into an executable sequence of instructions. Specifically, the generated decision recommendations, such as "implement dynamic price reductions to enhance competitiveness," are translated into machine instructions. The planning and decision orchestrator can then build upon an executable instruction set, such as adjust_pricing(station_id, new_price) and send_coupon(user_group, value), to transform natural language-style decisions into one or more ordered, executable atomic instructions, while ensuring that the instructions comply with security specifications. Step S7: Before the instruction is issued, the self-supervised training module will simulate the programmed instructions in the digital twin environment unit built based on historical data. For example, after simulating the execution of the price reduction strategy, it will predict the changes in passenger flow and revenue in the next few hours and the impact on the power grid load. It will evaluate the potential risks and benefits of the decision through the prediction and make adjustments and optimizations based on the prediction results, so as to minimize the losses caused by trial and error in real scenarios and take safety into account. After step S8, instruction arrangement and simulation verification are completed, the decision results are formally output to the charging station operation management module 1 through different control methods. The control methods include direct instruction issuance, strategy recommendation, and operation report generation. For mature and automated operations, the system directly executes atomic instructions through API or system interface. For complex decisions that require manual judgment, strategy suggestions are pushed to the operators for their decision-making reference. A comprehensive operation report containing current situation analysis, root cause inference, and decision expectations is generated to support the final decision-making of the management. Step S9: After receiving instructions or strategies, the charging station operation management module 1 will drive the hardware and software systems or guide the operators to perform operations, and continuously monitor the effects of the strategy execution. For example, after implementing a price reduction strategy, it will track the changes in core indicators such as the utilization rate and order volume of the charging pile in real time. The execution results and effect data are used as new feedback data, recorded and transmitted back to the data collection terminal to provide a basis for the continuous optimization of the business root cause reasoning model. The optimization process specifically involves comparing the collected feedback data with the expected effects to form reward signals. Based on these signals, the parameters of the operation decision-making big model module 2 are fine-tuned through mechanisms such as reinforcement learning or online learning.
[0021] The above methods enable the model to learn and update from each decision-making practice, continuously correcting its decision-making logic and making its intelligence level more efficient and of higher quality.
[0022] Although the embodiments of the present invention are described with reference to actual solutions, they do not constitute a limitation on the meaning of the present invention. Modifications to the embodiments and combinations with other solutions based on this specification will be obvious to those skilled in the art.
Claims
1. A charging station operation decision-making system, characterized in that: It includes a charging station operation and management module (1), an operation decision-making big model module (2), and a comprehensive charging station dataset (3). The charging station operation management module (1) is used to control the charging station and collect charging station operation data; The large-scale operation decision model module (2) is used to understand and train the multimodal information data of the comprehensive dataset (3) of the charging station, and to perform fusion reasoning and continuous analysis. It automatically diagnoses the root causes of business anomalies and generates operation strategies. The operation strategies are output to the charging station operation management module (1) and the manual control terminal. The large-scale operation decision model module (2) includes a multimodal recognition encoder, a business root cause reasoning model, a self-supervised training module, and a planning decision orchestrator. The charging station all-round dataset (3) is used to store multimodal information data and communicate with the operation decision big model module (2), which is then trained by the operation decision big model module (2) based on the multimodal information data.
2. The charging station operation decision-making system according to claim 1, characterized in that: The multimodal recognition encoder is used to parse and classify multimodal information data; the business root cause reasoning model is used to analyze the parsed and classified data information in combination with business knowledge base rules; the self-supervised training module is used to perform self-supervised learning and training based on the analysis results, and optimize the logical order of the operation strategy; the planning decision orchestrator is used to match the logical order according to preset rules, and generate the corrected operation strategy and the control mode of the charging station operation management module (1).
3. The charging station operation decision-making system according to claim 2, characterized in that: The self-supervised training module includes a digital twin environment unit, which is used to simulate and predict the effect of the strategy based on the modified operation strategy and the control mode of the charging station operation management module (1).
4. The charging station operation decision-making system according to claim 1, characterized in that: The multimodal information data includes text logs, time-series data, and images and videos related to charging station operation. The text logs include system logs and operation reports. The time-series data includes charging current, charging voltage, and device temperature. The images and videos include site photos and site monitoring videos.
5. The charging station operation decision-making system according to claim 1, characterized in that: After generating the operation strategy, the operation decision-making big model module (2) sends it to the charging station operation management module (1) and the manual control terminal through instruction issuance, strategy push or operation report.
6. An operational decision-making method, implemented based on a charging station operational decision-making system according to any one of claims 1 to 5, characterized in that, It includes the following specific steps: Step S1: The charging station operation management module (1) collects and summarizes data information of the charging station and its surrounding environment through the acquisition module to form multimodal information data, and stores the multimodal information data in the charging station all-round dataset (3); Step S2: The large-scale operation decision model module (2) retrieves multimodal information data from the all-round dataset (3) of the charging station and preprocesses the multimodal information data to form an analyzable dataset; Step S3: The multimodal recognition encoder performs multimodal encoding on different types of data and transforms them into vectorized representations that can be understood by large models; Step S4: Construct a business knowledge base containing industry knowledge, operation and maintenance procedures and practical results as the basis for reasoning, and use charging station data to pre-train and fine-tune the business root cause reasoning model. Step S5: The business root cause reasoning model combines multimodal information data and business knowledge base analysis to generate preliminary decision recommendations; Step S6: The planning decision orchestrator transforms the decision recommendations generated by the business root cause reasoning model into an executable sequence of instructions. Step S7: Before issuing the instruction, the strategy effect prediction is performed through the digital twin environment unit in the self-supervised training module; Step S8: Output the verified final operation strategy to the charging station operation management module (1). Step S9: The charging station operation management module (1) executes instructions and collects the effect data after execution to continuously optimize the business root cause reasoning model.