Operation management system and method

Abstract

The invention relates to an operation management system and method, and relates to the technical field of intelligent operation management, and the system comprises a data access and standardization unit which carries out the unified access and standardization processing of multi-source heterogeneous operation data including cross-brand and cross-protocol equipment, and lays a high-quality data foundation for subsequent analysis; the intelligent analysis and prediction unit is used for carrying out deep analysis by utilizing an advanced algorithm model based on the standardized data, outputting an accurate prediction result and business insight, and constructing a dynamic knowledge base; and the intelligent strategy making and executing unit is used for making and automatically executing operation strategies such as predictive maintenance and dynamic resource optimization scheduling according to a prediction result and knowledge base insight and in combination with an optimization target, and feeding back a strategy execution effect to realize closed-loop learning and system self-adaption. According to the method, the operation efficiency and the resource utilization level can be effectively improved, the equipment reliability and the active maintenance capability are enhanced, and prospective intelligent operation management is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent operation management, in particular to an operation management system and method. BACKGROUND

[0002] With the rapid development of Internet of Things technology and the continuous expansion of the new energy industry, especially in the field of new energy vehicle charging and battery swapping, the complexity and refinement of operation management are increasingly demanding. Traditional operation management modes gradually reveal their limitations in coping with massive, heterogeneous device data access, changing operating environments, and diverse user needs.

[0003] In the prior art, the data processing capability of the operation management system often cannot adapt to the wide access of cross-brand and cross-protocol devices, resulting in the widespread existence of data islands, and it is difficult to effectively integrate and utilize information. Decision-making often relies on experience or lagging statistical analysis, lacking accurate insight into real-time operating conditions and scientific prediction of future trends, which makes the allocation and scheduling of resources (such as charging piles, power, and maintenance manpower) often in a passive response state, making it difficult to achieve global optimization and maximum efficiency, thereby limiting the improvement of overall operational efficiency.

[0004] In addition, current device maintenance management mostly adopts fixed periodic maintenance or responsive maintenance after failure. This approach fails to fully consider the differences in actual operating conditions and individual health status of devices, lacking the ability to provide early warning of potential failure risks. As a result, on the one hand, unnecessary maintenance may be performed on devices in good condition, increasing operational costs; on the other hand, it is also unable to effectively avoid sudden device failures, leading to service interruptions, affecting user experience, and possibly causing higher emergency repair costs and safety hazards.

[0005] Furthermore, many existing operation management systems lack intelligence in terms of analysis models and strategy rules, which once deployed, often lack effective continuous learning and self-optimization mechanisms. The system is difficult to automatically iteratively update the accuracy of its prediction model and the adaptability of its strategy logic according to the continuous accumulation of operating data and dynamic changes in the external environment. At the same time, the data and service interaction capabilities with the external ecosystem are relatively weak, limiting the innovation of operation modes and the expansion of value chains, making it difficult for the system to maintain efficient operation and adapt to rapidly changing market demands in the long term. SUMMARY

[0006] The purpose of the present application is to provide an operation management system and method, which solves the technical problem that the existing operation management system is difficult to achieve efficient, proactive, and self-adaptive intelligent management of complex operation systems due to low data processing efficiency, insufficient prediction capability, and lagging strategy adjustment.

[0007] In a first aspect, the present application provides an operation management system, comprising: a data access and normalization module configured to uniformly access and normalize operation-related data from multiple heterogeneous data sources including cross-brand and cross-protocol devices, to output normalized data; an intelligent analysis and prediction module coupled to the data access and normalization module to receive the normalized data, and configured to perform deep analysis and intelligent prediction modeling based on the normalized data, to output analysis and prediction results; an intelligent strategy formulation and execution module coupled to the intelligent analysis and prediction module to receive the analysis and prediction results, and configured to formulate context-aware operation strategies based on the analysis and prediction results and drive their automated execution.

[0008] Preferably, the data access and normalization module, when performing uniform access of operation-related data, further comprises: automatically identifying the communication protocol of the accessed cross-brand and cross-protocol devices and establishing a secure data transmission link to collect device operation data; and integrating third-party services and external environment data sources through an open API interface to obtain supplementary operation-related data.

[0009] Preferably, the data access and normalization module, when performing normalization of operation-related data, further comprises: performing data cleaning, data format unification, and data fusion processing on the multiple heterogeneous operation-related data accessed through the data access and normalization module, to form the normalized data containing context information for use by the intelligent analysis and prediction module.

[0010] Preferably, the intelligent analysis and prediction module further comprises: using big data analysis technology and artificial intelligence algorithms to model the received normalized data to generate the analysis and prediction results, the modeling at least including one of the following functions: device health state assessment and failure probability prediction, wherein the device has a probability of failure within a future time window is determined by the following formula: ; wherein, is a specific device; is a predicted time window; is a discrete time point; is the device at time point ​A feature vector comprising at least one of historical equipment operation data, real-time sensor data, operating environment data, and operating condition data; is a pre-trained fault prediction model; are parameters of the fault prediction model; Or, user behavior pattern recognition and charging demand prediction.

[0011] Preferably, when the modeling includes user behavior pattern recognition and charging demand prediction functions, the charging demand prediction function further includes predicting the charging demand at a specific location by the following formula: and future time Charging demand : ; in, For a specific geographical location; The lead time for demand forecast; is a discrete time point; For location Deadline Historical demand data series; For time Calendar features; For location In the future Weather forecast data; A pre-trained demand forecasting model; are parameters of the demand forecasting model; Furthermore, the intelligent analysis and prediction module is further configured to extract business insights from the analysis and prediction results generated thereby, and to store the business insights and the trained models in a structured manner in a dynamic knowledge base.

[0012] Preferably, the intelligent strategy formulation and execution module is further configured to formulate and drive the execution of an operation strategy based on the analysis and prediction results, and the operation strategy includes at least one of the following: Predictive maintenance and proactive health management strategies based on the equipment failure probability prediction results; Dynamic resource optimization scheduling and coordination strategy, which maximizes the total utility of the system by solving To determine the target optimization problem, the It is expressed by the following objective function: ; in, is the scheduling period; is a collection of charging locations; for a particular location in a set of available devices for the location a particular device in a particular time point in a set of users having charging requests at time a particular user in a decision variable representing the assignment of charging requests of users to devices of locations at time a utility function for a single assignment; a predicted charging demand for the location at time a health status of the device at time or, based on real-time monitoring and anomaly detection results of operation-related data, intelligent early warning and emergency response strategies.

[0013] Preferably, the intelligent strategy formulation and execution module further comprises: records the execution process and results of the operation strategy, and feeds back the data of the execution process and results to the intelligent analysis and prediction module for evaluating the strategy effect and assisting the iteration of the model in the intelligent analysis and prediction module.

[0014] Preferably, the further comprises: an open API interface management module configured to support bidirectional data and service capability interaction with external ecosystems, including allowing the data access and normalization module to integrate external data through API, and allowing the processed valuable aggregated information in the analysis and prediction results output by the intelligent analysis and prediction module to be shared through the open API interface to authorized ecological partners.

[0015] Preferably, the system further comprises: having the ability of continuous learning and self-adaptive evolution, which is realized by the following ways: continuously evaluating the overall operation performance and the accuracy of the analysis and prediction results output by the intelligent analysis and prediction module; and based on the evaluation results, iteratively optimizing the model parameters in the intelligent analysis and prediction module and / or the strategy logic in the intelligent strategy formulation and execution module. ​​​​​​​​​

[0016] In a second aspect, the application provides an operation management method, comprising the following steps: S1, through a data access and normalization process, uniformly accessing and normalizing operation-related data from multi-source heterogeneous data sources including cross-brand and cross-protocol devices to output normalized data; S2, through an intelligent analysis and prediction process, receiving the normalized data and utilizing a unified data platform to perform deep analysis and intelligent prediction modeling based on the normalized data to generate business insights and prediction results; S3, through an intelligent strategy formulation and execution process, receiving the business insights and prediction results and formulating context-aware operation strategies and driving their automatic execution based on the business insights and prediction results.

[0017] In summary, the application includes at least one of the following beneficial technical effects: 1. The application significantly improves the overall efficiency and resource utilization level of operation management. Through the data access and normalization module, the unified access and standardized processing of multi-source heterogeneous data including cross-brand and cross-protocol devices are realized, laying a foundation for subsequent accurate analysis. The intelligent analysis and prediction module performs deep analysis based on the normalized data, such as accurately predicting user charging demand and device operating status, so that the intelligent strategy formulation and execution module can formulate context-aware dynamic resource optimization scheduling strategies. This automated resource allocation based on accurate prediction effectively avoids resource waste and uneven allocation, thereby comprehensively improving the turnover rate and economic benefits of operation assets; 2. The application effectively enhances the reliability of the operation system and realizes proactive and forward-looking maintenance management. The intelligent analysis and prediction module can model the device operating data accessed, realize fine evaluation of device health status and early prediction of failure probability. Based on these prediction results, the intelligent strategy formulation and execution module can automatically trigger predictive maintenance and proactive health management strategies to arrange maintenance or take preventive measures before actual failure occurs. This change from passive response to active intervention significantly reduces unplanned downtime, extends device service life, and ensures the continuity and stability of operation services; 3. The operation management system of the application has the ability of continuous learning and self-adaptive evolution, which enables it to dynamically adapt to changing operation environments and business demands. The system not only records the execution process and results of operation strategies and feeds these data back to the intelligent analysis and prediction unit for evaluating strategy effectiveness and assisting model iteration optimization, but also supports bidirectional data and service interaction with external ecosystems through an open API interface management unit. This combined internal and external continuous learning mechanism enables the system's prediction model accuracy and strategy effectiveness to continuously improve, thereby ensuring that operation management decisions remain at a high level. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 This is the system architecture diagram of this application; Figure 2 It is a flow chart of the application method. DETAILED DESCRIPTION

[0019] Combined with attachment Figure 1 , further details of this application are given.

[0020] Embodiment: An operation management system, comprising: The data access and normalization module is used to uniformly access and normalize operation-related data from multiple heterogeneous data sources, including cross-brand and cross-protocol devices, to output normalized data; In this embodiment, the Data Access and Normalization Module serves as the fundamental data processing hub of the operations management system. Its core task is to uniformly and efficiently access operational data from diverse and heterogeneous data sources, perform a series of normalization processes on this raw data, and ultimately output high-quality, standardized data that can be directly used by subsequent intelligent analysis and prediction modules. The implementation of this module lays a solid data foundation for intelligent decision-making throughout the operations management system.

[0021] When performing its functions, the data access and normalization module is first committed to solving the problem of unified data access across brands and cross-protocol devices.

[0022] Specifically, when new charging and swapping equipment, such as charging piles, battery swapping cabinets or energy storage modules produced by different manufacturers, is connected to the system, this module is configured to automatically identify the communication protocol type of the device.

[0023] This automatic identification capability can preferably rely on a built-in, dynamically updateable protocol feature library, which pre-stores handshake information, message structure characteristics or device fingerprint information of multiple mainstream and manufacturer-specific communication protocols.

[0024] During the initial device access phase, this module attempts to detect communication with the device and determines the best matching communication protocol by comparing interaction information with records in the protocol signature library.

[0025] In some implementations, identification can also be assisted by reading a preset configuration file of the device or by having an operation and maintenance person pre-register the device model and its corresponding protocol information in the system.

[0026] Once the communication protocol of the device is successfully identified, the data access and normalization module dynamically loads or configures the corresponding protocol parser module and communication interface parameters.

[0027] The communication interface parameters can include the network address (such as IP address and port number) of the target device, serial communication parameters (such as baud rate, data bits, stop bits, and parity bits), and security authentication credentials required by specific protocols.

[0028] To ensure the confidentiality and integrity of the data transmission process, the module prioritizes the use of secure communication mechanisms when establishing a data transmission link with the device.

[0029] For example, for TCP / IP-based protocols, the use of TLS / SSL encrypted transmission layers is preferred to ensure that data is not eavesdropped or tampered with during network transmission.

[0030] For devices and scenarios that support two-way authentication, X.509 digital certificates can be used for mutual authentication between the device and the system platform.

[0031] For IoT message queue protocols such as MQTT, their own security features, such as username / password authentication, client ID authentication, and topic-based access control lists, are used to ensure communication security.

[0032] Through the secure data transmission link established in this way, the data access and normalization module can collect core data related to the operation of the device in real time or according to a pre-set collection period.

[0033] The collected device data is rich in content, typically including but not limited to: real-time operating state parameters of the device, such as phase voltages, currents, active power, reactive power, power factor, temperatures of key components inside the device (such as power module temperature, battery pack temperature), connection status of the charging gun or battery replacement interface, remaining power or health status estimates; as well as various events and alarm information generated by the device, such as start, stop, charging start, charging end, fault codes, over-temperature alarms, and over-current alarms; in addition, it may also include cumulative operating data and historical operation records of the device.

[0034] To further enrich the dimensions of operational data and improve the accuracy of subsequent analysis and prediction, the data access and normalization module is also configured with an open API interface module.

[0035] This module is used to integrate supplementary operational data from external third-party service systems and public data sources.

[0036] The design of the API interface preferably follows the current industry mainstream RESTful architecture style, uses lightweight JSON as the main data exchange format, and can provide standardized interface documents for easy integration and integration of third-party systems.

[0037] Through these open API interfaces, the system can actively or passively obtain data from the external environment, for example: obtaining real-time and future weather information (such as temperature, humidity, rainfall, wind speed, and light intensity) of the target operating area from national or commercial meteorological service platforms; obtaining real-time grid load status, time-of-use electricity price information, and renewable energy power generation forecasts from the power grid dispatching center or power trading platform; obtaining charging transaction flows, user payment preferences, user identity, and membership level information from the payment gateway or user authentication center; and obtaining the geographic location information of the site, the distribution of surrounding points of interest (POIs), real-time traffic conditions, and large-scale event arrangements from map service providers or traffic information platforms.

[0038] The interaction process with third-party APIs also focuses on security, and usually adopts authorization frameworks such as OAuth2.0 for access token management and permission control to ensure the security and compliance of data access.

[0039] After accessing raw operational data from multiple heterogeneous sources, the key responsibility of the data access and normalization module is to comprehensively integrate and standardize these raw data to output high-quality normalized data that can be directly used by the intelligent analysis and prediction modules.

[0040] This preprocessing process usually includes the following core steps: Data cleaning: Carefully handle the inevitable noise, outliers, duplicate data, missing values ​​and other problems in the original data.

[0041] For the random noise existing in the continuous time series data collected by sensors, smoothing and denoising can be performed by selecting methods such as moving average filter, exponential smoothing method, or Kalman filter if the model complexity allows, according to the data characteristics.

[0042] For outliers or mutation points in data records, they can be identified and marked by combining statistical methods (such as the 3σ principle, box plot method) or simple business rules, and can be eliminated, corrected or recorded as special events according to the circumstances.

[0043] For the problem of missing data, different filling strategies can be flexibly adopted according to the type of missing data, the missing proportion and the importance of the data. For example, the global mean, median or mode of the feature can be used for simple filling; or the characteristics of time series can be used to fill in the previous valid value or the next valid value; in more complex cases, the missing values ​​can be predicted and filled based on the values ​​of other related features by building a regression model or using machine learning algorithms such as K-nearest neighbor interpolation.

[0044] Data format unification and standardization: Due to the diverse formats of accessed data sources, for example, devices may report custom binary streams, XML messages, or proprietary format log files, while third-party APIs may return JSON or XML objects.

[0045] Therefore, this module needs to convert these different structures and different encoded data into a unified and easily processed structured data representation form within the system, such as converting to a JSON-like object or a flattened table structure.

[0046] In this process, the naming conventions of data fields, data types (such as integer, floating point, string, date and time) are unified.

[0047] In particular, for numerical feature data, in order to eliminate the influence of different dimensions or too large value range differences between different features on subsequent model training (for example, making some features dominant in distance calculation or gradient descent), standardization or normalization processing is required.

[0048] A commonly used standardization method is Z-score standardization, whose calculation formula is: ; where, is the original observation value of a certain feature, is the mean of the feature on the training data set, is the standard deviation of the feature on the training data set, is the standardized value. This method makes the processed data have zero mean and unit variance. Another commonly used method is Min-Max normalization, whose calculation formula is: where, is the original observation value of a certain feature, and are the minimum and maximum values of the feature on the training data set, is the normalized value, usually in the range of [0, 1] or [-1, 1]. The choice of method depends on the sensitivity of the subsequent analysis model to data distribution.

[0049] Data fusion and context construction: The core of this link is to effectively associate and integrate data from different data sources that describe different aspects of the same operating entity or event.

[0050] A key step is timestamp alignment, ensuring that all data with time attributes are based on a unified time reference and can be organized and analyzed in chronological order.

[0051] Further, different source data records are associated by sharing identifiers such as device ID, user ID, site ID, geographic location coordinates, transaction order number. For example, the detailed operating parameters of a specific charging pile at a certain moment (such as current, voltage, power factor, internal temperature) can be associated with the external environment data (such as ambient temperature, humidity) at that time, the real-time electricity price information at that time, the recent maintenance record of the charging pile, and the user information (such as user membership level, vehicle type) being served, to form a standardized data record containing rich context information and multiple dimensions.

[0052] The construction of such context information greatly enhances the interpretability of data and the depth of subsequent intelligent analysis.

[0053] Data storage: the standardized data formed after the above series of cleaning, conversion, standardization and fusion processing is finally loaded and stored in the unified data platform configured by the system.

[0054] The unified data platform, according to the actual business needs and data characteristics, can preferably be a combination of one or more database systems. For example, for high-frequency time series data reported by devices, a specially optimized time series database can be selected; for structured business-related data and configuration information, a relational database can be used; for unstructured or semi-structured data (such as log text, JSON document), a NoSQL database or a data lake architecture can be used for unified storage and management.

[0055] The standardized data in the unified data platform provides a clean, consistent and easy-to-use data source for subsequent deep analysis, model training and intelligent decision-making.

[0056] Through the above detailed implementation, the data access and standardization module of the present application can effectively overcome the challenges of scattered data sources, different formats and uneven quality in traditional operation management, and provides a solid and reliable data guarantee for realizing fine and intelligent operation management.

[0057] The intelligent analysis and prediction module is coupled to the data access and standardization module to receive standardized data and is configured to perform deep analysis and intelligent prediction modeling based on the standardized data to output analysis and prediction results. In this embodiment, the intelligent analysis and prediction module, as the core driving engine of the intelligent decision-making of the operation management system, its main responsibility is to receive and deeply utilize the standardized data output by the aforementioned data access and standardization module, to construct and execute precise analysis and prediction models by using advanced big data analysis techniques and artificial intelligence algorithms, and finally generate insightful analysis and prediction results for the subsequent intelligent strategy formulation and execution module to make efficient decisions and precise execution.

[0058] After receiving the normalized data provided by the data access and normalization module, the intelligent analysis and prediction module may first perform further exploratory analysis and feature engineering of the data.

[0059] This stage aims to gain a deeper understanding of data characteristics, explore potential correlations, and construct or select more representative input features for subsequent modeling processes.

[0060] For example, for time series data, it may be necessary to perform periodic analysis, trend decomposition, or construct lag features; for multi-source fusion data, it may be necessary to perform more detailed cross-domain feature intersection and combination.

[0061] This process ensures the quality and information content of the data input into the prediction model and is a key prerequisite for improving model performance.

[0062] Once the data is ready, the intelligent analysis and prediction module is dedicated to using the normalized data to perform core modeling tasks to generate forward-looking analysis and prediction results. Modeling tasks include at least one or more of the following key functions: First, equipment health assessment and failure probability prediction. This function aims to provide early warning of potential equipment failure risks through deep learning and pattern recognition of equipment operating data, thereby providing a decision-making basis for implementing predictive maintenance, reducing unplanned downtime, and ensuring operational continuity.

[0063] Specifically, the intelligent analysis and prediction module builds and applies equipment failure prediction models The model is based on a specific device At discrete time points The comprehensive feature vector of as input.

[0064] Eigenvector It is carefully designed to fully reflect the current and historical operating status of the equipment, and preferably includes data from one or more of the following dimensions: historical equipment operation data, such as the equipment's cumulative operating hours, historical fault occurrence times, fault type statistics, and historical maintenance records; real-time sensor data, such as current values, voltage values, temperatures of key internal components (such as power modules, transformers, and battery cell temperatures), spectral characteristics of vibration signals, and insulation resistance values ​​directly collected from the equipment; equipment operating environment data, such as the ambient temperature, humidity, altitude, and the presence of corrosive gases at the equipment's location; and current equipment operating condition data, such as the current output power level, load rate, charge and discharge rate, and recent start and stop frequency.

[0065] The selection of these features is based on the understanding of device failure mechanisms and data-driven feature importance analysis.

[0066] Based on the above input feature vector ,Fault prediction model output device In a preset time window in the future The probability of failure within The probability is determined according to the following formula: ; In this formula, the symbols are defined as follows: Represents a specific operational equipment entity, such as a charging pile, a specific module of a battery swap station, or an energy storage module; Represents the time span for failure probability prediction, such as the next 24 hours, the next week, or the next maintenance cycle; Represents the current or most recent discrete data sampling time point; That is, the aforementioned description device At the time point The eigenvector of the state; Represents a pre-trained mathematical model or algorithm entity for performing failure probability prediction; Representative and Failure Prediction Model A set of associated optimal parameters is determined during the model training phase by learning a large amount of historical equipment operation data and corresponding fault labels.

[0067] Fault prediction model Depending on the complexity of the actual application scenario and the characteristics of the data, a variety of machine learning or deep learning algorithms can be used. For example, when the data features are relatively clear and the failure mode is relatively simple, logistic regression models, support vector machines (SVMs), or ensemble learning methods such as random forests and gradient boosting decision trees can be used. When it is necessary to capture the complex dynamic characteristics of the device state evolving over time, deep learning models that can process time series data, such as recurrent neural networks (RNNs) and their variants such as long short-term memory networks (LSTMs) or gated recurrent modules (GRUs), can be used. These models effectively predict future failure risks by learning the differences between normal operating modes and prefault patterns in historical data.

[0068] Second, user behavior pattern recognition and charging demand prediction functions.

[0069] The implementation of this function aims to accurately predict the demand for charging services at a specific time and location in the future by analyzing users' historical behavior data and related factors affecting demand, thereby providing key input for optimizing site layout, dynamically adjusting service prices, rationally dispatching charging resources, and improving users' charging experience.

[0070] Specifically, the intelligent analysis and prediction module builds and applies the charging demand prediction model The model is used to predict the At a specific point in the future Expected charging demand .

[0071] In order to make accurate predictions, the model takes into account a variety of factors that affect charging demand. Its input features preferably include one or more of the following aspects: location As of the current time Historical demand data series ,The sequence can contain historical observations such as the actual number of charging times, charging power or charging pile occupancy time at the same geographical location and the same time period over the past several periods (e.g., days, weeks, or even months) to capture the baseline level and periodicity of demand; Calendar Features This feature is used to characterize the impact of time factors on demand, such as the day of the week (the demand pattern on weekdays and weekends is usually different), whether it is a statutory holiday, seasonal factors, and whether there are special social activities (such as large conferences, sports events), etc.; as well as predict the target location At a future target time Weather forecast data For example, the expected temperature (too high or too low may affect travel and charging behavior), precipitation probability, wind level, etc. These environmental factors may also have a significant impact on users' travel decisions and charging needs.

[0072] Based on the above input features, the charging demand prediction model Output predicted charging demand The demand forecast follows the following formula: ; In this formula, the symbols are defined as follows: Represents a specific geographic location, such as a specific charging station, a city area, or a highway service area; Represents the time advance of demand forecast, that is, from the current time point The duration to the predicted target time; is the current discrete data sampling time point; i.e. the location the time of day a time series of historical demand data; i.e. the time a calendar-related feature; i.e. the location weather forecast data at the future time ; representing a pre-trained mathematical model or algorithmic entity for performing the charging demand prediction; representing a set of optimal parameters associated with the demand prediction model , which are also determined through learning a large amount of historical demand data and its influencing factors during the model training phase.

[0073] The charging demand prediction model may be selected from models suitable for handling time series prediction and multivariate regression problems. For example, classical statistical time series models such as ARIMA, or Facebook’s open-source Prophet model; or machine learning-based regression models such as Support Vector Regression (SVR), XGBoost, etc. For demand sequences with complex non-linear relationships and long-time dependencies, preferably, deep learning models such as LSTM networks can be adopted to capture more deep-level demand evolution patterns.

[0074] In addition to the above-mentioned core prediction modeling function, the intelligent analysis and prediction module can also be configured to perform other types of analysis tasks, such as operational resource utilization efficiency analysis, which identifies bottlenecks and optimization spaces of resource allocation by calculating key performance indicators such as average device utilization rate, average user waiting time, service turnover rate, etc. at each site; and operational cost-benefit analysis, which evaluates the economic efficiency of different operation strategies in combination with factors such as energy prices, maintenance costs, labor costs, etc.

[0075] After completing the above-mentioned various analysis and prediction modeling tasks, the intelligent analysis and prediction module is further configured to refine business insights with direct guiding significance for business operation from the generated analysis and prediction results (e.g. specific numerical values of and , intermediate outputs of the model, and other analysis reports).

[0076] These business insights are the result of deep processing and knowledge distillation of raw data and model outputs, such as: identifying the TopN list of devices with the highest risk of failure in the next week, along with their specific risk levels and possible failure components; profiling the charging behavior preferences of different user groups (such as private car owners, taxi drivers, logistics fleets) in different time periods and regions; revealing the actual pull effect of specific marketing activities on charging demand; or pinpointing specific links or sites that cause low operational efficiency.

[0077] To effectively manage and utilize these valuable analysis results and knowledge deposits, the intelligent analysis and prediction module will structure and store the refined business insights, along with the trained and validated prediction models themselves (including their specific model structure definitions, such as neural network levels and node information), and the optimal parameters of the models (such as the aforementioned and ), as well as model-related metadata (such as model version, training data set description, scope of application, performance indicators), in a structured manner in a dynamic knowledge base configured by the system.

[0078] The dynamic knowledge base is preferably built using technologies that support complex relationship storage and efficient retrieval, such as graph databases or data warehouses with semantic layers.

[0079] The construction of this dynamic knowledge base has a dual purpose: first, it provides direct and operational decision-making basis and intelligent input for the subsequent intelligent strategy formulation and execution module; second, it also provides a benchmark and feedback loop interface for the continuous learning and model iterative optimization of the intelligent analysis and prediction module itself, enabling the system to continuously evolve with the accumulation of new data and changes in the operating environment.

[0080] In the execution of the above complex analysis and modeling tasks, especially in the processing of massive historical data and training of deep learning models, the intelligent analysis and prediction module can preferably rely on a distributed computing framework to improve the efficiency and scalability of data processing and model training, but this is not a limitation of the specific hardware implementation of the system.

[0081] Finally, the intelligent analysis and prediction module will integrate and package all the analysis and prediction results it produces, including specific prediction values, probability assessments, business insight reports, and updates to the knowledge base, as its core output, and pass it to the intelligent strategy formulation and execution module in the operation management system, thereby driving subsequent intelligent operation decisions and automated control processes.

[0082] Through the above implementation, the intelligent analysis and prediction module of the present invention can transform the original operational data into highly valuable intelligent information and prediction capabilities, providing a powerful intelligent core for accurate decision-making and efficient operation of the entire operation management system.

[0083] The intelligent strategy formulation and execution module is coupled with the intelligent analysis and prediction module to receive analysis and prediction results, formulate context-aware operation strategies based on the analysis and prediction results, and drive their automated execution.

[0084] In this embodiment, the intelligent strategy formulation and execution module, as the key link for taking on the results of intelligent analysis and prediction and transforming them into actual operational actions, has the core function of receiving and interpreting the various analysis and prediction results output by the aforementioned intelligent analysis and prediction module, including but not limited to equipment failure probability, user charging demand forecasts, and business insights and model parameters extracted from the dynamic knowledge base. Based on these intelligent inputs, this module is committed to formulating forward-looking operational strategies that are highly aligned with the current operational context and further driving the automated or semi-automated execution of these strategies, thereby achieving closed-loop optimization and intelligent response of operational management activities.

[0085] After receiving the analysis and prediction results from the intelligent analysis and prediction module, the intelligent strategy formulation and execution module first parses and contextualizes this information.

[0086] For example, when a specific device is received In the future time window Failure probability within When the risk is detected, this module will compare it with the preset risk threshold.

[0087] When receiving a specific location In the future Charging demand When the device is in use, this module will conduct a comprehensive assessment based on the current resource availability, device health status and other information at the location.

[0088] This process is designed to ensure that the strategies subsequently developed can accurately address current or upcoming problems and maximize the expected operational benefits.

[0089] Based on the above understanding and assessment, the intelligent strategy formulation and execution module is further configured to formulate and drive the execution of at least the following core operational strategies: First, predictive maintenance and proactive health management strategies based on equipment failure probability prediction results.

[0090] This strategy aims to transform traditional passive and scheduled maintenance into active and predictive maintenance based on actual health status and future risk prediction, to improve equipment reliability, reduce operation and maintenance costs and unplanned downtime losses.

[0091] When the failure probability prediction result output by the intelligent analysis and prediction module of a certain operating equipment exceeds the pre-set maintenance trigger threshold (which can be dynamically adjusted according to factors such as equipment type, importance, and maintenance cost, etc.), the intelligent strategy formulation and execution module will automatically trigger the predictive maintenance process.

[0092] This process preferably includes: automatically generating a detailed predictive maintenance suggestion or electronic work order, which clearly indicates the unique identifier of the early warning equipment, its current location, the predicted potential failure type or faulty component, the recommended optimal maintenance time window (such as the period before the predicted failure occurs and has the least impact on operation), as well as the list of spare parts and accessories that may be needed and the recommended maintenance operation procedures according to the failure type.

[0093] This work order can then be automatically pushed to the enterprise's internal operation and maintenance management system, or directly notified to the designated operation and maintenance team or technical personnel through mobile applications, SMS, email, etc.

[0094] In some scenarios where remote control is allowed, if the predicted failure risk is still in the early stage or in order to gain more time for maintenance preparation, this module can also try to issue specific remote control instructions to the target equipment through linkage with the device control interface. These instructions may include: temporarily adjusting the operating parameters of the equipment, such as appropriately reducing its rated output power, limiting the use of some non-core functions, or guiding the equipment to enter a more conservative operating mode; or starting the built-in self-checking program or diagnostic sequence of the equipment to obtain more detailed fault location information.

[0095] Through such active intervention, the risk can be effectively avoided or the fault-free operation time of the equipment can be significantly prolonged before the actual failure occurs.

[0096] Secondly, dynamic resource optimization scheduling and collaborative strategy based on real-time and predicted operating situation and charging demand prediction results.

[0097] The core goal of this strategy is to maximize the overall utility of the operation system, such as improving energy utilization efficiency, increasing operation revenue, or balancing grid load, through intelligent resource allocation and scheduling, while meeting user demand and ensuring service quality.

[0098] When formulating such strategies, the intelligent strategy formulation and execution module will comprehensively consider the multi-dimensional information provided by the intelligent analysis and prediction module, mainly including: the specific geographical location​ At a specific point in the future Forecasted charging demand (This is the output of the aforementioned intelligent analysis and prediction module Application in specific scheduling scenarios); various operating equipment At the time point Real-time or predicted health status (For example, the aforementioned failure probability 1- The system converts the data into a health score. Devices with lower scores may be given lower priority or have their service capacity restricted in scheduling. The system also provides real-time or future energy cost information obtained from external systems (such as power grid companies or energy trading platforms). (e.g. time-of-use electricity prices, demand response compensation prices, etc.).

[0099] Based on these input information and combined with the preset operation optimization objectives (for example, maximizing the total revenue of charging services, minimizing the average waiting time of users, minimizing the total operation energy consumption cost, or maximizing the degree of friendly interaction with the grid), this module determines the optimal resource scheduling solution by solving a carefully constructed optimization problem. The goal of this optimization problem is usually to maximize a function representing the total utility of the system. , its mathematical expression is: In this objective function, the physical meaning of each symbol is as follows: Represents the entire resource scheduling planning period, such as the next hour, day, or a specific operating period. Represents the set of all geographic locations or sites that provide services in the system. Representative Set A specific geographic location or site in a. Representative in position A collection of currently available operational equipment (e.g., charging stations, battery swap stations). Representative Set A specific available device in . Represents the scheduling cycle A discrete scheduling time point or time slot within a schedule. Represents at a point in time A collection of users with clear service requests (such as charging requests, battery replacement requests). Representative Set A specific user in the . is the core decision variable, which represents a specific resource allocation behavior, such as at time point Decide to use The charging request is assigned to the location Specific devices Performing the service; this variable can be binary (allocated or not allocated) or continuous (e.g. the size of the power allocated). is the utility function generated by a single allocation action, the specific form of which depends on the operational objectives, which maps the decision variable and the context conditions (predicted demand), (equipment health status), and (energy cost) to a utility value (e.g. expected revenue, cost savings, or user satisfaction contribution).

[0100] The solution of this optimization problem, which usually also needs to satisfy a series of practical operational constraints, such as: the maximum output power limit of a single device, the total access capacity limit of a single charging station, the upper limit of the maximum queuing waiting time acceptable by users, the maximum access load limit allowed by the power grid for a single station or region, and the charge-discharge rate and SOC safety range limit of the battery energy storage system, etc.

[0101] For such a combined optimization problem that may have high complexity, according to the specific size of the problem, the real-time requirement and the available computing resources, this module can preferably use a variety of mathematical programming or artificial intelligence optimization algorithms for solution. For example, for small and medium-sized problems, a mixed integer programming solver can be used; for large-scale or fast response scenarios, efficient heuristic algorithms (such as genetic algorithms, simulated annealing algorithms, tabu search) or meta-heuristic algorithms can be used; in recent years, methods based on reinforcement learning have also shown potential, by training agents to learn the optimal dynamic scheduling strategy in a simulated operating environment.

[0102] The specific scheduling and coordination strategy obtained after solving the optimization problem has a rich variety of forms, such as: dynamically adjusting the charging service prices of different charging stations or at different times of the day to economically guide users to space-time shift and cut peaks and fill valleys; through the user mobile application (APP) or the vehicle-mounted navigation system, real-time push the optimal charging station navigation suggestions and estimated waiting time based on the current traffic conditions, station queuing conditions and personal preferences to users; intelligent queuing management and dynamic allocation of charging power for charging tasks accessed in the station, to prioritize high-value users or emergency needs; coordinated control of the battery energy storage system deployed in the station, charging and storing energy at low electricity prices, discharging to supply users or returning power to the grid at peak electricity prices or when the grid load is tight, participating in the demand response plan and auxiliary service market of the grid.

[0103] Thirdly, intelligent early warning and emergency response strategies based on real-time monitoring and abnormal detection results of operational related data.

[0104] This strategy aims to timely detect abnormal events or potential risks deviating from the normal operation trajectory through continuous monitoring and intelligent analysis of key state parameters of the operating system, and quickly initiate predefined emergency plans to minimize the negative impact of abnormal events and ensure the safety of operation and stability of the system.

[0105] The intelligent strategy formulation and execution module continuously receives and analyzes key operating parameters uploaded in real time by the data access and normalization module, which may include sudden or continuous over-limit of current, voltage, and temperature at the individual device level; abnormal fluctuations in total power consumption, communication network connection interruptions at the site level; and sudden increase in user service request concurrency, payment system failure alarms at the system level.

[0106] In order to effectively identify abnormalities from these massive data, this module can preferably use rules based on statistical process control (such as setting upper and lower limits of control charts), or apply more advanced machine learning-based anomaly detection models such as Isolation Forest, One-Class SVM, or reconstruction error detection based on Autoencoder. These models can learn the "normal" pattern from historical normal operation data, and thus label and warn new data points that significantly deviate from this pattern.

[0107] Once an abnormal event is detected and confirmed to have triggered a warning condition (such as abnormal score exceeding threshold or continuous multiple periods of abnormal signals), the intelligent strategy formulation and execution module will automatically or with the assistance of human confirmation, based on the rules in the emergency plan knowledge base pre-configured in the system and combined with real-time analysis results of the specific abnormal situation, decisively execute the corresponding emergency response strategy.

[0108] These emergency response strategies may include: at the device level, automatically attempting to isolate devices that have failed or pose serious safety risks, such as disconnecting them from the main system to prevent the spread of failure or impact on other devices and users; at the operation management level, immediately sending detailed alarm notifications (including abnormal type, location, possible cause, and recommended handling measures) and fault location information to relevant operation and maintenance personnel or emergency command center; at the resource scheduling level, if a site's service capacity decreases due to failure, dynamically adjusting the resource scheduling scheme of other normal operation sites in the surrounding area, such as temporarily increasing their service capacity or guiding users to these sites; at the user service level, through APP push, SMS, site display screen, etc., timely publish relevant safety tips, service interruption or adjustment information to users who may be affected, and provide alternative solutions.

[0109] In order to ensure the continuous optimization and self-adaptability of the entire operation management system, the intelligent strategy formulation and execution module also undertakes a crucial follow-up task: detailed recording and feedback on the entire life cycle of all formulated and executed operation strategies.

[0110] Specifically, this module will comprehensively and structurally record the detailed information of the above-mentioned various strategies (regardless of the issuance of predictive maintenance instructions, the implementation of resource scheduling schemes, or the initiation of emergency response measures). These recorded contents preferably include: the type and unique identification of the strategy; the timestamp of the strategy formulation and the input information (such as the prediction results at that time, market conditions) relied upon; the target object of the strategy (such as specific equipment, site, user group); the specific time of strategy execution, the content of control instructions issued or the adjusted parameter values; and the preliminary execution results and system state changes (such as whether the equipment failure after maintenance is eliminated, whether the site utilization rate is improved after scheduling, whether user complaints are reduced after emergency response) observable within a reasonable period of time after strategy execution.

[0111] These recorded strategy execution processes and result data will serve as valuable experience data and feedback information, which will be transmitted back to the data access and standardization module of the system and, after necessary processing, finally integrated into the aforementioned unified data platform. This part of feedback data is crucial for the closed-loop learning of the entire operation management system, as they are not only used to objectively evaluate the actual effect and return on investment of previously formulated strategies, but also to verify and calibrate the prediction accuracy of various models in the intelligent analysis and prediction module, and provide indispensable data support for subsequent system model iteration, strategy library update, and dynamic knowledge base enrichment.

[0112] Through the above-mentioned refined strategy formulation logic and automated execution mechanism, supplemented by comprehensive process recording and effect feedback, the intelligent strategy formulation and execution module of the present application can effectively convert the results of intelligent analysis into actual actions that improve operational efficiency, reduce operational risk, and optimize user experience, constituting the key execution hub of the complete intelligent closed loop of the entire operation management system from perception, analysis, decision-making to action.

[0113] In combination with the accompanying drawings, Figure 2 another embodiment of the present application provides an operation management method, comprising the following steps: S1, through the data access and standardization process, unified access and standardization of operation related data from multiple source heterogeneous data sources including cross-brand, cross-protocol equipment to output standardized data; S2, through the intelligent analysis and prediction process, receiving standardized data, and using the unified data platform, based on the standardized data, deep analysis and intelligent prediction modeling to generate business insights and prediction results; S3, receiving business insight and prediction results through an intelligent strategy formulation and execution process, and formulating context-aware operation strategies and driving automatic execution thereof based on the business insight and prediction results.

[0114] The method of the embodiment can be used to execute the above-mentioned system embodiment, and has similar principles and technical effects, which will not be described here again.

[0115] The embodiments of the specific implementation are the preferred embodiments of the present application, and are not intended to limit the protection scope of the present application, wherein the same parts are denoted by the same reference numerals. Therefore, equivalent changes made according to the structure, shape, principle of the present application should be covered within the protection scope of the present application.

Claims

1. An operation management system, characterized in that: include: The data access and normalization module is used to uniformly access and normalize operation-related data from multiple heterogeneous data sources, including cross-brand and cross-protocol devices, to output normalized data; an intelligent analysis and prediction module, coupled to the data access and normalization module to receive the normalized data, and configured to perform in-depth analysis and intelligent prediction modeling based on the normalized data to output analysis and prediction results; An intelligent strategy formulation and execution module is coupled to the intelligent analysis and prediction module to receive the analysis and prediction results, formulate a context-aware operation strategy based on the analysis and prediction results, and drive its automatic execution.

2. An operation management system according to claim 1, characterized in that: When the data access and normalization module performs unified access to operation-related data, it further includes: Automatically identify the communication protocol of the connected cross-brand, cross-protocol device and establish a secure data transmission link to collect device operation data; And through open API interfaces, third-party services and external environment data sources are integrated to obtain supplementary operation-related data.

3. An operation management system according to claim 1, characterized in that: When the data access and normalization module performs normalization of operation-related data, it further includes: The multi-source heterogeneous operation-related data accessed through the data access and normalization module are subjected to data cleaning, data format unification, and data fusion processing to form the normalized data containing contextual information for use by the intelligent analysis and prediction module.

4. An operation management system according to claim 1, characterized in that: The intelligent analysis and prediction module further includes: The received normalized data is modeled using big data analysis technology and artificial intelligence algorithms to generate the analysis and prediction results, wherein the modeling includes at least one of the following functions: Equipment health status assessment and failure probability prediction, including equipment In the future time window The probability of failure within Determined by the following formula: ； in, For specific equipment; is the time window for prediction; is a discrete time point; For devices At the time point A feature vector comprising at least one of historical equipment operation data, real-time sensor data, operating environment data, and operating condition data; is a pre-trained fault prediction model; are parameters of the fault prediction model; Or, user behavior pattern recognition and charging demand prediction.

5. An operation management system according to claim 1, characterized in that: When the modeling includes user behavior pattern recognition and charging demand prediction functions, the charging demand prediction function further includes predicting the charging demand at a specific location by the following formula: and future time Charging demand : ； in, For a specific geographical location; The lead time for demand forecast; is a discrete time point; For location Deadline Historical demand data series; For time Calendar features; For location In the future Weather forecast data; A pre-trained demand forecasting model; are parameters of the demand forecasting model; Furthermore, the intelligent analysis and prediction module is further configured to extract business insights from the analysis and prediction results generated thereby, and to store the business insights and the trained models in a structured manner in a dynamic knowledge base.

6. An operation management system according to claim 1, characterized in that: The intelligent strategy formulation and execution module is further configured to formulate and drive the execution of an operation strategy based on the analysis and prediction results, wherein the operation strategy includes at least one of the following: Predictive maintenance and proactive health management strategies based on the equipment failure probability prediction results; Dynamic resource optimization scheduling and coordination strategy, which maximizes the total utility of the system by solving To determine the optimization problem of the target, the It is expressed by the following objective function: ； in, is the scheduling period; is a collection of charging locations; for a specific location in the For location The set of available devices; for A specific device in for a specific point in time within For in time A collection of users who have charging requests; for A specific user in is the decision variable, indicating the The user The charging request is assigned to the location Equipment behavior; is the utility function for a single allocation; For location In time Forecasted charging demand; For equipment In time health status; Alternatively, intelligent early warning and emergency response strategies based on real-time monitoring of operational-related data and anomaly detection results.

7. An operation management system according to claim 1, characterized in that: The intelligent strategy formulation and execution module further includes: The execution process and results of the operation strategy are recorded, and the data of the execution process and results are fed back to the intelligent analysis and prediction module for evaluating the effectiveness of the strategy and assisting the iteration of the model in the intelligent analysis and prediction module.

8. An operation management system according to claim 1, characterized in that: The system further comprises: The open API interface management module is configured to support two-way data and service capability interaction with the external ecosystem. The interaction includes allowing the data access and normalization module to integrate external data through the API, and allowing the processed valuable aggregated information in the analysis and prediction results output by the intelligent analysis and prediction module to be shared with authorized ecological partners through the open API interface.

9. An operation management system according to claim 1, characterized in that: The system further includes: a capability of continuous learning and adaptive evolution, which is achieved by: Conducting continuous evaluation of overall operational performance and the accuracy of the analysis and forecasting results output by the intelligent analysis and forecasting module; And based on the evaluation results, the model parameters in the intelligent analysis and prediction module and / or the policy logic in the intelligent policy formulation and execution module are iteratively optimized.

10. An operation management method, according to an operation management system according to any one of claims 1 to 9, characterized in that: The following steps are involved: S1. Through the data access and normalization process, unified access and normalization of operation-related data from multiple heterogeneous data sources, including cross-brand and cross-protocol devices, to output normalized data; S2. Receive the normalized data through an intelligent analysis and prediction process, and utilize a unified data platform to perform in-depth analysis and intelligent predictive modeling based on the normalized data to generate business insights and prediction results; S3. Receive the business insights and prediction results through the intelligent strategy formulation and execution process, and formulate context-aware operational strategies based on the business insights and prediction results and drive their automated execution.

Images