Energy storage power station equipment state monitoring and maintenance decision-making system

Abstract

The invention relates to the technical field of energy storage equipment monitoring, and discloses an energy storage power station equipment state monitoring and maintenance decision system. The system comprises an equipment topology overview construction module, a multi-source monitoring module and a health state compensation module. Wherein the equipment topology overview construction module generates an equipment topology overview containing electrical connection parameters and working condition parameters based on a physical connection relationship of an energy storage power station; the multi-source monitoring module collects equipment operation state data streams from the distributed monitoring nodes and obtains environment disturbance data in real time; and the health state compensation module performs missing node compensation on the equipment operation state data stream in the equipment state fusion monitoring mode to obtain a health state index of the energy storage equipment. According to the system, equipment topology information and multi-source data are integrated, accurate assessment of the equipment health state is realized, support is provided for maintenance decision of energy storage power station equipment, and equipment operation and maintenance efficiency can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of energy storage equipment monitoring, and in particular to an energy storage power station equipment status monitoring and maintenance decision-making system. Background Art

[0002] With the rapid development of the new energy industry, energy storage power stations, as a crucial vehicle for energy storage and dispatch, are attracting increasing attention for their operational stability and reliability. Energy storage power stations typically consist of numerous energy storage battery packs, converters, combiner cabinets, and other equipment. These devices are interconnected through complex electrical connections to form an integrated whole. The coordinated operation of these devices directly impacts the overall efficiency of the power station.

[0003] Currently, equipment status monitoring in energy storage power plants often uses a decentralized management model. Monitoring data from different devices is often collected and processed by independent systems, lacking unified topological correlation analysis. This model separates the electrical connections between devices from their operating parameters, making it difficult to assess the overall operational status of the device cluster. For example, when a battery pack experiences a voltage anomaly, the inability to quickly locate its connections to surrounding devices often results in a time-consuming and delayed investigation of the source of the fault, delaying repairs.

[0004] Existing monitoring systems have limitations in data collection. Distributed monitoring nodes may lose data due to communication interruptions, sensor failures, and other factors, affecting accurate assessment of device status. Environmental factors such as temperature, humidity, and vibration significantly impact the operating status of energy storage devices, but existing systems are inadequate in collecting and integrating data on these environmental disturbances, making it difficult to fully assess the health of the equipment.

[0005] When it comes to maintenance decisions, due to a lack of comprehensive analysis of equipment topology, operating status data, and environmental factors, maintenance plans often rely on empirical judgment. This can lead to problems such as untimely, excessive, or insufficient maintenance. This not only shortens the lifespan of the equipment, but can also increase operating costs and even pose safety risks. Therefore, building a system that can integrate equipment topology information and multi-source monitoring data to accurately assess health status and make scientific maintenance decisions has become a critical requirement for energy storage power station operations and management. Summary of the Invention

[0006] The purpose of the present invention is to provide an energy storage power station equipment status monitoring and maintenance decision-making system to solve the problems raised in the above background technology.

[0007] To achieve the above objectives, the present invention provides an energy storage power station equipment status monitoring and maintenance decision-making system, the system comprising: A device topology overview building module is used to generate a device topology overview based on the physical connection relationship of the energy storage power station. The device topology overview includes electrical connection parameters and operating condition parameters of the energy storage device cluster; Multi-source monitoring module, used to collect equipment operation status data streams from distributed monitoring nodes and obtain environmental disturbance data of energy storage equipment in real time; The health status compensation module is used to compensate for missing nodes in the device operation status data stream in the device status fusion monitoring mode to obtain the health status indicator of the energy storage device.

[0008] Preferably, the health status compensation module is specifically used to: determining a state transition path of the energy storage device according to the device topology overview; Probabilistically compensating for missing state nodes in the device operation state data stream to generate a complete state event sequence; Performing feature aggregation on the state event sequence to obtain state entropy value information of the energy storage device; The health status index of the energy storage device is determined based on the state entropy value information.

[0009] Preferably, the system further comprises: The monitoring cycle decision module is used to determine the monitoring interaction cycle of the energy storage device during operation according to the health status indicator, specifically including: Constructing a degradation rate field model of the energy storage device according to the health status indicator; Outputting a monitoring cooperation boundary for dynamic monitoring from the degradation rate field model; The monitoring interaction period of the energy storage device during operation is determined by the monitoring cooperation boundary.

[0010] Preferably, the system further comprises: The risk feature analysis module is used to obtain the fault propagation path information of the energy storage device cluster, perform risk feature analysis on the fault propagation path information, and obtain node fault coupling data of the energy storage device during operation; specifically, it includes: Building a spatiotemporal correlation model between the dynamic behavior of device nodes and historical fault events based on the fault propagation path information; Outputting the fault coupling strength of the energy storage device during operation through the spatiotemporal correlation model; The fault coupling strength is used to determine the fault impact characteristics of each device node in the device cluster.

[0011] Preferably, the system further comprises: The risk situation determination module is used to determine the risk situation level during the operation of the equipment based on the node fault coupling data and the environmental disturbance data; specifically includes: determining fault correlation constraints during equipment operation according to the node fault coupling data; Determining the abnormal disturbance granularity during the operation of the equipment through the environmental disturbance data; A risk situation level is determined according to the fault association constraint and the abnormal disturbance granularity.

[0012] Preferably, the system further comprises: The maintenance blind spot warning module is used to warn of dynamic maintenance blind spots during equipment operation and maintenance based on the monitoring interaction cycle and the risk situation level; specifically, it includes: determining a resilience monitoring threshold of the maintenance process according to the monitoring interaction cycle; determining a blind spot exposure coefficient in a dynamic maintenance blind spot according to the risk situation level; The elastic monitoring threshold and the blind spot exposure coefficient are strategically matched to generate a dynamic maintenance blind spot warning map.

[0013] Preferably, the multi-source monitoring module specifically includes: An encryption transmission submodule, used to create an anti-interference monitoring channel and encrypt and transmit the device operation status data stream through a quantum key distribution channel and an encryption algorithm channel; The distributed processing submodule is used to construct N distributed monitoring nodes and adaptively generate N node data relay strategies based on the real-time network status and interference intensity of the N distributed monitoring nodes.

[0014] Preferably, the system further comprises: The deception device control module is used to set up the device status deception device and modulate and emit multi-source deception signals based on the device's real location information; specifically, it includes: Deploy a thermodynamic characteristic simulation unit and a radio frequency signal generator on the equipment status deception device; generating decoy signal modulation parameters according to the device position offset; The device location information is modulated by the decoy signal modulation parameter.

[0015] Preferably, the system further comprises: The maintenance strategy decision module is used to calculate the remaining time between the current time and the maintenance window period and determine the maintenance strategy based on the remaining time. Specifically, it includes: Retrieve the first relevant conversion relationship between the remaining time and the maintenance period; Determine the number of failure modes of the energy storage device cluster; Retrieving a second correlation conversion relationship between the number of fault mode types and the period adjustment amount; A dynamic maintenance strategy is output according to the first correlation conversion relationship and the second correlation conversion relationship.

[0016] Preferably, the system further comprises: A maintenance decision output module, configured to perform a three-dimensional overlay of the dynamic maintenance blind spot warning map and the dynamic maintenance strategy to generate a dynamically rendered maintenance decision map; The maintenance decision map is spatially mapped through the equipment topology overview building module.

[0017] Compared with the prior art, the present invention has the following beneficial effects: The device topology overview module generates a device topology overview based on the physical connections of the energy storage power station, including electrical connection parameters and operating condition parameters. This allows operators to intuitively understand the interconnectedness of the entire power station equipment and clearly understand the position and role of each device in the overall system. This global perspective helps quickly trace potentially affected associated devices when equipment anomalies occur, reducing the scope and time of troubleshooting.

[0018] The multi-source monitoring module collects device operating status data streams from distributed monitoring nodes and acquires environmental disturbance data in real time, achieving comprehensive awareness of device operating status. The operating status data stream reflects the device's real-time operating conditions, such as changes in parameters like voltage, current, and power, while the environmental disturbance data complements the impact of external factors on the device. The combination of the two provides a more comprehensive and robust view of device status, avoiding biased judgments caused by a single data source.

[0019] In the fusion monitoring mode for device status, the health status compensation module compensates for missing nodes in the device operating status data stream, effectively resolving the issue of inaccurate assessments caused by missing data. In actual monitoring, data loss is difficult to completely avoid. The compensation mechanism fills in data gaps by making reasonable inferences based on existing data and device topology, thereby obtaining a more accurate health status indicator for the energy storage device. This indicator provides a reliable basis for maintenance decisions, enabling more targeted maintenance work and formulating appropriate maintenance strategies based on the actual health of the device, avoiding blind repairs. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a timing diagram of the energy storage power station equipment status monitoring and maintenance decision-making system according to the present invention; Figure 2 This is a flowchart of the working principle of the health status compensation module; Figure 3 This is a flowchart of the working principle of the risk feature analysis module; Figure 4 This is a flow chart of the working principle of the maintenance blind spot warning module; Figure 5 Flowchart showing the working principle of the maintenance strategy decision module. DETAILED DESCRIPTION

[0021] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0022] See also Figure 1 The present invention provides an energy storage power station equipment status monitoring and maintenance decision-making system, the system comprising: The physical connection model of the energy storage power station is established through the device topology overview module. This module analyzes the electrical wiring diagram and mechanical connection relationships of the energy storage device cluster to generate a topological network that includes device cascade parameters and operating conditions. The multi-source monitoring module is deployed at key monitoring points of the energy storage equipment. It uses a distributed data acquisition architecture to obtain real-time voltage, current, and temperature operating parameters of devices such as battery packs, converters, and transformers, and simultaneously collects disturbance data such as ambient temperature, humidity, and vibration. The health status compensation module establishes a state transition model based on a Markov chain. When monitoring data is missing, data compensation is performed using the state transition probabilities of adjacent nodes, ultimately outputting a health status indicator that contains device degradation characteristics.

[0023] Example 1: See Figure 2 The device topology overview module first connects to the energy storage plant's SCADA system and parses its stored electrical wiring diagrams, equipment layout diagrams, and operating parameter database. This module uses graph theory modeling to transform the energy storage plant's physical connections into a weighted directed graph structure, where nodes represent key equipment such as battery modules, converters, and transformers, and edges represent electrical connections or mechanical couplings. Each node stores the device's rated parameters, real-time operating data, and historical maintenance records. These real-time operating data and historical maintenance records together constitute the specific data source for operating parameters. Taking battery cluster node B01-01 in an energy storage power station as an example, its operating parameters are as follows: ① Real-time operating parameters: current charge and discharge power 120kW, voltage balance 98.5%, harmonic distortion 1.1%, bus voltage 380V; ② Historical load parameters: cumulative operating time 7500h, number of charge and discharge cycles 850, single maximum load duration 50min, and maximum temperature frequency 2 times per week; ③ Operating mode parameters: current mode is grid-connected discharge, switching frequency 3 times per 24h, and grid-connected mode accounts for 92%. Edge weights reflect the energy transmission efficiency or mechanical stress transfer coefficient between devices. The topology map generation process uses a hierarchical modeling strategy, first constructing battery cluster-level connectivity relationships, then gradually aggregating them upward to the converter and grid access point, ultimately forming a complete power station-level topology network. This network supports dynamic updates. When devices are commissioned or decommissioned or the connection mode changes, the system automatically adjusts the topology and recalculates associated parameters.

[0024] During operation, the health status compensation module receives real-time device status data streams from the multi-source monitoring module. When the voltage or temperature data of a battery module is missing, the module first queries the topology map to determine the specific location of the module in the battery string and the connection method of its adjacent modules. Subsequently, the module starts the state compensation algorithm based on the hidden Markov model, and uses the historical state transition probabilities of adjacent modules to infer the most likely value of the missing node. The calculation process comprehensively considers the equipment degradation trend and environmental disturbance factors. For example, when the ambient temperature rises suddenly, the algorithm will increase the probability weight of the sudden change of the state of the adjacent module. After the compensation is completed, the module will generate a complete state event sequence with a timestamp, where each event records the operating parameters and compensation mark of the device at a specific moment.

[0025] The feature aggregation stage of the state event sequence utilizes a multi-scale analysis approach. The module first segments the sequence into sliding windows, with the window length adaptively adjusted based on the device type. For example, battery modules use a 5-minute window, while converters use a 1-minute window. The data within each window undergoes a wavelet transform to extract time-frequency features, and its energy entropy and permutation entropy are calculated as local health indicators. The module then applies time-series modeling to the window features using a long-short-term memory network to capture both the long-term degradation patterns and short-term fluctuations of the device state. The final output, the state entropy value, contains the device's health score and its confidence interval at the current moment. This score is nonlinearly correlated with the device's actual degradation level and reflects potential failure risks.

[0026] The monitoring cycle decision module adjusts the monitoring strategy based on the dynamic changes in health status indicators. The module maintains a degradation rate field model internally, which estimates the attenuation gradient of the equipment health in real time through a Kalman filter. When the entropy value change rate exceeds the preset threshold, the module automatically shortens the monitoring interaction cycle, for example, from the default 10 minutes to 2 minutes. The adjustment strategy uses fuzzy control logic, taking into account factors such as the importance of the equipment, historical failure rate, and current risk level. For example, for equipment located on the critical path of fault propagation, even if the health change rate does not reach the threshold, the module will actively increase the monitoring frequency. The dynamically adjusted monitoring cycle is distributed to each distributed monitoring node through a message queue to ensure the synchronization of data collection and status assessment.

[0027] The monitoring cycle decision module is used to determine the monitoring interaction cycle of the energy storage device during operation according to the health status indicator, specifically including: constructing a degradation rate field model of the energy storage device according to the health status indicator, the degradation rate field model collects the time series data of the health status indicator (such as the state entropy value information) in real time through the Kalman filter, calculates the attenuation gradient of the equipment health (i.e., degradation rate), so as to quantitatively reflect the decay speed of the equipment from the healthy state to the fault state; and outputs the monitoring collaborative boundary of dynamic monitoring from the degradation rate field model. The monitoring collaborative boundary refers to the quantitative critical threshold range of the equipment health state, stable operation stage-slow degradation stage-rapid degradation stage based on the degradation rate field model of the energy storage device. Its core function is to provide a unified judgment standard for the dynamic monitoring frequency to avoid excessive or insufficient monitoring. The monitoring collaborative boundary is specifically composed of the degradation rate lower limit threshold and degradation rate upper threshold Composition, of which is the critical value of the equipment in the stable operation stage, The output process is as follows: first retrieve the historical fault data of the same type of equipment in the past three years, combine it with the real-time degradation rate data of the current equipment, perform cluster analysis through Gaussian mixture model, and automatically generate and Specific values ​​(for example: battery module =0.005 health score / day, =0.02 health score / day, converter =0.003 health score / day, = 0.015 health score / day), the final output [ , ] The interval form is stored in the module database and is recalculated and updated every hour based on the real-time degradation rate of the equipment to ensure that the boundary is synchronized with the current health status of the equipment; the monitoring interaction cycle of the energy storage equipment during operation is determined by the monitoring collaborative boundary. Based on the monitoring collaborative boundary to determine the monitoring interaction cycle, a two-dimensional judgment rule of boundary interval matching + equipment importance weighting is adopted. The specific steps are: ① Obtain the real-time degradation rate of the current equipment , and monitoring collaborative boundaries[ , ] to perform interval matching. If < (Stable phase), the default monitoring interaction period is set to (e.g. 10 minutes); if ≤ ≤ (slow degradation phase), monitoring interaction cycle according to the formula = ×( - ) / ( - ) calculation (Example: =0.01 health score / day, = 0.005 health score / day, =0.02 health score / day, = 10 minutes, then =10×(0.02-0.01) / (0.02-0.005) = 6.7 minutes); if > (rapid degradation phase), the monitoring interaction period is set to ;② Combined with the weighted adjustment of equipment importance, for equipment located in the critical path of fault propagation, even if < , and the monitoring cycle will be shortened; finally, the adjusted cycle will be distributed to each distributed monitoring node through the message queue to ensure that the data collection frequency is synchronized with the equipment degradation status.

[0028] Taking the battery module B01-02 in the energy storage power station as an example, the monitoring collaborative boundary and cycle determination process are as follows: Monitoring collaborative boundary output: The monitoring cycle decision module calculates the real-time degradation rate of the battery module through the Kalman filter =0.012 health score / day; combined with historical data, the output monitoring coordination boundary of this type of battery module is [0.005,0.02] health score / day; the monitoring interaction cycle is determined: (0.005)≤ (0.012)≤ (0.02), according to the formula =10×(0.02-0.012) / (0.02-0.005) is calculated to be 5.3 minutes. Since the battery module is a key device in the main power supply circuit of the power station, weighted adjustment is required, and the final monitoring interaction cycle is determined to be 4 minutes. Cycle execution: This cycle is sent to the distributed monitoring node corresponding to B01-02 through the message queue, and the node automatically switches from the original 10-minute / time collection frequency to 4 minutes / time.

[0029] During actual system operation, the device topology overview module regularly synchronizes with the power plant monitoring system to ensure the timeliness of topology data. The health status compensation module records an operation log after each data compensation, including the location of the compensation node, the calculation basis, and the compensation results, for subsequent analysis and reference. Adjustment records made by the monitoring cycle decision module are stored in the operation and maintenance knowledge base for optimizing the parameters of the degradation rate field model.

[0030] In terms of specific implementation, the device topology overview module uses a distributed graph database to store topology data, supporting high-concurrency queries and real-time updates. The health status compensation module's algorithms are deployed on edge computing nodes, leveraging local computing resources to reduce network transmission latency. The monitoring cycle decision module's policy execution utilizes an event-driven architecture, immediately triggering policy recalculation when health status indicators change. System modules interact via lightweight APIs to ensure efficient data flow and low-latency response. Furthermore, all core algorithms support online learning, dynamically optimizing model parameters based on newly collected data to improve the accuracy of condition monitoring.

[0031] Taking full account of the actual operational characteristics of energy storage power station equipment, the system achieves accurate equipment health assessment and risk warning through the synergy of topological modeling, state compensation, and dynamic monitoring. The system adopts a modular design, allowing each functional component to be independently upgraded or replaced, easily adapting to the operation and maintenance needs of energy storage power stations of varying sizes. The entire implementation process emphasizes the adaptability of algorithms and the real-time nature of data, ensuring the system's stable operation and reliable decision-making support in complex environments.

[0032] Example 2: See Figure 3 The risk signature analysis module accesses the energy storage power station's central database and extracts a dataset of fault events and equipment operation logs recorded over the past five years. The module first constructs a graph-based fault propagation model, abstracting each energy storage device as a graph node and converting the electrical connections and energy transfer relationships between devices into directed edges. Directed edge weights are calculated by statistically analyzing historical fault propagation frequencies. For example, if a battery cluster experiences three voltage anomalies that spread to connected converters, the path weight is set to 0.75. Node attributes include dynamic data such as device type, service life, and the last thirty maintenance records.

[0033] The spatiotemporal correlation model utilizes a three-layer graph neural network architecture for processing. The underlying node encoder converts real-time device operating parameters (such as battery internal resistance fluctuation and converter switching frequency offset) into a 128-dimensional feature vector. These real-time device operating parameters are the core component of the device node's dynamic behavior. Taking converter node C02-03 as an example, the specific dynamic behavior data for this device node are as follows: ① Real-time operating parameter change dimension: output current change rate of 0.3 A / s, bus voltage sag frequency of 0.5 times / hour; ② State transition dimension: two start-up-run-shutdown transitions per day, with an 8-hour interval between grid-connected and off-grid modes; ③ Load response dimension: power regulation delay of 150 ms, current stabilization time of 3 seconds. This dynamic behavior data is matched with three overcurrent fault events at this node over the past five years using the spatiotemporal correlation model to provide data support for the output fault coupling strength. The middle-layer spatiotemporal convolution module uses a sliding time window to scan the device state change trajectory, capturing the migration patterns of fault characteristics over different time periods. The multi-head attention mechanism in the output layer calculates the strength of fault coupling between nodes. For example, if the harmonic distortion rate of the current at a transformer node is abnormal, the system automatically associates it with the battery nodes within its power supply range and labels the coupling coefficient. Historical fault cases are converted into semantic features through the embedding layer and then matched with real-time data streams in vector space for similarity.

[0034] Fault propagation path analysis uses a random walk algorithm. After setting an initial fault point, it simulates the spread of abnormal conditions across the topological network. The algorithm configures six typical propagation strategies, including physical paths such as electrical parameter fluctuation propagation, mechanical resonance propagation, and thermal conduction propagation. A single simulation generates a path tree diagram consisting of 500 walks, recording the number of times each node is affected and the propagation delay. After 3,000 Monte Carlo simulations, the module outputs a fault impact feature matrix, which includes three parameters: fault source identification index, path criticality score, and node vulnerability coefficient.

[0035] The risk situation assessment module deploys a fuzzy inference engine, which receives node fault coupling data from upstream. This engine is configured with seven sets of inference rules, quantifying fault correlation constraints into three dimensions: topological correlation (0-1), historical recurrence frequency (0-1), and equipment criticality (1-5). The granularity analysis of abnormal disturbances uses multi-sensor data fusion technology. The specific processing flow is as follows: ① Ambient temperature and humidity data processing: Real-time data from the ambient temperature and humidity sensors (temperature unit: °C, humidity unit: %RH) is collected and normalized using a sliding window (window duration: 10 minutes). The deviation between the data within the window and the normal operating temperature and humidity thresholds of the equipment (temperature: 25±2°C, humidity: 40±5%RH) is calculated to generate temperature and humidity offset data. ② Vibration data processing: Real-time data from the vibration monitoring sensor (unit: m / s²) is collected and Fourier transformed using a spectral feature extractor to convert the vibration signal into a spectral energy distribution vector. The effective value of the vibration signal is then calculated based on the energy distribution to generate mechanical vibration intensity data. ③ Electromagnetic interference data processing: Real-time data from the electromagnetic interference sensor (unit: dBμV / m) is collected and converted into a standardized value in the range of 0-1 using an interference intensity quantization algorithm to generate electromagnetic interference level data. The above temperature and humidity offset, mechanical vibration intensity, and electromagnetic interference level data together form the basis for evaluating the granularity of abnormal disturbances. The disturbance level classification adopts an adaptive clustering algorithm. When the environmental parameters of a certain area deviate from the normal range by more than three times the standard deviation, it will automatically upgrade to a level three disturbance.

[0036] The risk level calculation implements a weighted comprehensive evaluation system. The total weight of the fault correlation constraint is set to 70%, of which the topological correlation accounts for 40%, the historical recurrence frequency accounts for 20%, and the equipment criticality accounts for 10%. The abnormal disturbance granularity accounts for 30% of the weight, and its sub-items include temperature and humidity offset, mechanical vibration intensity, and electromagnetic interference level. The evaluation rule base presets 24 conditional judgment statements, such as "when the topological correlation of a battery cabinet exceeds 0.8 and the ambient temperature rises by 8°C", a specific risk calculation path is triggered. The final generated risk situation level is output using a five-level system. The risk situation level is specifically divided into five levels. The definition and judgment criteria of each level are as follows: Level 1 (Smooth Operation): The equipment has no potential failure risks and is operating stably. Criteria for this evaluation are: node fault coupling strength <0.3, abnormal disturbance granularity <0.2. Typical scenarios include: battery module voltage balance >98%, ambient temperature and humidity maintained at 25±2°C, 40±5%RH, and no vibration or electromagnetic interference.

[0037] Level 2 (Minor Risk): The equipment has a very minor fault risk, and the impact of environmental interference on operation is negligible. The criteria for determination are: 0.3 ≤ node fault coupling strength < 0.5, 0.2 ≤ abnormal disturbance granularity < 0.4. Typical scenarios include: individual battery cell voltage fluctuations of ±0.08V (normal ≤ ±0.05V), a short-term increase in ambient temperature to 28°C, no signs of fault propagation, and the equipment continuing to operate normally.

[0038] Level 3 (Medium Risk): The device is showing a clear trend toward failure, and environmental disturbances are already impacting local performance. The criteria for this are: 0.5 ≤ node fault coupling strength < 0.7, 0.4 ≤ abnormal disturbance granularity < 0.6. Typical scenarios include: battery cluster internal resistance deviation reaching 15% (normal ≤ 10%), ambient humidity rising to 55% accompanied by slight vibration (vibration frequency 20-30Hz). The fault may spread to one or two adjacent device nodes, necessitating increased monitoring frequency.

[0039] Level 4 (High Risk): A localized fault has occurred in the equipment, and environmental interference is exacerbating the fault's development. The criteria are: 0.7 ≤ node fault coupling strength < 0.9, 0.6 ≤ abnormal disturbance granularity < 0.8. Typical scenarios include: the converter output current harmonic distortion reaches 5% (normal ≤ 3%), the ambient temperature rises to 35°C, and the electromagnetic interference intensity exceeds 80dB. The fault has formed a localized propagation chain, requiring preventive maintenance preparations.

[0040] Level 5 (Emergency Failure): A serious equipment failure has occurred, and environmental interference has caused the fault to spread rapidly, posing a direct threat to power plant safety. The criteria for this failure are: node fault coupling strength ≥ 0.9, and abnormal disturbance granularity ≥ 0.8. Typical scenarios include: a battery pack showing signs of thermal runaway (temperature rise of 10°C / min), an ambient vibration frequency exceeding 50Hz, and the fault has spread to three or more critical equipment nodes, requiring immediate shutdown for repair.

[0041] The judgment thresholds for each level are calibrated based on energy storage power station failure cases and environmental impact experimental data from the past five years to ensure the accuracy and operability of risk assessment.

[0042] During the system's operation cycle, the risk signature analysis module automatically updates the fault propagation model at dawn each day, integrating the previous day's equipment operation and maintenance records. Parameter training for the spatiotemporal correlation model is performed three times a week, fine-tuning network weights using newly added fault cases. The risk situation assessment module refreshes risk level data every five minutes, initiating real-time calculations when environmental sensors report sudden disturbances. All intermediate results are stored in a time series database, including traceable records such as the historical coupling strength matrix, environmental disturbance spectrum, and risk level change curves.

[0043] At the engineering implementation level, the fault propagation model uses a graph database for persistent storage, supporting millisecond-level path queries. The inference process of the spatiotemporal correlation model is deployed on edge servers equipped with TensorRT acceleration, keeping the response time for a single calculation within 300 milliseconds. The risk level calculation engine utilizes a distributed stream processing framework to ensure stable output even with high-concurrency data input. The system incorporates an exception circuit breaker mechanism that automatically triggers a site-wide early warning protocol if a level 5 risk is calculated three times consecutively. The module also includes a reserved expansion interface, allowing the configurator to adjust the weighting coefficients of different device types in risk calculations to meet the specific needs of various energy storage power plants.

[0044] Example 3: See Figure 4 The maintenance blind spot warning module is built on a timed Petri net model, which transforms the monitoring cycle and maintenance activities of the energy storage power station into a discrete event system with timing constraints. The place nodes in the model represent the health status of the equipment, the transition nodes correspond to monitoring behaviors or maintenance operations, and the time intervals reflect the time spans required for different state transitions. During system initialization, an independent Petri net sub-model is established for each key device, and the corresponding timing parameters are configured according to the device type. For example, the health status place of the battery module is set with a 30-minute time delay, while the maintenance operation transition of the inverter is configured with a variable time window of 15-45 minutes.

[0045] The elastic monitoring threshold is calculated using a sliding window mechanism. The window size W is dynamically adjusted according to the device type. The calculation formula is:

[0046] in: is the smoothing factor (value is 0.6), is the average interval of the device's past 10 health status assessments, This is the maximum historically permitted interval. The health data within the window is processed through a Savitzky-Golay filter, and its second-order derivative is calculated as the state change acceleration indicator. When this indicator exceeds the device's sensitivity coefficient, the system automatically shortens the monitoring interval, with the extent of the shortening linearly proportional to the excess.

[0047] The calculation process for the blind spot exposure factor integrates failure mode and effects analysis (FMEA) and risk situation data. The system maintains a three-dimensional fault monitoring coverage matrix, with dimensions corresponding to equipment type, failure mode, and monitoring method. Each cell stores the historical monitoring success rate for that combination, with data sourced from the past three years of operation and maintenance records. The risk situation level is converted into an impact weight through normalization, and after performing a Hadamard product operation with the coverage matrix, the exposure risk value for each maintenance link is obtained. The system has a dynamic adjustment mechanism. If a certain type of fault is not detected in a timely manner for three consecutive times, the weight of the relevant cell is automatically increased by 20%.

[0048] Early warning maps are generated using a 3D visualization engine, spatially integrating the device topology, time dimension, and risk heat map. In the visualization coordinate system, the X-axis represents the physical location of the equipment within the power plant, the Y-axis represents time progression, and the Z-axis quantifies the risk level. The heat map rendering algorithm utilizes an improved kernel density estimation algorithm. The radius R of each data point is determined by the following factors: the deviation between the current monitoring interval and the threshold, the blind spot exposure coefficient, and the criticality of the equipment in the fault propagation path. The map supports multi-scale display, allowing operators to freely switch between a macro view of the power plant and a micro view of the equipment.

[0049] During system operation, the maintenance blind spot warning module maintains real-time data synchronization with the health status monitoring system. Whenever a new health assessment result arrives, the module immediately updates the Petri net model state of the corresponding device and recalculates relevant timing parameters. Adjustment signals for elastic monitoring thresholds are broadcast to all monitoring nodes via a publish-subscribe model, ensuring a rapid response from the data acquisition system. Blind spot exposure coefficients are updated in batch mode, integrating the previous 24 hours' worth of maintenance data and recalculating the matrix values ​​at dawn each day.

[0050] The push notification strategy for early warning information utilizes a tiered mechanism. A primary alert is triggered when the monitoring interval exceeds the threshold by 15%, displaying a prompt message only on the work terminal. An intermediate alert is activated when the threshold exceeds 30% and the exposure factor exceeds 0.7, sending a text message notification to the responsible engineer. A high-level alert requires the simultaneous fulfillment of three conditions: an interval exceeding 50%, an exposure factor exceeding 0.9, and the equipment being on the critical path. At this point, an audible and visual alarm is activated, and an emergency repair work order is automatically generated. All alert events are recorded in the audit log, including the trigger time, relevant parameters, and the handling status.

[0051] In terms of implementation architecture, the timed Petri net model uses a distributed event stream processor to perform state calculations, ensuring real-time performance in high-concurrency scenarios. Elastic monitoring threshold calculations are assigned to edge computing nodes, leveraging local cache to accelerate data access. The visualization engine, developed based on WebGL technology, supports cross-platform access and multi-person collaborative viewing. The system also includes a reserved API interface for integration with third-party maintenance management systems, enabling automatic integration of warning information with work order systems.

[0052] The Maintenance Strategy Optimizer regularly analyzes warning records to identify frequently occurring blind spot patterns. This analysis utilizes an association rule mining algorithm to extract underlying patterns between monitoring interval settings and maintenance effectiveness from historical data. Optimization recommendations are output as configuration adjustment plans, encompassing three types of measures: revising monitoring frequency, supplementing monitoring methods, and improving equipment layout. The system automatically generates blind spot trend reports quarterly to assist management in developing long-term preventive maintenance plans.

[0053] The module's exception handling mechanism incorporates a multi-level fallback strategy. When health status data arrives late, the system generates temporary estimates based on historical patterns. When computing resources are limited, the system automatically reduces visualization rendering accuracy to ensure the real-time performance of core early warning functions. In the event of a network outage, edge nodes independently maintain basic monitoring functions. All abnormal events trigger a self-diagnosis process, recording a snapshot of the system status for subsequent analysis.

[0054] The maintenance blind spot warning module forms a closed-loop control loop with other power plant subsystems. Dynamic adjustments to the monitoring cycle affect the density of health data collection, which in turn changes the accuracy of risk assessments. Changes in risk levels are then fed back into the calculation of blind spot exposure coefficients, forming an adaptive optimization loop. The system continuously accumulates operational data and, through online learning mechanisms, gradually improves the accuracy and timeliness of warnings.

[0055] Example 4: In actual deployment, the encrypted transmission submodule of the multi-source monitoring module was specifically implemented for the battery management system of a 100MWh energy storage power station. The power station contains 32 battery compartments, and 8 battery cluster monitoring nodes are arranged in each compartment. The system adopts a layered encryption architecture, and the edge layer monitoring nodes are equipped with hardware encryption chips. After collecting battery voltage, temperature and other data, they immediately perform AES-256 algorithm encryption processing. The encrypted data packet is added with a timestamp and node ID identification and transmitted through a pre-configured quantum key distribution channel. The quantum key is updated every 30 minutes and distributed to each node by the central key management server through the optical fiber network. During the transmission process, the system continuously monitors the channel quality, and automatically switches to the backup national secret SM4 encryption channel when it detects that the bit error rate exceeds 10^-5.

[0056] Based on the actual layout of the power plant, the distributed processing submodule divides the 32 battery compartments into four monitoring zones, deploying an edge computing gateway in each zone. The gateway device collects real-time operating data from the eight monitoring nodes within the zone and implements the following adaptive processing strategy (see Table 1).

[0057] Table 1: Monitoring node network status evaluation parameters

[0058] The parameters in the table are updated hourly, and relay strategy codes correspond to different data transmission schemes. R01 indicates a direct connection to the central server, R02 relays data via adjacent nodes, R03 uses multi-path concurrent transmission, and R05 enables local caching and delayed transmission. The gateway device dynamically adjusts the communication strategy for each node based on real-time network assessment results. For example, if the signal strength at node B02-01 drops to -81dBm, the system automatically switches to strategy R05, temporarily buffering data locally and sending it after channel quality improves.

[0059] The real-time network status and interference intensity serve as the core basis for the distributed processing submodule to generate the node data relay strategy. The specific meanings and quantitative standards are as follows: Real-time network status refers to the real-time operating status of the data transmission link between the distributed monitoring node and the edge gateway / central server. The core is used to judge the link stability and data transmission capacity. It specifically includes three quantitative indicators: ① Signal strength: uses dBm (decibel milliwatt) as the unit to measure the power of the node receiving / sending signal. The value range is [-110dBm, -50dBm], where the [-70dBm, -50dBm] interval indicates a good signal, the (-90dBm, -70dBm) interval indicates a general signal, and the [-110dBm, -90dBm] interval indicates a weak signal; ② Network delay: refers to the time it takes for data to be sent from the monitoring node to the receiving end The difference (in milliseconds) ranges from [0ms to 1000ms], where [0ms to 50ms] indicates low latency (suitable for real-time monitoring), [50ms to 200ms] indicates medium latency (needs transmission path optimization), and [200ms to 1000ms] indicates high latency (needs relay strategy switching). (3) Data integrity rate: This refers to the proportion of successfully transmitted packets per unit time to the total number of sent packets. The range is [0% to 100%], where [0% to 95%] indicates insufficient integrity (needs retransmission), [95% to 99%] indicates acceptable integrity, and [99% to 100%] indicates excellent integrity. These metrics are collected every 30 seconds using the network diagnostic tools built into the distributed monitoring nodes (such as the ping command and the packet statistics module) and synchronized in real time to the distributed processing submodule.

[0060] Interference intensity refers to the degree of external interference that affects the data transmission of distributed monitoring nodes, focusing on two core interference sources: electrical interference and electromagnetic interference. It specifically includes two quantitative indicators: ① Electromagnetic interference intensity: uses dBμV / m (decibel microvolts per meter) as the unit to measure the interference level of the electromagnetic environment around the node. The value range is [0dBμV / m, 120dBμV / m], where [0dBμV / m, 60dBμV / m] indicates slight interference, (60dBμV / m, 90dBμV / m) indicates moderate interference, and [90dBμV / m, 120dBμV / m] indicates severe interference; ② Signal bit error rate: refers to the ratio of the number of erroneous data packets in the transmission process to the total number of transmitted data packets, with the unit being [0dBμV / m, 120dBμV / m]. (For example =1 error per 100,000 packets). Very low bit error rate range: ≤ (i.e. the percentage of erroneous data packets is ≤ 1 in a million); acceptable range of bit error rate: ( , ) (i.e. the percentage of erroneous data packets is between one in a million and one in ten thousand, excluding the boundary value); Bit error rate is too high range: ≥ (That is, if the proportion of error packets is ≥ 0.1%, the encrypted anti-interference channel needs to be activated.) The electromagnetic interference intensity is collected by electromagnetic interference sensors deployed on the nodes, and the signal bit error rate is calculated in real time using the packet verification function of the distributed processing submodule. Both indicators are updated once every minute.

[0061] The decoy device control module deploys six multifunctional decoy terminals around the power station perimeter. Each terminal contains a thermodynamic simulation unit and a radio frequency signal generator. The thermodynamic simulation unit has a built-in programmable heating array that can reproduce the infrared radiation characteristics of the battery compartment during operation. When the system detects an unauthorized device scan, the decoy terminal initiates a preset interference protocol: first, the position offset is calculated based on the device's true position coordinates to generate a decoy coordinate set with a random error of ±3 meters; then, the Doppler frequency shift algorithm is used to simulate the device's movement trajectory; finally, the heating array power and radio frequency signal strength are synchronously adjusted to form a false target with spatiotemporal continuity. Each decoy cycle lasts 15 minutes, during which the system continuously monitors the response pattern of the intruding device and dynamically adjusts the decoy parameters.

[0062] Inside the battery compartment, the system installs miniature decoy beacons. These beacons, about the size of a matchbox, are discreetly installed within the equipment cabinet. Their operating frequency is offset by 20 MHz from that of the actual monitoring nodes. When they detect abnormal signal collection behavior, they actively transmit jamming signals containing falsified device parameters. The beacons maintain synchronization via power line carrier communication to ensure the timing consistency of the decoy signals. The system maintains a decoy strategy library containing 12 typical device parameter falsification patterns, automatically matching the optimal jamming strategy based on the characteristics of the intrusion.

[0063] The encrypted transmission submodule's exception handling mechanism includes a three-level response strategy. The primary response addresses communication anomalies at individual nodes, attempting to switch encryption algorithms or transmission paths. The intermediate response, initiated when a regional gateway fails, activates a backup gateway to restructure the data transmission network. The advanced response addresses system-wide security threats, immediately severing external network connections and activating site-wide deception mode. All response processes are recorded in the security audit log, including trigger time, disposition, and final outcome.

[0064] During system implementation, several adaptive improvements were made to address the unique environment of power stations. The strong electromagnetic environment within the battery compartment required the encryption chip to possess enhanced anti-interference capabilities. The industrial-grade chip ultimately selected can operate normally under a field strength of 100V / m. High temperature and high humidity conditions prompted the system to incorporate enhanced heat dissipation design and three-proofing solutions. All outdoor decoy terminals meet IP67 protection levels. To account for potential signal blind spots within power stations, the system solution incorporates a 4G / 5G wireless backup channel to maintain basic monitoring functionality in the event of a complete wired network outage.

[0065] Operations and maintenance personnel manage the entire system through a dedicated console, which provides a panoramic view of the encrypted channel status, a map of decoy deployment, and a real-time security event dashboard. The user interface utilizes a layered design: the first level displays the overall system status, the second level displays detailed regional information, and the third level provides real-time data streams for individual nodes. All management operations require two-factor authentication and maintain a comprehensive log.

[0066] The system integrates loosely with the power plant's existing monitoring platform, exchanging data via a standard OPC UA interface. Encrypted monitoring data is decrypted and stored in a real-time database, accessible to other system modules. Information on the decoy device's status is transmitted via an independent channel, ensuring its concealment. The system automatically generates a weekly security assessment report, analyzing key metrics such as the encryption transmission success rate, decoy trigger counts, and network anomalies.

[0067] The network evaluation parameters listed in the table are derived from three consecutive months of system operation data. The relay strategy selection algorithm has been repeatedly validated in the field and is capable of optimizing network resource utilization while ensuring data reliability. Actual operational data shows that the combination of different strategies enables the system to maintain a data integrity rate of over 98% even in adverse weather conditions. The effectiveness of the deception system is evaluated through regular penetration testing, which demonstrates that the decoy targets can effectively mislead over 85% of conventional scanning devices.

[0068] Example 5: See Figure 5The core function of the maintenance strategy decision module is to establish a dynamic maintenance window calculation model, which is continuously optimized based on the real-time health status data of the equipment and historical maintenance records. The system first establishes a life prediction curve for various types of equipment in the energy storage power station and uses an improved Weibull distribution function to describe the equipment degradation process. The remaining time of the maintenance window The calculation uses the following formula:

[0069] in: is the equipment operating condition correction factor, is the allowed failure probability threshold, and is the shape parameter of the Weibull distribution, Indicates the equipment's operating time. This calculation is automatically performed every hour, and when the remaining operating time falls below the preset warning level, the maintenance strategy adjustment process is triggered.

[0070] The number of fault mode types is counted using a graph database-based association analysis method. The system maintains a multidimensional fault feature space, with each dimension corresponding to a fault manifestation, such as battery capacity degradation and converter harmonic distortion. When a new fault event occurs, the feature extractor maps it to this space and uses a density clustering algorithm to determine whether it belongs to an existing fault mode. The system uses an adaptive threshold mechanism. When a certain type of fault recurs more than three times within two weeks, it is automatically upgraded to an independent fault mode category. The number of currently active fault modes serves as an important input parameter for policy adjustment and directly affects the calculation weight of the maintenance cycle.

[0071] The dynamic generation of maintenance strategies utilizes a two-tiered decision-making mechanism. The basic strategy library contains 32 standard maintenance plans for different equipment types, such as battery balancing maintenance and converter filter capacitor replacement. The dynamic adjustment layer optimizes the parameters of the basic strategies based on real-time calculations of remaining maintenance time and the number of failure modes. The strategy matching engine evaluates the progress of the current maintenance task hourly and, based on trends in equipment health, predicts maintenance needs for the next 24 hours. The output includes decision data on three dimensions: maintenance priority ranking, recommended operation duration, and resource allocation plan.

[0072] The first related conversion relationship between the remaining time and the maintenance cycle is a technical rule that describes the relationship between the remaining time between the current moment and the maintenance window period, and the maintenance cycle of energy storage equipment. Its core logic is: the shorter the remaining time, the shorter the maintenance cycle needs to be; the longer the remaining time, the equipment is still in a relatively safe operating range, and the maintenance cycle can be appropriately extended to reduce unnecessary consumption of operation and maintenance resources.

[0073] In actual applications, the relationship needs to adjust the strength of the association in combination with the type of energy storage equipment and the differences in core functions: for core equipment such as energy storage battery packs and converters, the correlation between the remaining time and the maintenance cycle is higher. For non-core equipment such as auxiliary cooling systems and lighting equipment, the correlation is relatively low. If the remaining time of the cooling system maintenance window is shortened from 72 hours to 36 hours, the maintenance cycle only needs to be shortened from the original 90 days to 75 days, balancing operation and maintenance costs and equipment safety. At the same time, the conversion relationship needs to exclude the influence of temporary interference factors, such as misjudgment of the remaining time due to temporary failure of the environmental sensor. The system will compare historical data with the equipment health status indicators, correct the misjudgment data, and then apply the conversion relationship to ensure the accuracy of the maintenance cycle adjustment.

[0074] The second related conversion relationship between the number of fault mode types and the cycle adjustment amount is a technical rule that describes the correlation between the number of fault mode types in the energy storage device cluster and the maintenance cycle adjustment amount. Its core logic is: the more fault mode types there are, the more fault types need to be covered by increasing the maintenance cycle adjustment amount (i.e., shortening the maintenance cycle); the fewer fault mode types there are, the smaller the maintenance cycle adjustment amount can be to avoid excessive operation and maintenance.

[0075] When applying this relationship, the fault mode impact and severity should be tiered. For minor localized faults (such as single-cell voltage deviation or unusual noise from a small fan), the maintenance interval should be adjusted by 5%-8% of the current maintenance interval for each additional fault mode. For cluster-affecting faults (such as unbalanced battery cluster charge and discharge or excessive converter output harmonics), the maintenance interval should be adjusted by 10%-15% of the current maintenance interval for each additional fault mode. If a new cluster-affecting fault mode is added to the aforementioned equipment, the adjustment will be 3-4.5 days, resulting in a corrected maintenance interval of 25.5-27 days. For safety-critical faults (such as precursors to battery thermal runaway or DC bus insulation degradation), the maintenance interval should be adjusted by 20%-25% of the current maintenance interval for each additional fault mode. A temporary special investigation should be initiated immediately, and targeted maintenance should be completed before the correction interval is initiated. In addition, the statistics of fault mode types need to distinguish between confirmed faults and suspected faults. Suspected faults (preliminarily judged by health status indicators but not confirmed) are included in the total number with a weight of 0.5 to avoid excessive shortening of the maintenance cycle due to misjudgment.

[0076] The maintenance strategy decision module integrates the first and second related conversion relationships through a three-step process of basic cycle calculation, adjustment amount correction, and strategy refinement, and outputs a dynamic maintenance strategy that can be directly implemented. The specific steps are as follows: Step 1: Determine the basic maintenance cycle based on the first correlation conversion relationship. The system first obtains the remaining maintenance window duration of the current device (derived from the maintenance window calculation model), then matches the corresponding association rules based on the device type to calculate the basic maintenance cycle.

[0077] Step 2: Based on the second correlation conversion relationship, the maintenance cycle is corrected. The system counts the number and types of current fault modes in the device cluster where the converter is located, calculates the maintenance cycle adjustment amount, and corrects the basic cycle.

[0078] Step 3: Refine and output the dynamic maintenance strategy. Combined with the revised maintenance cycle, supplement the operation and maintenance execution details to form a complete strategy, including: Maintenance frequency: According to the 54-day cycle, a simple status check is carried out every 18 days (focusing on key parameters such as voltage and current), and a comprehensive in-depth maintenance is carried out every 54 days (covering the entire process of disassembly inspection, component calibration, and troubleshooting); Maintenance content priority: Prioritize two types of cluster-affecting faults (such as excessive output harmonics of the converter, adding harmonic detector calibration and filter capacitor performance testing to maintenance; for unbalanced charging and discharging of battery clusters, adding inter-cluster voltage balancing adjustment), and then deal with one local minor fault; Resource allocation: One engineer with converter commissioning qualifications and two auxiliary operation and maintenance personnel are deployed for comprehensive in-depth maintenance, and special equipment such as harmonic detectors and insulation resistance testers are prepared; Emergency triggering conditions: If the remaining time of the equipment is further shortened to less than 36 hours within the 18-day simple inspection interval, or a new safety-critical failure mode is added, the emergency maintenance process will be immediately initiated, the next comprehensive maintenance time will be advanced to within 7 days, and spare parts will be deployed simultaneously.

[0079] The maintenance decision output module's three-dimensional overlay algorithm utilizes voxel processing technology. The system first converts the equipment topology overview into a three-dimensional grid model, with each grid cell corresponding to a physical space measuring 20 cm x 20 cm x 20 cm. The maintenance blind spot warning map is quantized as a risk density field and mapped to the grid's transparency properties. Dynamic maintenance strategies are converted into vector arrows, with direction representing the maintenance path, length representing the duration of the operation, and color used to distinguish priority levels. The overlay process utilizes a ray casting algorithm to ensure that the visualization of different data layers is independent of each other and clearly defined.

[0080] The augmented reality presentation system is deployed on the smart glasses of maintenance personnel. Device identification utilizes a spatial positioning method based on SLAM technology, achieving sub-centimeter positioning accuracy by comparing real-time images with feature points in a 3D device model library. Maintenance instructions are superimposed on the actual device as floating labels, including fault codes, operating procedures, and safety warnings. The system supports voice interaction, allowing maintenance personnel to access deeper device parameters and maintenance history through natural language commands.

[0081] In actual power plant operation, the maintenance strategy decision module is deeply integrated with the work order management system. The generated dynamic strategy is automatically converted into a standardized work order format, containing structured data such as estimated work time, required spare parts, and risk warnings. During work order execution, the system receives real-time feedback from the field. When the actual maintenance time exceeds the estimated value by 20%, the strategy recalculation process is automatically triggered. After the maintenance is completed, equipment status data is collected and returned to the system to verify the effectiveness of the strategy and optimize model parameters.

[0082] The system employs a multi-level caching mechanism to ensure real-time decision-making. The short-term cache stores the last 15 minutes of monitoring data for immediate policy fine-tuning; the medium-term cache retains a complete record of the day's operations to support trend analysis; and the long-term cache archives historical maintenance data for model training. Cached data is managed differently, retaining original accuracy for critical device data while allowing lossy compression for non-critical device data. Data updates utilize copy-on-write technology to ensure that the decision-making process is unaffected by data refreshes.

[0083] The fault pattern recognition module utilizes incremental learning for continuous optimization. After each completed maintenance task, the system automatically extracts the correspondence between fault characteristics and repair measures and updates the knowledge graph. Graph nodes represent three types of entities: fault symptoms, equipment components, and repair methods. Edge weights reflect the effectiveness of solutions. When a new fault occurs, the system uses a graph neural network to infer the most likely solution and, after implementation, adjusts the graph structure based on the actual results.

[0084] The maintenance decision map uses progressive detail loading technology to render basic equipment outlines and key risk areas, with detailed annotations and auxiliary information loaded as needed. Visualization parameters can be dynamically adjusted, including the color scale range of risk heat, the thickness of the maintenance path display, and the level of detail of the annotations. The system supports multiple viewing angles, allowing operators to freely switch between three viewing modes: a bird's-eye view, a close-up view of the equipment, and a close-up view of the maintenance process.

[0085] The module's exception handling mechanism includes data validation and fallback strategies. When sensor data is abnormal, the system automatically switches to a conservative calculation mode based on the device's rated parameters. When network latency causes a decision timeout, the system uses the latest valid strategy cached locally. When AR device positioning fails, the system switches to a two-dimensional display of key information. All abnormal events generate a diagnostic report, fully documenting environmental parameters, system status, and remediation measures.

[0086] The maintenance strategy decision module forms a closed data loop with other power plant systems. Real-time data from the monitoring system drives strategy generation, maintenance execution results provide feedback to optimize the decision model, and historical equipment data supports long-term trend forecasting. The system automatically generates monthly strategy effectiveness analysis reports, analyzing key indicators such as timely repair rate, fault recurrence rate, and resource utilization, providing data support for management decisions. The module utilizes a microservices architecture, allowing each functional component to be independently upgraded and expanded to adapt to the changing operation and maintenance requirements of different energy storage power plants.

[0087] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0088] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A system for monitoring the status of energy storage power station equipment and making maintenance decisions, characterized in that: The system comprises: A device topology overview building module is used to generate a device topology overview based on the physical connection relationship of the energy storage power station. The device topology overview includes electrical connection parameters and operating condition parameters of the energy storage device cluster; Multi-source monitoring module, used to collect equipment operation status data streams from distributed monitoring nodes and obtain environmental disturbance data of energy storage equipment in real time; A health status compensation module is used to compensate for missing nodes in the device operation status data stream in the device status fusion monitoring mode to obtain the health status indicator of the energy storage device; The health status compensation module is specifically used to: determining a state transition path of the energy storage device according to the device topology overview; Probabilistically compensating for missing state nodes in the device operation state data stream to generate a complete state event sequence; Performing feature aggregation on the state event sequence to obtain state entropy value information of the energy storage device; The health status index of the energy storage device is determined based on the state entropy value information.

2. The energy storage power station equipment status monitoring and maintenance decision system according to claim 1, characterized in that: The system further comprises: The monitoring cycle decision module is used to determine the monitoring interaction cycle of the energy storage device during operation according to the health status indicator, specifically including: Constructing a degradation rate field model of the energy storage device according to the health status indicator; Outputting a monitoring cooperation boundary for dynamic monitoring from the degradation rate field model; The monitoring interaction period of the energy storage device during operation is determined by the monitoring cooperation boundary.

3. The energy storage power station equipment status monitoring and maintenance decision-making system according to claim 2, characterized in that: The system further comprises: The risk feature analysis module is used to obtain the fault propagation path information of the energy storage device cluster, perform risk feature analysis on the fault propagation path information, and obtain node fault coupling data of the energy storage device during operation; specifically, it includes: Building a spatiotemporal correlation model between the dynamic behavior of device nodes and historical fault events based on the fault propagation path information; Outputting the fault coupling strength of the energy storage device during operation through the spatiotemporal correlation model; The fault coupling strength is used to determine the fault impact characteristics of each device node in the device cluster.

4. The energy storage power station equipment status monitoring and maintenance decision-making system according to claim 3, characterized in that: The system further comprises: The risk situation determination module is used to determine the risk situation level during the operation of the equipment based on the node fault coupling data and the environmental disturbance data; specifically includes: determining fault correlation constraints during equipment operation according to the node fault coupling data; Determining the abnormal disturbance granularity during the operation of the equipment through the environmental disturbance data; A risk situation level is determined according to the fault association constraint and the abnormal disturbance granularity.

5. The energy storage power station equipment status monitoring and maintenance decision-making system according to claim 4, characterized in that: The system further comprises: The maintenance blind spot warning module is used to warn of dynamic maintenance blind spots during equipment operation and maintenance based on the monitoring interaction cycle and the risk situation level; specifically, it includes: determining a resilience monitoring threshold of the maintenance process according to the monitoring interaction cycle; determining a blind spot exposure coefficient in a dynamic maintenance blind spot according to the risk situation level; The elastic monitoring threshold and the blind spot exposure coefficient are strategically matched to generate a dynamic maintenance blind spot warning map.

6. The energy storage power station equipment status monitoring and maintenance decision-making system according to claim 1, characterized in that: The multi-source monitoring module specifically includes: An encryption transmission submodule, used to create an anti-interference monitoring channel and encrypt and transmit the device operation status data stream through a quantum key distribution channel and an encryption algorithm channel; The distributed processing submodule is used to construct N distributed monitoring nodes and adaptively generate N node data relay strategies based on the real-time network status and interference intensity of the N distributed monitoring nodes.

7. The energy storage power station equipment status monitoring and maintenance decision-making system according to claim 6, characterized in that: The system further comprises: The deception device control module is used to set up the device status deception device and modulate and emit multi-source deception signals based on the device's real location information; specifically, it includes: Deploy a thermodynamic characteristic simulation unit and a radio frequency signal generator on the equipment status deception device; generating decoy signal modulation parameters according to the device position offset; The device location information is modulated by the decoy signal modulation parameter.

8. The energy storage power station equipment status monitoring and maintenance decision-making system according to claim 5, characterized in that: The system further comprises: The maintenance strategy decision module is used to calculate the remaining time between the current time and the maintenance window period and determine the maintenance strategy based on the remaining time. Specifically, it includes: Retrieve the first relevant conversion relationship between the remaining time and the maintenance period; Determine the number of failure modes of the energy storage device cluster; Retrieving a second correlation conversion relationship between the number of fault mode types and the period adjustment amount; A dynamic maintenance strategy is output according to the first correlation conversion relationship and the second correlation conversion relationship.

9. The energy storage power station equipment status monitoring and maintenance decision-making system according to claim 8, characterized in that: The system further comprises: A maintenance decision output module, configured to perform a three-dimensional overlay of the dynamic maintenance blind spot warning map and the dynamic maintenance strategy to generate a dynamically rendered maintenance decision map; The maintenance decision map is spatially mapped through the equipment topology overview building module.

Images