State monitoring and real-time evaluation system and method for energy storage power station

Abstract

The invention discloses a state monitoring and real-time evaluation system and method for an energy storage power station, and relates to the field of energy storage power station monitoring, and the method comprises the steps: collecting multi-source heterogeneous data of the energy storage power station through constructing a hierarchical multi-scale sensing system; the collected multi-source heterogeneous data is preprocessed; inputting the processed data into the first model and the second model, carrying out classification judgment on the state of the energy storage battery, and outputting an evaluation result; the assessment result is transmitted to a safety early warning unit, the early warning unit maps and judges the state assessment result according to a preset risk grading rule, and early warning grade information and a risk label are output; and transmitting an evaluation result and an early warning result to an energy management system, and performing visual display and early warning prompt on the real-time state. An evaluation method combining a battery behavior mechanism and an artificial intelligence algorithm is introduced, dynamic prediction and trend analysis of key indexes are realized, and the accuracy and adaptability of the second model are improved.

Description

Technical Field

[0001] This invention relates to the field of energy storage power station monitoring, and in particular to a system and method for condition monitoring and real-time assessment of energy storage power stations. Background Technology

[0002] With the rapid development of distributed clean energy (such as wind and solar power) and its widespread integration into smart grids, electrochemical energy storage systems, especially lithium-ion battery-based energy storage power stations, are playing an increasingly crucial role in peak shaving, frequency regulation, voltage regulation, and energy transfer. However, lithium batteries pose a certain risk of thermal runaway, and under conditions of overcharging, overheating, or structural damage, they are prone to serious safety accidents such as fires and explosions. Existing fire protection and alarm technologies mainly rely on high-temperature or smoke detection, which is a delayed response mechanism and often fails to provide timely warnings in the early stages of an accident, resulting in passive safety protection measures and uncontrollable risks.

[0003] On the other hand, energy storage power stations are typically characterized by complex hierarchical structures, diverse operating parameters, and rapid changes in operating status. Especially in unattended or remotely dispatched environments, traditional monitoring methods are insufficient to meet the requirements for full-cycle, hierarchical, real-time, and intelligent management of system operating status. Furthermore, battery status assessments often remain at the level of single-point data judgment or empirical rules, lacking in-depth analytical capabilities based on models and structures, making it difficult to support the grid collaborative dispatch needs across multiple scenarios and time scales.

[0004] Furthermore, the current subsystems lack a unified interface and linkage logic in the status assessment and risk judgment stages, resulting in low data utilization efficiency, fragmented early warning results, insufficient overall intelligence level, and an inability to form an effective "perception-assessment-response" closed loop.

[0005] Therefore, there is an urgent need to establish a technology system for monitoring and real-time evaluation of energy storage power stations that integrates multi-source sensing, mechanism modeling, data analysis and intelligent response, so as to achieve accurate identification of battery operating status, early judgment of risk trends, and graded response and system linkage of early warning results. Summary of the Invention

[0006] In view of the above-mentioned problems, the present invention is proposed.

[0007] Therefore, the problem that this invention aims to solve is that battery status assessment often remains at the level of single-point data judgment or empirical rules, lacking in-depth analysis capabilities in terms of modeling and structure, and is difficult to support the grid collaborative scheduling needs of multiple scenarios and multiple time scales.

[0008] To address the aforementioned technical problems, this invention provides the following technical solution: a status monitoring and real-time evaluation system for energy storage power stations, comprising: a data acquisition unit for acquiring multi-source heterogeneous data from the energy storage power station at different levels; a data processing and fusion unit for preprocessing the acquired multi-source heterogeneous data; a status evaluation unit for constructing a first model reflecting the operating status of the energy storage battery and a second model for evaluating the battery's operating status, and for performing multi-dimensional analysis and evaluation of the energy storage power station's operating status based on the first and second models and real-time data; a safety early warning unit for receiving the evaluation results and issuing safety alarms based on the evaluation results and risk classification rules; and a communication and management interface unit for establishing data interaction with the energy management system.

[0009] As a preferred embodiment of the condition monitoring and real-time evaluation system for energy storage power stations described in this invention, the data acquisition unit constructs a hierarchical multi-scale sensing system, targeting each operating level in the energy storage power station, and collects multi-source heterogeneous data by deploying monitoring equipment at different levels.

[0010] As a preferred embodiment of the state monitoring and real-time evaluation system for energy storage power stations described in this invention, the first model expresses the behavioral mechanism of energy storage battery operation based on preprocessed data, the first model and the second model are fused together, and the operating status of the energy storage power station is evaluated based on real-time data.

[0011] The beneficial effects of this preferred technical solution are as follows: by integrating battery behavior mechanisms and operational data, a state assessment system with both interpretability and real-time performance is constructed. This system can not only accurately characterize the current performance of the energy storage system, but also flexibly switch assessment paths under different power grid application scenarios, realizing personalized and adaptable assessments, and significantly improving the scheduling decision support capability and operational safety of energy storage power stations.

[0012] As a preferred embodiment of the status monitoring and real-time assessment system for energy storage power stations described in this invention, the safety early warning unit is used to receive the operating status data and risk assessment results output by the status assessment unit, and classify and judge the assessment results based on preset risk classification rules, and output early warning level and linkage response signal.

[0013] As a preferred embodiment of the condition monitoring and real-time evaluation system for energy storage power stations described in this invention, the operation levels include cell-level monitoring, battery pack-level monitoring, and power station-level monitoring; the multi-source heterogeneous data includes operating parameters, environmental parameters, and structural safety parameters.

[0014] The beneficial effects of this preferred technical solution are as follows: by constructing a multi-parameter intelligent sensing system covering the three levels of cells, battery packs and power stations, it is possible to monitor the operating status of energy storage power stations in the whole time domain, in the whole space, and in all dimensions, to discover potential safety hazards in a timely manner, and to effectively improve the accuracy, reliability and response speed of data collection.

[0015] As a preferred embodiment of the condition monitoring and real-time evaluation system for energy storage power stations described in this invention, the cell-level monitoring includes: installing voltage acquisition devices at both ends of each individual cell to collect cell voltage data; attaching a thermistor to the cell casing to collect the casing surface temperature; installing a gas sensor inside the encapsulation chamber to track the concentration of released gas; and recording the natural decreasing trend of the open-circuit voltage when the cell is in standby mode, and estimating the dynamic resistance value of the internal short-circuit path based on the slight changes in steady-state current and voltage disturbance.

[0016] As a preferred embodiment of the condition monitoring and real-time evaluation system for an energy storage power station described in this invention, the battery pack-level monitoring includes: configuring a differential voltage acquisition unit at the series terminal of the battery pack to monitor the overall voltage stability and voltage fluctuation characteristics during charging and discharging; deploying a Hall current sensor in the bus circuit to acquire the input and output current of the battery pack in real time, and dynamically estimating the charging and discharging power in conjunction with the total voltage signal; deploying a temperature and humidity sensor inside the battery pack housing to collect temperature and humidity data inside the housing; and installing a corrosion sensor at the wiring terminals to identify the insulation aging trend.

[0017] As a preferred embodiment of the state monitoring and real-time evaluation system for an energy storage power station described in this invention, the power station-level monitoring includes: setting a differential voltage detector at the main DC bus of the energy storage station to monitor the total voltage level of each group of batteries connected in parallel in real time; setting a current sensor in the main circuit to monitor the direction, amplitude and dynamic changes of the total current flow of the power station and the load; and installing an online insulation monitoring device between the main DC circuit of the energy storage and ground to detect the resistance value and monitor the voltage imbalance state of each electrode to ground.

[0018] As a preferred embodiment of the condition monitoring and real-time evaluation system for energy storage power stations described in this invention, the second model is constructed by extracting multi-dimensional feature variables from historical operating data, constructing a unified feature vector through feature engineering, combining known evaluation indicators, training the second model based on machine learning algorithms, and classifying and judging the state of the energy storage system.

[0019] To address the aforementioned technical problems, this invention provides the following technical solution: a method for state monitoring and real-time assessment of energy storage power stations, comprising: constructing a hierarchical multi-scale sensing system, deploying sensors at the cell level, battery pack level, and power station level to collect multi-source heterogeneous data from the energy storage power station; preprocessing the collected multi-source heterogeneous data and organizing it in a structured manner according to the spatial hierarchy to form a unified data representation; inputting the processed data into a first model and a second model to quantitatively predict and classify the state of the energy storage battery, and outputting assessment results; transmitting the assessment results to a safety early warning unit, which maps and judges the state assessment results according to preset risk classification rules, and outputs early warning level information and risk labels; and transmitting the assessment results and early warning results to an energy management system for visual display and early warning prompts of the real-time state.

[0020] The beneficial effects of this invention are as follows: by constructing a hierarchical multi-scale sensing system, covering the operating status and environmental information at the cell level, battery pack level, and power station level, the granularity and completeness of data acquisition are improved, providing high-quality basic data for system-level evaluation.

[0021] An evaluation method combining battery behavior mechanisms and artificial intelligence algorithms is introduced to achieve dynamic prediction and trend analysis of key indicators, thereby improving the accuracy, adaptability, and scenario generalization ability of the second model. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart of a condition monitoring and real-time evaluation method for an energy storage power station in Example 2. Detailed Implementation

[0024] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0025] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0026] Example 1, the first embodiment of the present invention, provides a condition monitoring and real-time assessment system for an energy storage power station, comprising:

[0027] The data acquisition unit is used to collect multi-source heterogeneous data from different levels of the energy storage power station.

[0028] The data processing and fusion unit is used to preprocess the collected multi-source heterogeneous data.

[0029] The status assessment unit constructs a first model to reflect the working status of the energy storage battery and a second model to assess the working status of the battery. Based on the first and second models, and combined with real-time data, it performs multi-dimensional analysis and assessment of the operating status of the energy storage power station.

[0030] The safety early warning unit is used to receive assessment results and issue safety alarms based on the assessment results and risk classification rules.

[0031] The communication and management interface unit is used to establish data interaction with the energy management system.

[0032] It should be noted that traditional monitoring methods are insufficient to meet the requirements for full-cycle, hierarchical, real-time, and intelligent management of system operation status. Furthermore, battery status assessment often remains at the level of single-point data judgment or empirical rules, lacking in-depth analytical capabilities based on models and structures, making it difficult to support the grid collaborative scheduling needs across multiple scenarios and time scales. In addition, currently, there is a lack of unified interfaces and linkage logic between subsystems in the status assessment and risk judgment stages, resulting in low data utilization efficiency, fragmented early warning results, and an overall insufficient level of intelligence.

[0033] Therefore, to address the aforementioned issues, a hierarchical, multi-scale sensing system is constructed using a data acquisition unit, a data processing and fusion unit, a status assessment unit, a safety early warning unit, and a communication and management interface unit. This system collects multi-source operational data, including voltage, current, temperature, gas, humidity, and insulation status, at the cell, battery pack, and power station levels. The data is then cleaned and fused before being input into the status assessment unit. Based on the battery's first model and data-driven algorithms, the unit assesses and analyzes status indicators such as SOC, SOH, internal resistance, thermal trend, and fault risk. The assessment results are then transmitted to the safety early warning unit, which performs risk assessment based on preset rules and outputs corresponding early warning levels and control responses. Finally, the early warning information and operational status are fed back to the energy management system via an interface, completing an integrated intelligent monitoring closed loop of "sensing—assessment—early warning—response."

[0034] Example 2, the second embodiment of the present invention, differs from the first embodiment in that: a condition monitoring and real-time assessment system for energy storage power stations further includes:

[0035] The data acquisition unit constructs a hierarchical, multi-scale sensing system that targets each operational level in the energy storage power station. By deploying monitoring equipment at different levels, it collects multi-source heterogeneous data.

[0036] The operational level includes cell-level monitoring, battery pack-level monitoring, and power station-level monitoring; the multi-source heterogeneous data includes operational parameters, environmental parameters, and structural safety parameters.

[0037] Specifically, cell-level monitoring includes installing voltage acquisition devices at both ends of each individual cell to collect cell voltage data; attaching thermistors to the cell casing to collect the casing surface temperature; installing gas sensors inside the packaging chamber to track the concentration of released gases; and recording the natural decreasing trend of the open-circuit voltage when the cell is in standby mode, estimating the dynamic resistance value of the internal short-circuit path based on the slight changes in steady-state current and voltage disturbances. The specific implementation is as follows:

[0038] Voltage sensor deployment: A precision voltage acquisition module is installed at both ends of each individual cell to support rapid sampling and temperature drift compensation, which is used to identify abnormal discharge and overcharging behavior of the cell.

[0039] Temperature sensor deployment: Thermistors / thermocouples are attached to the battery cell housing to collect the surface temperature of the housing. At the same time, redundant temperature sensors are deployed to achieve temperature cross-verification and anomaly correction.

[0040] Diagnostic mechanism based on self-discharge characteristics: When the cell is in a static / standby state, record the natural downward trend of its open-circuit voltage, and use this trend to compare with the standard discharge curve to determine whether there is a micro short-circuit channel (such as lithium dendrite penetration).

[0041] By modeling the minute changes in steady-state current and voltage disturbances, the dynamic resistance of the internal short-circuit path is estimated.

[0042] Internal gas signal acquisition: A miniature gas sensor is introduced into the encapsulation chamber to track the concentration of CO, CO2, and HF released by the destruction of the SEI membrane (Solid Electrolyte Interphase) and the decomposition of the electrolyte.

[0043] It should be noted that by deploying a high-precision sensor network and a micro-short-circuit identification mechanism at the cell level, refined health status perception and proactive protection of the smallest unit of the energy storage system are achieved. This effectively identifies early battery faults, dendrite growth, and localized heat source evolution trends, significantly improving the granularity of fault identification and the overall system safety level, and providing crucial data support for upper-level status assessment and early warning models.

[0044] To further explain, battery pack-level monitoring includes: configuring a differential voltage acquisition unit at the series terminal of the battery pack to monitor the overall voltage stability and voltage fluctuation characteristics during charging and discharging; deploying Hall current sensors in the busbar circuit to acquire the battery pack's input and output currents in real time, and dynamically estimating the charging and discharging power in conjunction with the total voltage signal; deploying temperature and humidity sensors inside the battery pack housing to collect temperature and humidity data; and installing corrosion sensors at the wiring terminals to identify insulation aging trends. The specific implementation is as follows:

[0045] Total voltage acquisition: A high-precision differential voltage acquisition module is configured at the series terminal of the battery pack, which supports millisecond-level data refresh cycles and is used to monitor the overall voltage stability and voltage fluctuation characteristics during charging and discharging.

[0046] Bus current detection: Hall current sensors are deployed in the bus circuit to acquire the input and output current of the battery pack in real time. Combined with the total voltage signal, the charging / discharging power is dynamically estimated, providing the basic input for the state-second model.

[0047] Wireless temperature and humidity sensor array: Temperature and humidity sensors are deployed in multiple locations inside the battery pack enclosure to achieve full coverage monitoring of the spatial temperature field. Abnormal temperature gradients can indicate safety hazards such as abnormal heat dissipation and localized overheating.

[0048] Salt and alkalinity and corrosive gas monitoring: By installing corrosion sensors or salt spray probes at the terminals, insulation aging trends in complex environments such as the seaside and high humidity can be identified in advance.

[0049] It should be noted that by setting up comprehensive operating parameter and environmental condition sensing devices at the battery pack level, a comprehensive monitoring system for the overall health status, operational consistency, and insulation performance of the battery pack has been constructed. This system can promptly identify inter-pack imbalances caused by temperature and voltage differences, thereby improving the operational stability and pack-level risk control capabilities of the energy storage system.

[0050] To further explain, the power station-level monitoring includes: installing a differential voltage detector at the main DC bus of the energy storage station to monitor the total voltage level of all battery groups connected in parallel in real time; installing a current sensor in the main circuit to monitor the direction, amplitude, and dynamic changes of the total current flow in the power station; and installing an online insulation monitoring device between the main DC circuit of the energy storage and ground to detect the resistance and simultaneously monitor the voltage imbalance between each electrode and ground. The specific implementation is as follows:

[0051] Total voltage detection: A high-voltage differential voltage detection module is installed at the main DC bus of the energy storage station to monitor the total voltage level of each group of batteries connected in parallel in real time. The data is used to analyze the power supply capacity of the power station and to determine whether there is system voltage instability or string interruption.

[0052] Busbar total current detection: A high current sensor is installed in the main circuit to monitor the direction, amplitude and dynamic changes of the total current flow in the power plant and the load. It is used to coordinate with the EMS (Energy Management System) to evaluate the power regulation capability under the current operating conditions.

[0053] Main circuit insulation resistance monitoring: An online insulation monitoring device is installed between the main DC circuit of energy storage and ground to detect the resistance value and monitor the voltage imbalance of each electrode to ground to determine whether there is a risk of offset potential or floating ground in the system.

[0054] It should be noted that by deploying a power station-level sensor network and establishing a safety monitoring mechanism, an overall perception of the operating status of the entire energy storage system, cross-group judgment, and fault isolation response are achieved, effectively ensuring the safe operation capability of the energy storage power station in unattended or remotely dispatched scenarios.

[0055] Furthermore, in the data processing and fusion unit, data from different levels and types is received. Through data interface adaptation and parsing, the original collected data is uniformly encoded and format converted to form a consistent internal data standard. Missing data is interpolated and completed, and abnormal data is identified and removed.

[0056] The data processing and fusion unit can also align and fuse data from different levels and sampling periods based on the timestamps, spatial identifiers and logical relationships of each data source to form a complete "running status snapshot" and ensure that the evaluation unit obtains a globally consistent data view at the same time.

[0057] Furthermore, a first model and a second model are constructed in the state assessment unit, and the first model and the second model are merged to assess the operating status of the energy storage power station based on real-time data.

[0058] In this embodiment of the application, the first model expresses the behavioral mechanism of the energy storage battery based on the preprocessed data, and is used to reflect the electrical mechanism. The specific model construction process includes the following steps S311-S315:

[0059] S311: Establish an equivalent circuit model, consisting of a voltage source representing the battery's open-circuit voltage; a series ohmic resistor reflecting the battery's intrinsic resistance (including internal resistance caused by conductors, electrolytes, etc.); and a parallel resistor-capacitor polarization branch representing the influence of electrochemical polarization, charge transfer, and double-layer behavior on the voltage response.

[0060] The output voltage of the overall model is the open-circuit voltage minus the sum of the ohmic voltage drop and the polarization voltage drop.

[0061] S312: Perform constant current charge-discharge experiments on the battery and allow it to stand still for a long time at each SOC (State of Charge) point until the terminal voltage stabilizes. Record the stable voltage and the corresponding SOC value, and plot and fit to form an "open circuit voltage-state of charge curve". This curve provides a voltage reference value for subsequent dynamic state estimation.

[0062] S313: Apply a current step or pulse signal (such as switching the current change from 0 to 1C) and record the corresponding voltage response curve.

[0063] S314: Calculate the series ohmic resistance using the instantaneous current change corresponding to the voltage drop abrupt change; fit the voltage time response curve using the slow voltage change phase; extract the polarization resistance and polarization capacitance according to the formula for the first-order capacitor charging and discharging process; the above fitting process is numerically optimized using the least squares method to make the simulated voltage curve as close as possible to the measured voltage curve.

[0064] S315: Validate the fitted model using experimental data under different magnifications or temperatures, compare the error between the model output voltage and the measured voltage. If the error exceeds the allowable range, refit the polarization parameters or adopt a piecewise modeling strategy.

[0065] In an optional embodiment, the first model can also be a model reflecting the thermal mechanism, and the specific model construction process includes the following steps S321-S326:

[0066] S321: The thermal equivalent model structure includes treating the battery as a physical body with heat capacity concentrated at a single point; the internal heat source comes from ohmic heat (generated by current passing through internal resistance); and heat is dissipated to the external environment through convection, conduction, and radiation.

[0067] S322: The basic assumptions include that the internal temperature distribution of the battery is approximately uniform (applicable to small single cells or transient temperature rise modeling); the heating rate is proportional to the square of the current; the ambient temperature is stable, and the heat dissipation path and heat conduction parameters can be considered constant in a short period of time.

[0068] S323: Based on the energy conservation equation, the thermal differential equation is expressed as follows:

[0069]

[0070] Where C represents the battery's heat capacity, which is the amount of heat absorbed per unit temperature change; T represents the overall temperature of the battery cell, which changes over time; and t represents time. Let R represent the rate of temperature change, I represent the current, and R represent the current. n Represented as internal resistance, the instantaneous ohmic resistance identified from the electrical model, T c Represented as battery temperature, T hRepresented as ambient temperature, R r It is expressed as thermal resistance, representing the overall thermal conduction path thermal resistance from the battery to the environment.

[0071] S324: Under constant ambient temperature, a constant current charge and discharge load is applied to the battery, and the battery temperature rise and fall curves over time are recorded; current, voltage and ambient temperature data are recorded simultaneously for calculating heat generation power.

[0072] S325: Calculate the power generated per unit time by multiplying the square of the current by the known internal resistance; based on the measured temperature curve, use the least squares method to fit the above thermal differential equation and solve for the two model parameters of heat capacity and thermal resistance.

[0073] S326: Simulate and test the constructed thermal model under other discharge conditions, and compare the error between the predicted temperature and the measured temperature; if the model error is greater than the set threshold (e.g., ±2℃), then correct it by iteratively optimizing the heat capacity and thermal resistance parameters.

[0074] In another alternative embodiment, the first model may also be a model reflecting the aging mechanism, specifically including:

[0075] Capacity decay model: used to predict the decline trend of usable capacity of a battery under different cycling conditions.

[0076] Internal resistance growth model: used to reflect the increase in impedance caused by electrochemical polarization and interface aging during battery use.

[0077] Furthermore, the second model is constructed by extracting multidimensional feature variables from historical operating data, constructing a unified feature vector through feature engineering, combining known evaluation indicators, and training the second model based on machine learning algorithms to classify and judge the state of the energy storage system.

[0078] In this embodiment of the application, the second model is constructed based on a model built using principal component analysis and support vector machine. The specific model construction process includes the following steps S331-S338:

[0079] S331: Collects data including more than ten operating parameters such as cell voltage, cell temperature, state of charge, current, battery pack bus current, charge / discharge rate, ambient temperature, and humidity.

[0080] S332: Data cleaning and processing includes outlier removal, missing value imputation, unit standardization, and standardization of all data to ensure that features participate in subsequent processing on an equal footing under the same unit of measurement.

[0081] S333: Construct a covariance matrix, analyze the correlation between features, and extract the top principal components whose cumulative contribution rate reaches a set threshold (e.g., 90%).

[0082] S334: Each principal component is divided into a linear combination of key operational features to form a principal component matrix, which serves as the input vector for the second model, reducing dimensional redundancy.

[0083] S335: Second model construction, including the following four layers:

[0084] Input layer: Receives vector data after principal component analysis; each input node corresponds to a principal component variable (such as the first principal component, the second principal component, etc.).

[0085] Kernel function mapping layer: maps the input vector from the original space to a high-dimensional feature space; in this embodiment, a radial basis kernel function is used to transform the nonlinear feature distribution into a high-dimensional linearly separable structure; this layer does not perform explicit feature expansion, but achieves implicit mapping through the inner product operation of the kernel function.

[0086] Margin optimization layer: finds the optimal classification hyperplane to maximize the classification margin between different categories; only support vectors near the boundary are retained to define the hyperplane, thereby compressing the model size; if there is noise or a small number of mislabeled data, a penalty coefficient is introduced to balance classification accuracy and model tolerance.

[0087] Output layer: Outputs the current operating status category label of the battery; the label includes three categories: "normal operation", "minor abnormality" and "serious abnormality", and the output is in one-hot encoded form; at the same time, it outputs the classification confidence value of each category for subsequent warning level mapping.

[0088] S336: Training data construction includes constructing a training set by combining the principal component analysis results with manually labeled state category labels, and then training the second model.

[0089] S337: Parameter tuning and cross-validation include using a grid search method to adjust the penalty coefficient and kernel function width; and using five-fold cross-validation to test the generalization ability of the second model and prevent overfitting.

[0090] S338: Deploys the trained model to the evaluation engine, receives real-time data collected by the monitoring system, transforms it through principal component analysis, inputs it into the support vector machine model, and outputs the real-time running status category and its confidence level for the safety early warning module to call.

[0091] In an optional embodiment, the second model can also be constructed based on K-means clustering and random forest, and the specific model construction process includes the following steps S341-S3410:

[0092] S341: Collect power data and preprocess the data.

[0093] S342: Clustering objective setting, to identify the potential distribution structure of the running data in an unsupervised manner, and to divide a large amount of unlabeled data into several categories based on similarity (e.g., normal high load, mild high temperature, abnormal discharge, etc.).

[0094] S343: Specify the number of cluster categories, initialize cluster centers, and assign them according to the distance of the samples to each center. Continuously update the cluster centers until convergence, and obtain the final cluster labels.

[0095] S344: Combining expert knowledge or manual annotation, the clustering results are mapped to the corresponding operating state categories, which serve as the target variables for subsequent classification models. These categories can be divided into three states: "normal operation", "boundary operation", and "severe anomaly".

[0096] S345: Second model construction, including the following three layers:

[0097] Input layer: Receives the feature vector corresponding to each sample after K-means clustering; each input node represents an original monitoring parameter (such as temperature, current, voltage change rate, etc.).

[0098] Decision tree ensemble layer: contains multiple classification decision trees trained in parallel (usually dozens to hundreds); each decision tree uses only a random subset of samples during training and randomly selects some variables from the features for splitting judgment; each tree determines which class state a sample belongs to through several decision nodes, forming independent voting opinions.

[0099] Output layer: Collects the classification results of all decision trees; uses majority voting to finally output the running state category of the sample; if necessary, the voting ratio of each category can be output at the same time as a state confidence index for subsequent safety warning level classification.

[0100] S346: The training set construction includes using K-means clustered and labeled data as training samples, containing input features and state category labels.

[0101] S347: The model training process includes constructing multiple decision trees, each tree being trained using different subsets of samples and features, and each tree recursively constructing nodes to split feature variables with the goal of minimizing classification error.

[0102] S348: Model tuning and validation includes adjusting hyperparameters such as the number of trees, maximum depth, and minimum number of sample splits, and using cross-validation to evaluate classification accuracy, overfitting risk, and generalization ability.

[0103] S349: Deploy the trained random forest model to the state assessment unit. The model automatically receives monitoring data from the energy storage system and completes the identification of the operating status in real time.

[0104] S3410: Outputs the current sample's running state category in real time, and outputs the voting ratio of each category, which can be used to quantify the degree of state abnormality.

[0105] In another optional embodiment, the second model can also be constructed based on grey relational analysis and neural network algorithms. The specific model construction process includes the following steps: First, historical data of the energy storage system under typical operating cycles is collected, including key variables such as temperature, current, voltage, internal resistance, and state of charge. Then, grey relational analysis is used to calculate the correlation between each variable and system performance degradation, and highly correlated feature variables are selected for modeling input. During the model construction phase, a three-layer feedforward neural network is designed. The number of nodes in the input layer corresponds to the number of selected features. Two hidden layers are set, containing 64 and 32 neurons respectively. The activation function is a modified linear unit function, and the output layer is a state category probability value or a continuous state score. During the training phase, the backpropagation algorithm is used to optimize the weights, historical labeled data is used for supervised learning, and hyperparameters are adjusted through cross-validation. After training, the model is used to identify the current operating state in real time and predict the state evolution trend.

[0106] It should be noted that by integrating battery behavior mechanisms and operational data, a state assessment system that combines interpretability and real-time performance has been constructed. This solution can not only accurately characterize the current performance of the energy storage system, but also flexibly switch assessment paths under different grid application scenarios, achieving personalized and adaptable assessments, and improving the scheduling decision support capabilities and operational safety of energy storage power stations.

[0107] Furthermore, the safety early warning unit is used to receive the operating status data and risk assessment results output by the status assessment unit, and classify and judge the assessment results based on the preset risk classification rules, and output the early warning level and linkage response signal.

[0108] In this embodiment of the application, the preset risk classification rule adopts a scoring method based on rule thresholds, including the following steps S411-S45:

[0109] S411: The safety warning unit receives quantitative assessment results from the condition assessment unit, including battery state of charge, health status, temperature rise trend, short circuit risk score and other key indicators.

[0110] S412: For each key indicator, the early warning unit pre-sets fixed threshold standards. For example, when the temperature rise rate exceeds the preset value, it is judged as "high risk"; when the state of charge is lower than the safe lower limit, it is judged as "slight abnormality", etc.

[0111] Each parameter is independently assessed for a threshold, generating a single risk result.

[0112] S413: Based on the preset rule matrix, combine the individual risk results: if any key parameter reaches "high risk", the overall warning level is set to "severe abnormality" or "emergency"; otherwise, if multiple parameters are "mildly abnormal", the overall warning level is set to "minor abnormality"; the "if...then..." rule decision is adopted here to ensure that the warning judgment is simple and intuitive.

[0113] S414: Based on the rule mapping results, output the specific warning level, such as "normal", "minor abnormality", "serious abnormality" or "emergency alarm".

[0114] S415: Based on different warning levels, trigger corresponding linkage response measures: For the "minor anomaly" level, it can prompt maintenance personnel to observe and display warning information on the control interface; for the "serious anomaly" level, it automatically issues partial load reduction adjustment instructions and sends remote alarm information to the monitoring center; for the "emergency alarm" level, it immediately triggers emergency power outage, starts backup power supply, links the fire protection system, and notifies on-site personnel to carry out emergency handling.

[0115] In an optional embodiment, the preset risk grading rule may also employ a multi-factor weighted scoring method, including the following steps S421-S424:

[0116] S421: The safety warning unit receives quantitative assessment results from the condition assessment unit, including battery state of charge, health status, temperature rise trend, short circuit risk score and other key indicators.

[0117] S422: Based on actual operating experience and engineering risk requirements, assign weights to each key indicator. For example, the temperature rise trend has a higher weight, while the state of charge and internal resistance change have a lower weight. The weight allocation is predetermined and fixed in the early warning algorithm.

[0118] S423: Calculate a comprehensive risk score by using the risk scores of each indicator and their corresponding weights. This score is the weighted sum of all indicators and is used to reflect the overall operational risk level.

[0119] S424: Divide the comprehensive risk score into several segments. When the comprehensive risk score is below the low threshold, the overall status is judged as "normal"; when the score is in the middle range, it is judged as "slightly abnormal"; when the score is above the high threshold, it is judged as "severely abnormal" or "emergency alarm". This step uses a pre-set score range table to achieve hierarchical mapping.

[0120] Generate a linkage response signal: Based on the comprehensive risk score and its corresponding warning level, output the corresponding linkage command:

[0121] If the assessment result is in the "minor anomaly" area, an alert will be sent to the monitoring center and the operation and maintenance personnel will be notified to pay attention. If the level reaches "serious anomaly" or "emergency alarm", a safety protection measure instruction will be automatically issued, such as partial load reduction, starting backup power, or directly performing emergency shutdown operation, and a report will be sent to the relevant emergency center at the same time.

[0122] In another optional embodiment, the preset risk classification rules can also employ a probabilistic early warning method based on Bayesian inference. Specifically, this includes: First, statistically calculating the prior probabilities of each early warning level (e.g., normal, abnormal, severe abnormal) based on historical data, and constructing a conditional probability model by combining key indicators output by the state assessment unit (e.g., temperature rise rate, internal resistance change, etc.). Then, using Bayes' theorem, the posterior probability of the current state belonging to each risk level is calculated in real time. During the judgment phase, thresholds are set to classify levels based on the posterior probability values ​​of each level. For example, when the posterior probability of the "severe abnormal" level exceeds the set threshold, an early warning response for that level is triggered. Finally, the system generates control strategies based on the judgment results, including alarms, power regulation, or shutdown protection measures. This method can comprehensively handle uncertainties and achieve dynamic early warning for fuzzy states, making it suitable for energy storage power station operation scenarios where abnormal boundaries are unclear and state evolution is complex.

[0123] It should be noted that by decoupling the safety early warning unit and establishing a logical connection between the status assessment unit and the safety early warning unit, the early warning module focuses on attribution judgment and graded response of the assessment results, while the assessment module focuses on multi-parameter analysis and modeling prediction. This avoids duplication of assessment strategies and improves the clarity, scalability and engineering integration efficiency of the system modules.

[0124] Furthermore, the communication and management interface unit is used to establish data interaction with the energy management system.

[0125] In this embodiment of the application, the communication and management interface unit transmits data based on the data uploading method of industrial Ethernet, including the following steps S511-S514:

[0126] S511: The status assessment unit and the safety early warning unit generate structured output, including assessment data such as the current health status index, risk assessment level, key indicator values, and timestamps.

[0127] S512: The communication and management interface unit encapsulates the above evaluation data into standardized data frames via industrial Ethernet and sends them to the energy management system backend periodically or triggeredly.

[0128] S513: After receiving the data, the energy management system updates the status dashboard in real time and links it with automatic scheduling strategies (such as load adjustment, maintenance plan generation, etc.).

[0129] S514: The remote monitoring platform can receive the same data, enabling unified visual supervision and historical data traceability across multiple stations.

[0130] In an optional embodiment, the communication and management interface unit can also transmit data based on a wireless serial communication data uploading method, including the following steps S521-S524:

[0131] S521: In cabling-constrained sites, status assessment results and early warning results are encapsulated into compact messages through the serial communication interface of the communication module.

[0132] S522: Messages are sent to the nearest data central node via the wireless transceiver module and then aggregated and sent to the central controller.

[0133] S523: The main controller forwards the received evaluation data to a remote server or cloud management platform to achieve linkage between edge computing and cloud monitoring.

[0134] S524: Back-end administrators can handle the situation remotely or manually based on the warning level.

[0135] In another optional embodiment, the communication and management interface unit can also transmit data based on power line carrier communication technology. Specifically, this includes: first, connecting each monitoring terminal (such as a battery management unit, temperature sensor, etc.) to a node module with power line communication functionality; the node module then superimposes the data signal onto the power line using a modulation method (such as orthogonal frequency division multiplexing); the communication and management interface module, acting as the master control node, parses the data frames sent by each sub-node using power line demodulation to achieve periodic acquisition and status synchronization of the energy storage system's operating parameters. This communication method supports point-to-multipoint networking and features strong anti-interference capabilities, low deployment costs, and high adaptability.

[0136] The received data is parsed, cached, and forwarded within the main control module, and can be uploaded to the energy management system or cloud platform via local bus or remote interface to achieve unified monitoring and scheduling at the upper level.

[0137] Example 3, referring to Figure 1 This is the third embodiment of the present invention, which differs from the previous two embodiments in that: a method for condition monitoring and real-time assessment of an energy storage power station includes:

[0138] S1: By constructing a hierarchical multi-scale sensing system, sensors are deployed at the cell level, battery pack level and power station level to collect multi-source heterogeneous data from energy storage power stations.

[0139] S2: Preprocess the collected multi-source heterogeneous data and organize it in a structured manner according to the spatial hierarchy to form a unified data representation.

[0140] S3: Input the processed data into the first and second models to quantitatively predict and classify the state of the energy storage battery, and output the evaluation results.

[0141] S4: Transmit the assessment results to the safety early warning unit, which maps and judges the status assessment results according to the preset risk classification rules, and outputs early warning level information and risk labels.

[0142] S5: Transmits the assessment results and early warning results to the energy management system to visualize the real-time status and provide early warning prompts.

[0143] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A condition monitoring and real-time assessment system for energy storage power stations, characterized in that: include, The data acquisition unit is used to collect multi-source heterogeneous data from different levels of the energy storage power station. The data processing and fusion unit is used to preprocess the collected multi-source heterogeneous data; The status assessment unit constructs a first model to reflect the working status of the energy storage battery and a second model to assess the working status of the battery. Based on the first and second models, and combined with real-time data, it performs multi-dimensional analysis and assessment of the operating status of the energy storage power station. The safety early warning unit is used to receive assessment results and issue safety alarms based on the assessment results and risk classification rules; The communication and management interface unit is used to establish data interaction with the energy management system.

2. The condition monitoring and real-time assessment system for an energy storage power station as described in claim 1, characterized in that: The data acquisition unit constructs a hierarchical, multi-scale sensing system that targets each operational level in the energy storage power station. By deploying monitoring equipment at different levels, it collects multi-source heterogeneous data.

3. The condition monitoring and real-time assessment system for an energy storage power station as described in claim 2, characterized in that: The first model expresses the behavioral mechanism of energy storage battery operation based on preprocessed data. The first model is fused with the second model to evaluate the operating status of the energy storage power station based on real-time data.

4. The condition monitoring and real-time assessment system for an energy storage power station as described in claim 3, characterized in that: The safety early warning unit is used to receive the operating status data and risk assessment results output by the status assessment unit, and classify and judge the assessment results based on the preset risk classification rules, and output the early warning level and linkage response signal.

5. The condition monitoring and real-time assessment system for an energy storage power station as described in claim 4, characterized in that: The operational levels include cell-level monitoring, battery pack-level monitoring, and power station-level monitoring. The multi-source heterogeneous data includes operating parameters, environmental parameters, and structural safety parameters.

6. The condition monitoring and real-time assessment system for an energy storage power station as described in claim 5, characterized in that: The cell-level monitoring includes installing voltage acquisition devices at both ends of each individual cell to collect cell voltage data. A thermistor is attached to the cell casing to collect the surface temperature of the casing; a gas sensor is installed in the packaging chamber to track the concentration of the released gas. When the battery cell is in standby mode, the natural downward trend of the open-circuit voltage is recorded. Based on the slight changes in steady-state current and voltage disturbance, the dynamic resistance value of the internal short-circuit path is estimated.

7. The condition monitoring and real-time assessment system for an energy storage power station as described in claim 6, characterized in that: The battery pack-level monitoring includes configuring a differential voltage acquisition unit at the series terminal of the battery pack to monitor the overall voltage stability and voltage fluctuation characteristics during charging and discharging. Hall current sensors are deployed in the bus circuit to acquire the input and output current of the battery pack in real time, and to dynamically estimate the charging and discharging power in combination with the total voltage signal. Temperature and humidity sensors are deployed inside the battery pack enclosure to collect temperature and humidity data inside the enclosure. Corrosion sensors are installed at the terminals to identify the trend of insulation aging.

8. The condition monitoring and real-time assessment system for an energy storage power station as described in claim 7, characterized in that: The power station-level monitoring includes setting up a differential voltage detector at the main DC bus of the energy storage station to monitor the total voltage level of each group of batteries connected in parallel in real time. A current sensor is installed in the main circuit to monitor the direction and magnitude of the total current flow in the power plant and the dynamic changes in the load. An online insulation monitoring device is installed between the main DC circuit of the energy storage and ground to detect the resistance value and monitor the voltage imbalance between each electrode and ground.

9. The condition monitoring and real-time assessment system for an energy storage power station as described in claim 8, characterized in that: The second model is constructed by extracting multidimensional feature variables from historical operating data, constructing a unified feature vector through feature engineering, combining known evaluation indicators, and training the second model based on machine learning algorithms to classify and judge the state of the energy storage system.

10. A method for condition monitoring and real-time assessment of an energy storage power station, using a condition monitoring and real-time assessment system for an energy storage power station as described in any one of claims 1 to 9, characterized in that: include, By constructing a hierarchical, multi-scale sensing system, sensors are deployed at the cell level, battery pack level, and power station level to collect multi-source heterogeneous data from energy storage power stations. The collected multi-source heterogeneous data is preprocessed and organized in a structured manner according to the spatial hierarchy to form a unified data representation; The processed data is input into the first and second models to quantitatively predict and classify the state of the energy storage battery, and the evaluation results are output. The assessment results are transmitted to the safety early warning unit, which maps and judges the status assessment results according to the preset risk classification rules, and outputs early warning level information and risk labels. The assessment results and early warning results are transmitted to the energy management system to visualize the real-time status and provide early warning prompts.

Images