A multi-parameter risk grading alarm method for an energy storage system

Abstract

The application relates to the technical field of hierarchical alarm, and particularly discloses a multi-parameter risk hierarchical alarm method for an energy storage system, which comprises the following steps: step S1: determining all primary risk parameters, obtaining historical fault data to obtain risk scenes, and determining abnormal risk parameters under each risk scene and the cause-effect relationship between the abnormal risk parameters and the scenes; step S2: taking the primary parameters and the risk scenes as leaf nodes and intermediate nodes respectively, determining the leaf node risk threshold based on historical data, and configuring the transmission rate of the risk signals along the directed edges; step S3: establishing a root node, establishing upper-layer directed edges between the intermediate nodes and the root node, and outputting the upper-layer risk signals and transmitting the upper-layer risk signals to the root node after the intermediate nodes receive the risk signals; and step S4: manually determining the risk sensitivity of each intermediate node, obtaining a risk grade index according to all the received upper-layer risk signals and the corresponding sensitivities by the root node, and thus determining the risk grade of the energy storage system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of hierarchical alarm, in particular to a multi-parameter risk hierarchical alarm method for energy storage system. BACKGROUND

[0002] With the large-scale application and complex development of energy storage systems, the risks faced in the operation process present diversified characteristics, including but not limited to overcharging, overdischarging, temperature abnormalities, internal resistance increases, insulation failures, and imbalances between modules. These risk factors are coupled and dynamically changing, making it difficult for traditional single-parameter threshold judgment mode to meet the precise control requirements.

[0003] In the prior art, the current mainstream scheme adopts the strategy of setting fixed thresholds for each risk factor, which not only requires storing a large number of preset parameters such as voltage upper and lower limits, temperature difference ranges, and SOC fluctuation intervals, but also causes significant memory occupation problems due to multi-dimensional data parallel processing. Especially in large-scale battery array scenarios, real-time data collection and comparison of hundreds of monitoring points further exacerbate the consumption of computing resources. In addition, static thresholds cannot adapt to changes in working conditions, such as fluctuations in environmental temperature and humidity, and performance degradation caused by increased cycle times, which can easily cause false alarms or missed alarms, reducing the reliability and economy of the system.

[0004] Therefore, there is an urgent need to develop a new risk warning method that can integrate multi-parameter correlation characteristics, dynamically optimize hierarchical strategies, and efficiently utilize storage space to achieve dual improvement of safety and efficiency of energy storage systems. SUMMARY

[0005] The purpose of the present application is to provide a multi-parameter risk hierarchical alarm method for energy storage systems, which solves the following technical problems.

[0006] The purpose of the present application can be achieved by the following technical solutions:

[0007] A multi-parameter risk hierarchical alarm method for energy storage systems, comprising the following steps:

[0008] Step S1: Determine all primary risk parameters of the energy storage system, and obtain historical failure data of the energy storage system to obtain several risk scenarios; for any risk scenario, determine all abnormal risk parameters under the risk scenario, and record that each abnormal risk parameter has a causal relationship with the risk scenario;

[0009] Step S2: each primary risk parameter is recorded as a leaf node, each risk scenario is recorded as an intermediate node, for any leaf node, a directed edge is established between the intermediate nodes which have causal relationship with the leaf node; a risk threshold of the leaf node is determined based on the historical failure data, the leaf node generates a risk signal, and the transmission rate of the risk signal along the directed edge is configured according to the risk threshold, so that when the physical value of the leaf node reaches the risk threshold, the risk signal reaches the intermediate node;

[0010] Step S3: a root node is established, for any intermediate node, an upper layer directed edge is established between the root node and the intermediate node, if the intermediate node receives a risk signal, the intermediate node outputs an upper layer risk signal, and the upper layer risk signal is transmitted to the root node along the upper layer directed edge;

[0011] Step S4: the risk sensitivity of each intermediate node is determined artificially, the risk level index is obtained by the root node according to all the received upper layer risk signals and the risk sensitivity of the corresponding intermediate nodes, and the risk level of the energy storage system is determined according to the risk level index.

[0012] As a further scheme of the present application: the primary risk parameters include voltage, current, internal resistance and temperature; the historical failure data includes the failure data of the energy storage system at each time of failure in the historical operation period, the failure data includes the physical values of all primary risk parameters and environmental data when the failure occurs, the environmental data includes air temperature, wind power, air temperature and sunshine intensity, and the type of each failure of the energy storage system is recorded as a risk scenario.

[0013] As a further scheme of the present application: the process of determining all abnormal risk parameters in the risk scenario includes:

[0014] The risk scenario is recorded as a target risk scenario, the change feature of the target risk scenario is extracted based on a machine learning model; all failure data in the historical failure data which are in the target risk scenario are screened out, all primary risk parameters belonging to the same failure in all failure data are recorded as a combination; for any primary risk parameter in any group of primary risk parameters, whether the primary risk parameter and the change feature exist correlation is judged based on Pearson correlation coefficient method, if yes, the combination in which the primary risk parameter is located is recorded as a relevant combination, and the correlation proportion P of the primary risk parameter is obtained, n is the number of relevant combinations of the primary risk parameter, N is the total number of combinations, if the correlation proportion of the primary risk parameter is greater than a preset proportion threshold, the primary risk parameter is recorded as an abnormal risk parameter.

[0015] As a further scheme of the present application: the process of determining the risk threshold of the leaf node based on the historical failure data includes:

[0016] acquire any intermediate node in causal relationship with the leaf node, record the risk scenario corresponding to the intermediate node as a to-be-tested scenario, filter out all fault data in the historical fault data under the to-be-tested scenario and record as fault trigger data;And acquire normal operation data in the historical operation period, the normal operation data includes physical values of all primary risk parameters and environment data when the energy storage system does not occur fault;

[0017] Take the environment data and the physical values of each primary risk parameter as independent variables, and take the operation state of the energy storage system as dependent variable, establish an initial logistic regression model, the operation state of the energy storage system includes fault and normal;Take the normal operation data as positive sample data, and take the fault trigger data as negative sample data, input the positive sample data and negative sample data into the initial logistic regression model, train the initial logistic regression model, and obtain a logistic regression model;

[0018] acquire current environment data, record as current environment data, input the current environment data into the logistic regression model as a constraint condition, obtain a limited logistic regression model;For any primary risk parameter, exhaust several physical value gradient values of the primary risk parameter, and input them into the limited logistic regression model in turn to obtain output results;For any physical value gradient value of the primary risk parameter, repeat the operation of inputting into the limited logistic regression model several times to obtain several output results, acquire the proportion of the operation state being fault in all output results, if the proportion is greater than or equal to 90%, record the physical value gradient value as the abnormal physical value of the primary risk parameter;Select the minimum value from all abnormal physical values of the primary risk parameter, record as the risk threshold of the primary risk parameter, that is, the risk threshold of the leaf node corresponding to the primary risk parameter.

[0019] As a further scheme of the application: the process of configuring the transmission rate of the risk signal along the directed edge according to the risk threshold comprises:

[0020] Set the starting physical value Q of the leaf node, acquire the length L of the directed edge, and acquire the risk threshold Qt of the current leaf node, obtain the transmission rate v of the risk signal along the directed edge L / (Qt-Q);If Qt≤Q, the transmission rate of the risk signal along the directed edge is directly recorded as 0.

[0021] As a further scheme of the application: the direction of the directed edge is from the leaf node to the intermediate node, and the direction of the upper directed edge is from the intermediate node to the root node.

[0022] As a further scheme of the application: the process of the risk level index comprises:

[0023] The risk level index wherein K i , wherein K represents the risk sensitivity of the intermediate node corresponding to the i-th upper layer risk signal received by the root node, num is the total number of the i-th upper layer risk signal received by the root node, i [1, num] and i is a positive integer.

[0024] As a further scheme of the present application: the process of determining the risk level of the energy storage system according to the risk level index comprises:

[0025] The number of preset risk levels is Z, and Z-1 risk index thresholds are set, and a plurality of risk index intervals {[-∞, R1), [R1, R2),..., [R Z-2 , R Z-1 ), [R Z-1 , +∞]} are determined according to the risk index thresholds, wherein R Z-1 represents the Z-1th risk index threshold; the risk levels are sequentially assigned to the risk index intervals, and the risk index interval in which the real-time risk level index is located is determined, which is denoted as the current risk index interval, and the risk level corresponding to the current risk index interval is denoted as the current risk level of the energy storage system.

[0026] The present application has the following beneficial effects:

[0027] The present application can accurately locate the risk inducement by clearly defining the causal relationship between the primary risk parameter and the risk scenario; the leaf node risk threshold is determined based on the historical failure data, and the risk signal transmission rate is configured, which not only ensures the objectivity and accuracy of risk identification, but also enables the risk signal to be efficiently and timely transmitted to the corresponding risk scenario; with the aid of the hierarchical architecture of the root node and the upper layer directed edge, the system-level aggregation and conduction of risks are realized; the risk level index is calculated in combination with the intermediate node risk sensitivity determined by artificial, which not only integrates professional experience, but also clearly defines the system risk level through quantitative means, and can provide scientific and efficient support for the safe operation monitoring, early failure warning and operation and maintenance decision-making of the energy storage system, and is helpful to improve the overall reliability and risk management level of the energy storage system. BRIEF DESCRIPTION OF DRAWINGS

[0028] The present application will be further described below with reference to the accompanying drawings.

[0029] Figure 1 is a step schematic diagram of a multi-parameter risk grading alarm method of an energy storage system of the present application;

[0030] Figure 2 is a schematic diagram of the architecture of a multi-parameter risk grading alarm method of an energy storage system of the present application. DETAILED DESCRIPTION

[0031] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described, obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the protection scope of the present application.

[0032] Please refer to Figure 1 As shown in the drawings, the present application is a kind of energy storage system multi-parameter risk classification alarm method, comprising the following steps:

[0033] Step S1: determine all primary risk parameters of the energy storage system, and obtain historical failure data of the energy storage system, obtain several risk scenarios;For any risk scenario, determine all abnormal risk parameters under the risk scenario, record each abnormal risk parameter and the risk scenario exist causal relationship;

[0034] Specifically, the energy storage system includes battery cluster, converter, energy management system, temperature control device, all primary risk parameters include battery temperature, charge and discharge circuit current, battery state of charge SOC, electrolyte concentration, system bus voltage, equipment running vibration amplitude;From the operation and maintenance database of the energy storage system, sensor real-time monitoring history record, fault alarm log and other multi-source channels, comprehensive obtain historical failure data, and obtain several typical risk scenarios by data clustering method, for example, battery thermal runaway scenario, electrolyte leakage scenario, charge and discharge overcurrent scenario, voltage abnormal fluctuation scenario, equipment mechanical failure scenario;

[0035] As a preferred embodiment of the present application, the primary risk parameters include voltage, current, internal resistance and temperature;The historical failure data includes the failure data of the energy storage system when each failure occurs in the historical operation cycle, the failure data includes the physical values of all primary risk parameters when the failure occurs and environmental data, the environmental data includes air temperature, wind power, air temperature and sunshine intensity, and the type of each failure of the energy storage system is recorded as a risk scenario;

[0036] As a preferred embodiment of the present application, the process of determining all abnormal risk parameters under the risk scenario includes:

[0037] record the risk scenario as a target risk scenario, extract a change feature of the target risk scenario based on a machine learning model, filter all fault data in the historical fault data under the target risk scenario, record all primary risk parameters belonging to the same fault in all fault data as a combination, determine whether a primary risk parameter in any combination of primary risk parameters and the change feature exist correlation based on a Pearson correlation coefficient method, if yes, record the combination in which the primary risk parameter is located as a relevant combination, and obtain a relevant proportion P of the primary risk parameter n / N, wherein n is the number of relevant combinations of the primary risk parameter, and N is the total number of combinations, if the relevant proportion of the primary risk parameter is greater than a preset proportion threshold, record the primary risk parameter as an abnormal risk parameter;

[0038] The setting range of the proportion threshold is [50%, 70%];

[0039] As a preferred embodiment of the present application, the process of extracting the change feature of the target risk scenario based on the machine learning model comprises:

[0040] The machine learning model is specifically a feature extractor based on unsupervised learning, the machine learning model takes all fault data under the target risk scenario as a training set, and compresses and reconstructs high-dimensional and redundant fault data into a low-dimensional feature vector through a hybrid architecture of a long short-term memory network and a self-encoder, and records the low-dimensional feature vector as the change feature of the target risk scenario.

[0041] It can be understood that the change feature is a data representation of the core dynamic mode of the target risk scenario autonomously learned by the model under the training target of reconstructing input data with minimum information loss or capturing the inherent invariance of data, and each numerical element in the dimension collectively constitutes the change feature representing the macro state of the risk scenario.

[0042] Step S2: record each primary risk parameter as a leaf node, record each risk scenario as an intermediate node, for any leaf node, establish a directed edge between the intermediate nodes having a causal relationship with the leaf node, determine a risk threshold of the leaf node based on the historical fault data, the leaf node generates a risk signal, and configure the transmission rate of the risk signal along the directed edge according to the risk threshold, so that when the physical value of the leaf node reaches the risk threshold, the risk signal reaches the intermediate node.

[0043] Specifically, an early warning mechanism based on a dynamic risk propagation network is constructed, primary risk parameters such as voltage, current and temperature are set as leaf nodes, risk scenarios such as thermal runaway and insulation failure are set as intermediate nodes, and causal relationships are established through directed edges; a risk threshold calculation model that is self-adaptive to environmental temperature and equipment aging degree is established based on historical fault data, and through an intelligent transmission mechanism, the signal transmission rate is dynamically configured according to the risk emergency level;

[0044] As a preferred embodiment of the present application, the direction of the directed edge is from the leaf node to the intermediate node;

[0045] As a preferred embodiment of the present application, the process of determining the risk threshold of the leaf node based on the historical fault data comprises:

[0046] Any intermediate node having a causal relationship with the leaf node is obtained, the risk scenario corresponding to the intermediate node is recorded as a to-be-tested scenario, all fault data in the historical fault data that are in the to-be-tested scenario are screened out and recorded as fault trigger data; and normal operation data in a historical operation period is obtained, the normal operation data comprising physical values of all primary risk parameters and environmental data when the energy storage system does not fail;

[0047] An initial logistic regression model is established with environmental data and physical values of each primary risk parameter as independent variables and the operating state of the energy storage system as a dependent variable, the operating state of the energy storage system comprising failure and normal; the normal operation data is taken as positive sample data and the fault trigger data is taken as negative sample data, the positive sample data and the negative sample data are input into the initial logistic regression model, the initial logistic regression model is trained, and a logistic regression model is obtained;

[0048] Current environmental data is obtained and recorded as current environmental data, the current environmental data is input into the logistic regression model as a constraint condition to obtain a limited logistic regression model; for any primary risk parameter, a plurality of physical value gradient values of the primary risk parameter are enumerated and input into the limited logistic regression model in turn to obtain output results; for any physical value gradient value of the primary risk parameter, the operation of inputting into the limited logistic regression model is repeated several times to obtain a plurality of output results, the proportion of the operating state being failure in all output results is obtained, if the proportion is greater than or equal to 90%, the physical value gradient value is recorded as an abnormal physical value of the primary risk parameter; the minimum value is selected from all abnormal physical values of the primary risk parameter and recorded as a risk threshold of the primary risk parameter, that is, a risk threshold of the leaf node corresponding to the primary risk parameter;

[0049] It can be understood that, based on the intermediate node having a causal relationship with the leaf node, the to-be-tested risk scenario is determined; the fault trigger data under the scenario is extracted from the historical fault data, and complete data during normal operation of the system is collected as a positive sample; a logistic regression model considering the coupling of multiple factors such as environmental temperature, humidity, and system aging degree is established, and a basic discrimination model is obtained through positive and negative sample training; the current environmental data is introduced as a constraint condition to form a limited model, and an adaptive gradient search algorithm is used to quickly locate the critical physical value that makes the fault probability reach 90%;

[0050] As a preferred embodiment of the application, the process of configuring the transmission rate of the risk signal along the directed edge according to the risk threshold value comprises:

[0051] The starting physical value Q of the leaf node is set, the length L of the directed edge is obtained, and the risk threshold value Qt of the current leaf node is obtained, so as to obtain the transmission rate v of the risk signal along the directed edge, which is L / (Qt-Q); if Qt≤Q, the transmission rate of the risk signal along the directed edge is directly recorded as 0;

[0052] Specifically, when the monitoring value Q is far away from the threshold value Qt, the transmission rate v is low, indicating that the system is in a safe state and only needs to be monitored regularly; when the monitoring value Q gradually approaches the risk threshold value Qt, the transmission rate v will significantly increase;

[0053] Step S3: establishing a root node, for any intermediate node, establishing an upper directed edge between the intermediate node and the root node, if the intermediate node receives a risk signal, the intermediate node outputs an upper layer risk signal, and the upper layer risk signal is transmitted to the root node along the upper layer directed edge;

[0054] As a preferred embodiment of the application, the direction of the upper directed edge is pointed from the intermediate node to the root node;

[0055] It should be noted that the root node is the highest decision point, responsible for comprehensive evaluation of all activated risk scenarios, realizing judgment from local risk perception to global security situation; the upper directed edge forms a dedicated channel for the convergence of risk information from the intermediate node to the root node, ensuring that important risk information can be orderly and reliably transmitted to the decision center; after receiving the lower layer signal, the intermediate node outputs an upper layer risk signal in a unified format, providing standardized input for the comprehensive risk assessment of the root node;

[0056] Step S4: determining the risk sensitivity of each intermediate node based on artificial determination, and the root node obtains a risk level index according to all received upper layer risk signals and the risk sensitivity of the corresponding intermediate nodes; determining the risk level of the energy storage system according to the risk level index;

[0057] Specifically, based on expert experience and historical failure analysis results, each intermediate node is given a risk sensitivity weight that can be dynamically adjusted, wherein historical occurrence rate, consequence severity and diffusion speed of different risk scenarios are considered; the root node calculates a quantitative risk level index by a weighted aggregation algorithm, by comprehensively considering all input upper risk signals and their corresponding sensitivity; based on the index, the current risk level of the energy storage system is dynamically determined through a preset multi-level threshold interval with a buffer zone;

[0058] As a preferred embodiment of the present application, the process of the risk level index comprises:

[0059] The risk level index , wherein K i represents the risk sensitivity of the intermediate node corresponding to the i-th upper risk signal received by the root node, num is the total number of the i-th upper risk signal received by the root node, i∈[1, num] and i is a positive integer;

[0060] As a preferred embodiment of the present application, the process of determining the risk level of the energy storage system according to the risk level index comprises:

[0061] The number of preset risk levels is Z, then Z-1 risk index thresholds are set, and a plurality of risk index intervals {[-∞, R1), [R1, R2),..., [R Z-2 , R Z-1 ), [R Z-1 , +∞]} are determined according to the risk index thresholds, wherein R Z-1 represents the Z-1th risk index threshold; the risk levels are sequentially assigned to the risk index intervals, and the risk index interval in which the real-time risk level index is located is determined, which is recorded as the current risk index interval, and the risk level corresponding to the current risk index interval is recorded as the current risk level of the energy storage system;

[0062] It can be understood that the priority judgment problem in the concurrent multiple risk scenarios is solved, and the complex multi-parameter risk state is converted into an intuitive and operable level signal.

[0063] The above has described one embodiment of the present application in detail, but the content described is only a preferred embodiment of the present application, and cannot be considered as limiting the scope of the present application. Any equivalent changes and improvements made within the scope of the present application should still belong to the scope of the present application.

Claims

1. A multi-parameter risk classification and alarm method for energy storage systems, characterized in that, Includes the following steps: Step S1: Determine all primary risk parameters of the energy storage system and obtain historical fault data of the energy storage system to obtain several risk scenarios; For any risk scenario, determine all abnormal risk parameters under the risk scenario, and record the causal relationship between each abnormal risk parameter and the risk scenario. Step S2: Record each primary risk parameter as a leaf node and each risk scenario as an intermediate node. For any leaf node, establish a directed edge between it and the intermediate nodes that have a causal relationship with it. Based on the historical fault data, a risk threshold for the leaf node is determined. The leaf node generates a risk signal. The transmission rate of the risk signal along the directed edge is configured according to the risk threshold, so that when the physical value of the leaf node reaches the risk threshold, the risk signal reaches the intermediate node. Step S3: Establish a root node. For any intermediate node, establish an upper-level directed edge between it and the root node. If the intermediate node receives a risk signal, the intermediate node outputs an upper-level risk signal, which is transmitted to the root node along the upper-level directed edge. Step S4: Based on the risk sensitivity of each intermediate node determined manually, the root node obtains the risk level index according to all the received upper-level risk signals and the risk sensitivity of the corresponding intermediate nodes; The risk level of the energy storage system is determined based on the aforementioned risk level index; In step S1, the process of determining all abnormal risk parameters under the risk scenario includes: The risk scenario is designated as the target risk scenario. Change features of the target risk scenario are extracted based on a machine learning model. All fault data belonging to the target risk scenario are selected from the historical fault data. All primary risk parameters belonging to the same fault in all fault data are grouped into a single combination. For any primary risk parameter within any group of primary risk parameters, the correlation between the primary risk parameter and the change features is determined based on the Pearson correlation coefficient method. If a correlation exists, the combination of the primary risk parameter is designated as a related combination, and the correlation ratio of the primary risk parameter is calculated as P = n / N, where n is the number of related combinations of the primary risk parameter and N is the total number of combinations. If the correlation ratio of the primary risk parameter is greater than a preset ratio threshold, the primary risk parameter is designated as an abnormal risk parameter. In step S2, the process of configuring the transmission rate of the risk signal along the directed edge according to the risk threshold includes: Set the initial physical value Q of the leaf node, obtain the length L of the directed edge, and obtain the risk threshold Qt of the current leaf node. Then, obtain the transmission rate v of the risk signal along the directed edge v = L / (Qt-Q). If Qt≤Q, then the transmission rate of the risk signal along the directed edge is directly recorded as 0.

2. The multi-parameter risk classification and alarm method for an energy storage system according to claim 1, characterized in that, In step S1, the primary risk parameters include voltage, current, internal resistance, and temperature; the historical fault data includes fault data of the energy storage system at each fault during its historical operating cycle. The fault data includes the physical values ​​of all primary risk parameters and environmental data at the time of the fault. The environmental data includes air temperature, wind force, and solar radiation intensity, and the type of each fault of the energy storage system is recorded as a landscape scene.

3. The multi-parameter risk classification and alarm method for an energy storage system according to claim 1, characterized in that, In step S2, the process of determining the risk threshold of the leaf node based on the historical fault data includes: Obtain any intermediate node that has a causal relationship with the leaf node, and record the risk scenario corresponding to the intermediate node as the scenario to be tested. Filter out all fault data in the scenario to be tested from the historical fault data and record them as fault trigger data. Obtain normal operation data within the historical operating cycle. The normal operation data includes the physical values ​​and environmental data of all primary risk parameters when the energy storage system has not experienced a fault. Using environmental data and the physical values ​​of each primary risk parameter as independent variables, and the operating state of the energy storage system as the dependent variable, an initial logistic regression model is established. The operating state of the energy storage system includes fault and normal operation. The normal operation data is used as positive sample data, and the fault triggering data is used as negative sample data. The positive sample data and negative sample data are input into the initial logistic regression model to train the initial logistic regression model and obtain the logistic regression model. Acquire the current environmental data, denoted as the current environmental data, and input the current environmental data as a constraint into the logistic regression model to obtain a constrained logistic regression model. For any primary risk parameter, exhaustively enumerate the physical value gradient values ​​of several primary risk parameters and input them sequentially into the constrained logistic regression model to obtain output results. For any physical value gradient value of the primary risk parameter, repeat the operation of inputting into the constrained logistic regression model several times to obtain several output results. Obtain the proportion of the running state as faulty in all output results. If the proportion is greater than or equal to 90%, then the physical value gradient value is recorded as the abnormal physical value of the primary risk parameter. Select the minimum value among all abnormal physical values ​​of the primary risk parameter and record it as the risk threshold of the primary risk parameter, that is, the risk threshold of the leaf node corresponding to the primary risk parameter.

4. The multi-parameter risk classification and alarm method for an energy storage system according to claim 1, characterized in that, In step S3, the direction of the directed edge is from the leaf node to the middle node, and the direction of the upper-level directed edge is from the middle node to the root node.

5. The multi-parameter risk classification and alarm method for an energy storage system according to claim 1, characterized in that, In step S4, the process of establishing the risk level index includes: The risk level index K i This represents the risk sensitivity of the intermediate node corresponding to the i-th upper-level risk signal received by the root node, where num is the total number of the i-th upper-level risk signals received by the root node, i∈[1,num] and i is a positive integer.

6. The multi-parameter risk classification and alarm method for an energy storage system according to claim 1, characterized in that, In step S4, the process of determining the risk level of the energy storage system based on the risk level index includes: If the number of risk levels is preset to Z, then Z-1 risk index thresholds are set, and several risk index intervals {[-∞, R1), [R1, R2), ..., [R] are determined based on each risk index threshold. Z-2 R Z-1 ), [R Z-1 ,+∞]},where R Z-1 This represents the Z-1th risk index threshold; risk levels are assigned to each risk index interval in sequence, and the risk index interval in which the real-time risk level index is located is determined and recorded as the current risk index interval. The risk level corresponding to the current risk index interval is recorded as the risk level of the current energy storage system.

Images