A method and device for energy storage battery cluster multi-source data credible fusion and cluster-level fault feature extraction

Abstract

The application discloses a kind of energy storage battery cluster multi-source data credible fusion and cluster level fault feature extraction method and device, belong to energy storage security risk control technical field.This method runs in general computing equipment, obtains BMS cluster, PCS, environmental sensing, station operation gateway multi-channel energy storage dataset;From the four-dimensional calculation of integrity, timeliness, consistency, data authority, each channel credibility weight coefficient is completed multi-source data weighted fusion, parameter conflict is preferentially sensed high credibility channel data;Extract inter-cluster circulating amplitude, float degeneration rate, standing self-discharge rate and other energy storage battery cluster exclusive fault features;Also according to the conversion coefficient and threshold value of corresponding electrochemical system can be loaded according to the identification of energy storage battery cluster electrochemical system.The software and hardware of the application are decoupled, and can be deployed in industrial computer, industrial tablet and other general equipment.

Description

This application is a divisional application of the invention patent application with application number 2026113832456 and application date of September 8, 2026. Technical Field

[0001] This invention relates to the field of safety detection technology for electrochemical energy storage in novel power systems, specifically to a method and apparatus for reliable fusion of multi-source data from energy storage battery clusters and extraction of cluster-level fault features. Background Technology

[0002] Large-scale energy storage battery clusters operate using multiple data sources, including data from BMS clusters, PCS converters, environmental sensor terminals, and site maintenance gateways. Existing energy storage risk control solutions directly reuse a single BMS data source, lacking a reliable quantitative evaluation mechanism for multi-channel data. When data conflicts occur for the same parameter across different channels, there are no unified acceptance rules, leading to distorted evaluation results and prominent issues of false alarms and missed alarms.

[0003] Meanwhile, traditional testing methods often refer to the characteristic parameters of other electrochemical energy storage devices, lacking specific fault characteristics for energy storage battery clusters under long-term float charging, static self-discharge, and large-scale parallel operation conditions, such as inter-cluster circulating current amplitude and float charging degradation rate; they cannot accurately characterize the unique aging and fault phenomena of energy storage, and directly applying other system models will lead to judgment bias.

[0004] Existing technologies are often tied to specific hardware acquisition controllers, making them inconvenient to deploy on general-purpose computing devices such as industrial PCs and industrial tablets, which hinders the licensing and commercialization of the technology. Therefore, there is an urgent need for a technical solution that can run on general-purpose computing devices to achieve quantitative assessment of the reliability of multi-channel data, acceptance of conflicting data, and extraction of specific fault characteristics of energy storage battery clusters. Summary of the Invention

[0005] This invention aims to solve the technical problems of existing energy storage risk control data sources being singular, lacking a credibility assessment mechanism for multi-channel data conflicts, lacking energy storage cluster-specific fault characteristics, and having poor portability due to the binding of solutions to dedicated hardware; it provides a method and device for reliable fusion of multi-source data and extraction of cluster-level fault characteristics for energy storage battery clusters, which can run on general computing devices, complete the calculation and weighted fusion of multi-source data credibility weights, and extract energy storage battery cluster-specific fault characteristic parameters. Technical solution

[0006] A method for trusted fusion of multi-source data and cluster-level fault feature extraction of energy storage battery clusters, running on a general-purpose computing device, the general-purpose computing device including a processor and a memory, the memory storing a computer-executable program, the processor executing the program, including the following steps: S1. Multi-channel dataset acquisition steps: Acquire the energy storage battery cluster operation dataset through at least four logical data channels; the logical data channels correspond to the BMS cluster interface, PCS interaction interface, environmental sensor acquisition terminal, and site operation and maintenance acquisition gateway, respectively. The energy storage battery cluster operation dataset includes cluster-level operation logs, battery cluster profile data, charging and discharging history records, grid connection and off-grid switching records, environmental temperature and humidity and gas sensor data, fire alarm linkage signals, as well as operation and maintenance inspection records and unstructured text data such as fault alarms within a preset time window before and after the fault occurs.

[0007] S2. Multi-channel four-dimensional credibility weight calculation steps: Each data channel is scored separately from four dimensions: data integrity, data timeliness, data consistency, and data source authority, and the credibility weight coefficient corresponding to each channel is calculated; where... Channel i integrity score: \(C_i = 1 − (number of missing fields / total number of fields)\); Timeliness score: \(T_i = e^{−λ·Δt_i}\), \(Δt_i\) is the data acquisition delay of this channel, and λ is the preset attenuation coefficient; Consistency score: \(A_i = 1 − |x_i−median(X)| / median(X)\), where \(x_i\) is the current parameter value of this channel, and \(median(X)\) is the median value of the same parameter across multiple channels; The authority score (R_i) is a preset fixed value: BMS cluster interface (R_i) = 0.95, PCS interaction interface (R_i) = 0.85, environmental sensor acquisition terminal (R_i) = 0.75, and site operation and maintenance acquisition gateway (R_i) = 0.70. Total credibility weight of channel i: \(W_i = αC_i + βT_i + γA_i + δR_i\); where \(α+β+γ+δ=1\), α, β, γ, and δ are weighting coefficients that can be dynamically configured according to the site's security level.

[0008] S3. Multi-channel data weighted fusion steps: Perform weighted fusion of multi-channel data based on the credibility weight coefficient of each channel; when different data channels output conflicting values ​​for the same parameter, the data from the channel with the higher credibility weight coefficient is adopted, and the fused dataset is output.

[0009] S4. Steps for extracting specific fault features of energy storage clusters: Perform data cleaning on the fused dataset and extract specific fault feature parameters of energy storage battery clusters; The specific fault characteristic parameters of the energy storage battery cluster include at least: inter-cluster circulating current amplitude, float charging degradation rate, and static self-discharge rate; It also includes at least one of the following: inter-cluster voltage difference sequence, energy storage unit voltage consistency deviation, internal resistance growth rate curve, capacity decay rate, temperature rise rate gradient distribution, charge / discharge depth distribution histogram, SOC balance deviation, and insulation resistance change rate.

[0010] The float-charge degradation rate characterizes the irreversible capacity loss rate per unit time within a preset rest / float-charge time window: \(λ_{float}=ΔQ_{irr} / (Δt·Q_{rated})\); where \(ΔQ_{irr}\) is the irreversible capacity loss estimated by pulse testing or voltage rebound method; \(Δt\) is the float-charge / rest time window; and \(Q_{rated}\) is the rated capacity of the battery cluster.

[0011] Furthermore, step S4 also includes: reading the electrochemical system identifier of the energy storage battery cluster, dynamically loading the characteristic conversion coefficients and judgment threshold set of the corresponding electrochemical system, and adapting the energy storage battery cluster to at least two different electrochemical systems.

[0012] This invention also provides a device for trusted fusion of multi-source data and cluster-level fault feature extraction of energy storage battery clusters. This device is a general-purpose computing device, comprising: a processor; a memory; the memory storing a computer-executable program; and when the computer-executable program is executed by the processor, it implements any of the aforementioned methods for trusted fusion of multi-source data and cluster-level fault feature extraction of energy storage battery clusters. Beneficial effects

[0013] 1. This solution runs on general-purpose computing devices and does not rely on customized dedicated hardware controllers. It can be deployed on industrial PCs, industrial tablets, and servers. The software and hardware are decoupled, which facilitates external patent licensing. 2. Innovatively adopts a four-dimensional quantitative model of integrity, timeliness, consistency, and authority to calculate the channel credibility weight. In case of multi-source data conflict, the system automatically adopts the high credibility data source, reducing false alarms and missed alarms caused by differences in data sources. 3. Extract specific fault characteristic parameters of energy storage battery clusters, and introduce energy storage-specific indicators such as inter-cluster circulating current amplitude, float charging degradation rate, and static self-discharge rate to distinguish them from other electrochemical energy storage equipment characteristic systems and adapt them to long-term float charging, static storage, and large-scale cluster parallel operation conditions. 4. Supports reading the electrochemical system identifier of energy storage battery clusters, dynamically loading conversion coefficients and thresholds for different electrochemical systems, and is compatible with energy storage battery clusters with various electrochemical systems; 5. The output fusion dataset and feature parameters can be directly used by the backend fault attribution, risk simulation, and health assessment modules as standardized input for upper-level security assessment. Detailed Implementation

[0014] Example 1: Deployment of Industrial and Commercial Energy Storage Sites on Industrial Flat Panels. This method is deployed on industrial flat panels. The flat panel receives raw datasets uploaded by the BMS cluster, PCS, environmental sensor terminal, and site operation and maintenance gateway via the network. It performs four-dimensional reliability calculations to complete the weighted fusion of multi-source data. It extracts energy storage-specific features such as inter-cluster circulating current amplitude, float charging degradation rate, and static self-discharge rate, and outputs the fused dataset and feature parameters for use by the backend fault analysis module. The flat panel only performs calculations and does not directly collect physical signals of the energy storage units.

[0015] Example 2: Deployment of this method on a large grid-side energy storage site - industrial control computer. The method runs on a regular industrial control computer at the site, receives multi-source heterogeneous data, automatically calculates the reliability of each channel, and automatically accepts the data of the higher reliability channel when multiple channels have conflicting values ​​for the same parameter. It outputs the cleaned fused dataset and the specific fault characteristics of the energy storage cluster.

[0016] Example 3: Adaptation scenario for energy storage battery clusters with a third electrochemical system. The system reads the electrochemical system identifier of the energy storage battery cluster as a third electrochemical system, and automatically switches to load the feature conversion coefficients and judgment threshold set corresponding to the third electrochemical system to complete the feature extraction of the energy storage battery cluster. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall architecture of the energy storage battery cluster multi-source data reliable fusion and cluster-level fault feature extraction system of the present invention; Figure 2 This is a schematic diagram of the multi-channel credibility weight calculation process of the present invention.

Claims

1. A method for reliable fusion of multi-source data and cluster-level fault feature extraction of energy storage battery clusters, characterized in that, Running on a general-purpose computing device, the general-purpose computing device including a processor and memory, the method includes: S1. Multi-channel Dataset Acquisition: Acquire the energy storage battery cluster operation dataset through at least four logical data channels; the logical data channels correspond to the BMS cluster interface, PCS interaction interface, environmental sensor acquisition terminal, and site operation and maintenance acquisition gateway; the dataset includes cluster-level operation logs before and after the fault within a preset time window, battery cluster files, charging and discharging history, grid connection and disconnection records, environmental sensors, fire alarm linkage signals, and unstructured text of operation and maintenance inspections; S2. Four-dimensional credibility weight calculation: Each channel is scored from four dimensions: data integrity, data timeliness, data consistency, and data source authority, and the credibility weight coefficient of each channel is calculated. Channel i Completeness score (C_i=1−(number of missing fields / total number of fields)); Timeliness score (T_i=e^{−λ·Δt_i}\); Consistency score \(A_i=1−|x_i−median(X)| / median(X)\); The authority (R_i) is a preset base value. The BMS cluster interface (R_i) = 0.95, the PCS interaction interface (R_i) = 0.85, the environmental sensor acquisition terminal (R_i) = 0.75, and the site operation and maintenance acquisition gateway (R_i) = 0.

70. The total credibility weight of the channel is \(W_i=αC_i+βT_i+γA_i+δR_i\); \(α+β+γ+δ=1\), and the weighting coefficient can be dynamically configured according to the security level of the site; S3. Multi-channel weighted fusion: Multi-channel data is weighted and fused based on the confidence weight coefficient; when multiple channels with the same parameter have conflicting values, the data from the channel with the higher confidence weight is adopted, and the fused dataset is output. S4. Cluster-level fault feature extraction: Perform data cleaning on the fused dataset and extract specific fault feature parameters for energy storage battery clusters; the specific fault feature parameters include at least the inter-cluster circulating current amplitude, float charging degradation rate, and static self-discharge rate; The float degradation rate is given by \(\lambda_{float}=\Delta Q_{irr} / (\Delta t\cdot Q_{rated})\), where \(\Delta Q_{irr}\) is the irreversible capacity loss, \(\Delta t\) is the float / rest time window, and \(Q_{rated}\) is the rated capacity.

2. The method according to claim 1, characterized in that, The specific fault characteristic parameters also include at least one of the following: inter-cluster voltage difference sequence, energy storage unit voltage consistency deviation, internal resistance growth rate curve, capacity decay rate, temperature rise rate gradient distribution, charge / discharge depth distribution histogram, SOC equalization deviation, and insulation resistance change rate.

3. The method according to claim 1, characterized in that, Step S4 also includes: reading the electrochemical system identifier of the energy storage battery cluster, dynamically loading the characteristic conversion coefficients and judgment threshold set of the corresponding electrochemical system, and adapting the energy storage battery cluster to at least two different electrochemical systems.

4. A device for reliable fusion of multi-source data and cluster-level fault feature extraction of energy storage battery clusters, characterized in that, General-purpose computing devices, including: processor; The memory stores a computer-executable program that, when executed by a processor, implements the method of any one of claims 1 to 3.

Images