Search

CN-122024420-A - Multi-source decision-making hierarchical safety execution early warning method for charging cabin and corresponding product

CN122024420ACN 122024420 ACN122024420 ACN 122024420ACN-122024420-A

Abstract

The application relates to the technical field of safety control, and provides a multi-source decision grading safety execution early warning method for a charging cabin. The method comprises the steps of obtaining multi-source asynchronous safety data in a charging cabin, conducting time alignment and feature fusion processing on the multi-source asynchronous safety data to generate a comprehensive safety state vector, outputting a target safety level corresponding to the comprehensive safety state vector and at least one execution plan under the level by a multi-level safety decision model based on the comprehensive safety state vector, adaptively adjusting fusion weight of data acquisition frequency and the data fusion model according to the target safety level, continuously obtaining and processing real-time safety data and calculating decision confidence coefficient based on the adjusted data acquisition frequency and the fusion weight, and synchronously triggering multiple heterogeneous alarm operations and linkage control logic of multiple execution mechanisms associated with the target safety level by a charging cabin controller according to the execution plan when the decision confidence coefficient exceeds a confidence coefficient threshold of the corresponding safety level.

Inventors

  • ZHANG WEI

Assignees

  • 深圳市德塔电池科技有限公司

Dates

Publication Date
20260512
Application Date
20260113

Claims (10)

  1. 1. A multi-source decision-making hierarchical safety execution pre-warning method for a charging pod, the method comprising: Acquiring multisource asynchronous safety data which are acquired by at least two different types of sensors and are not completely synchronous in time in a charging cabin; Performing time alignment and feature fusion processing on the multi-source asynchronous safety data based on a preset data fusion model to generate a comprehensive safety state vector; Inputting the comprehensive security state vector into a pre-trained multi-level security decision model, wherein the multi-level security decision model outputs a target security level corresponding to the comprehensive security state vector and at least one execution plan under the level, and the multi-level security decision model is generated based on historical security event data training and is used for mapping continuous or discrete security state characteristics to a predefined plurality of discrete security levels; according to the target security level, self-adaptively adjusting fusion weight of data acquisition frequency and the data fusion model; Continuously acquiring and processing real-time safety data based on the adjusted data acquisition frequency and the fusion weight so as to dynamically confirm the target safety level and calculate decision confidence; and when the decision confidence coefficient exceeds a confidence coefficient threshold value of the corresponding safety level, synchronously triggering various heterogeneous alarm operations and linkage control logic of a plurality of execution mechanisms associated with the target safety level by the charging cabin controller according to the execution plan.
  2. 2. The multi-source decision hierarchical security execution pre-warning method for a charging cabin according to claim 1, wherein the performing time alignment and feature fusion processing on the multi-source asynchronous security data based on a preset data fusion model to generate a comprehensive security state vector comprises: identifying a timestamp for each data stream in the multi-source asynchronous secure data; taking a time sequence of a data stream with highest sampling frequency in the multi-source asynchronous safety data as a reference time axis; for other data streams with sampling frequency lower than the reference time axis, generating estimated values at each time point of the reference time axis by adopting an interpolation algorithm based on the physical characteristics of the sensor and the rules of the historical data so as to realize time alignment of multi-source data; And inputting the time-aligned multidimensional data into an attention mechanism network, dynamically calculating the weight of each dimensional data on the comprehensive safety state at the current moment by the attention mechanism network, and carrying out weighted fusion according to the weight to generate the comprehensive safety state vector.
  3. 3. The multi-source decision-making hierarchical security enforcement pre-warning method for a charging pod of claim 2, wherein said adaptively adjusting the fusion weights of the data acquisition frequency and the data fusion model according to the target security level comprises: When the target security level is increased, increasing the data acquisition frequency of a preset key sensor; in the data fusion model, an initial weight bias of a data dimension related to thermal runaway, arcing features in the attention mechanism network is increased.
  4. 4. The multi-source decision-making hierarchical security enforcement pre-warning method for a charging pod of claim 1, wherein the calculating decision confidence comprises: In the dynamic confirmation process, a plurality of comprehensive safety state vectors continuously calculated in a time window are obtained; inputting the comprehensive security state vectors into the multi-stage security decision model again to obtain a prediction sequence of the security level in the time window; Analyzing the stability, consistency and variation trend of the predicted sequence, and calculating a numerical value between 0 and 1 as the confidence of the current decision by using an algorithm based on the statistical variance and the state transition probability.
  5. 5. The multi-source decision-making hierarchical safety execution pre-warning method for a charging pod of claim 1, wherein the synchronizing triggers a plurality of heterogeneous alarm operations associated with the target safety level and coordinated control logic of a plurality of actuators, comprising: analyzing the execution plan to generate a control instruction set with a plurality of time sequence marks; issuing the control instruction set to a corresponding executing mechanism controller, and strictly monitoring the starting time and executing time of each control instruction; the linkage control logic requires that the execution confirmation signal of the preamble key action is a necessary condition for starting the subsequent specific action.
  6. 6. The multi-source decision-making hierarchical safety enforcement pre-warning method for a charging pod of claim 5, further comprising the step of dynamic routing when executing coordinated control logic: Receiving feedback states of the plurality of execution mechanisms in real time; If the feedback of an executing mechanism of a certain preset critical action is abnormal or overtime, dynamically skipping or replacing the critical action according to a preset emergency path rule, starting an alternative emergency action sequence immediately, and temporarily improving the security level by one step.
  7. 7. The multi-source decision-making hierarchical safety enforcement pre-warning method for a charging pod of claim 1, further comprising the step of evidence chain solidification after triggering any alarm and coordinated control operations: automatically associating and packaging the following data to generate a tamper-proof security event evidence chain data packet: The original multisource asynchronous safety data fragment which is based on the trigger decision and is subjected to time alignment processing; The calculation process and result of the comprehensive safety state vector; input, output and decision confidence of the multi-stage security decision model; All control instruction sets issued and the time sequence thereof; All feedback status and time stamps for each actuator.
  8. 8. A multi-source decision-making hierarchical safety execution early warning device for a charging pod, the device comprising: the first acquisition module is used for acquiring multi-source asynchronous safety data which are acquired by at least two different types of sensors in the charging cabin and are not completely synchronous in time; The fusion module is used for carrying out time alignment and feature fusion processing on the multi-source asynchronous safety data based on a preset data fusion model to generate a comprehensive safety state vector; a plan output module, configured to input the comprehensive security state vector into a pre-trained multi-level security decision model, where the multi-level security decision model outputs a target security level corresponding to the comprehensive security state vector and at least one execution plan under the level, and the multi-level security decision model is generated based on historical security event data training, and is configured to map continuous or discrete security state features to a predefined plurality of discrete security levels; The adjusting module is used for adaptively adjusting the fusion weight of the data acquisition frequency and the data fusion model according to the target security level; The second acquisition module is used for continuously acquiring and processing real-time safety data based on the adjusted data acquisition frequency and the fusion weight so as to dynamically confirm the target safety level and calculate decision confidence; And the execution module is used for synchronously triggering various heterogeneous alarm operations associated with the target safety level and linkage control logic of a plurality of execution mechanisms by the charging cabin controller according to the execution plan when the decision confidence coefficient exceeds a confidence coefficient threshold value of the corresponding safety level.
  9. 9. An electronic device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the steps of the method according to any one of claims 1 to 7 when the computer program is executed by the processor.
  10. 10. A storage medium storing a computer program which, when executed by a processor, implements the steps of the method according to any one of claims 1 to 7.

Description

Multi-source decision-making hierarchical safety execution early warning method for charging cabin and corresponding product Technical Field The application relates to the technical field of safety control, in particular to a multi-source decision grading safety execution early warning method for a charging cabin and a corresponding product. Background With the popularization of electric vehicles and various electrochemical energy storage devices, the number and scale of centralized charging stations, power stations and energy storage compartments (hereinafter collectively referred to as "charging compartments") are rapidly increasing. A large number of high-energy-density battery packs are densely deployed in the charging cabin, and potential safety hazards such as thermal runaway, electrical short circuit, overcharge and overdischarge exist in the charging, standing or maintenance processes. Once a safety accident occurs, fire and explosion can be caused, and serious personal and property loss and secondary disasters are caused. Therefore, the charging cabin is monitored and early-warned in real time and accurately in a safe state, and grading intervention measures are timely and effectively executed when risk is confirmed, so that the safety and stability of the charging cabin are ensured. Currently, common charging pod safety monitoring schemes rely primarily on deploying multiple sensors (e.g., temperature, smoke, voltage current sensors) for data acquisition and triggering alarms based on preset single or combined thresholds. More advanced systems introduce simple decision logic, such as performing combination judgment on different sensor alarm signals, corresponding to different early warning levels, and triggering corresponding preset actions such as audible and visual alarm, ventilation, or power failure. These schemes enable automation to some extent from perceived motion. However, the prior art scheme still has the obvious defects in practical application that 1) safety data in a charging cabin is sourced from sensors with different types and sampling periods, the prior art scheme generally carries out threshold judgment or simple logic AND/OR operation on various types of data independently, lacks a mechanism for carrying out effective time alignment and deep fusion on multi-source asynchronous safety data, leads to insensitivity and inaccuracy on early and coupling risks, and is easy to generate false alarm or miss report, 2) decision rules (threshold and logic) of most schemes are static preset, monitoring strategies cannot be adaptively adjusted according to risk levels estimated in real time, response can be delayed due to untimely information updating and resource waste due to continuous high-frequency monitoring when the risk is upgraded, and 3) the prior art scheme generally lacks an evaluation link on self decision reliability, namely a dynamic calculation and confirmation process of decision confidence degree, and once noise or instantaneous interference triggers high-level early warning and carries out strong factor early warning (for example, the threshold and logic) causes unnecessary interruption of the whole system, and the whole system is difficult to run and lose reliability. Disclosure of Invention The application provides a multi-source decision-making hierarchical safety execution early warning method, a multi-source decision-making hierarchical safety execution early warning device and a storage medium for a charging cabin. In one aspect, the application provides a multi-source decision-making hierarchical safety execution early warning method for a charging cabin, which comprises the following steps: Acquiring multisource asynchronous safety data which are acquired by at least two different types of sensors and are not completely synchronous in time in a charging cabin; Performing time alignment and feature fusion processing on the multi-source asynchronous safety data based on a preset data fusion model to generate a comprehensive safety state vector; Inputting the comprehensive security state vector into a pre-trained multi-level security decision model, wherein the multi-level security decision model outputs a target security level corresponding to the comprehensive security state vector and at least one execution plan under the level, and the multi-level security decision model is generated based on historical security event data training and is used for mapping continuous or discrete security state characteristics to a predefined plurality of discrete security levels; according to the target security level, self-adaptively adjusting fusion weight of data acquisition frequency and the data fusion model; Continuously acquiring and processing real-time safety data based on the adjusted data acquisition frequency and the fusion weight so as to dynamically confirm the target safety level and calculate decision confidence; and when the decision confidence coefficient exceeds