Battery Virtual Sensor Module for Spatial-Temporal Temperature Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current battery energy storage systems face operational safety risks due to temperature spikes and thermal runaways, largely because of limited and inaccurate thermal monitoring capabilities, which can lead to catastrophic failures.
Innovation Solution
A virtual sensor module combining a Proportional-Integral-Derivative (PID) controller with a trained convolutional neural network—long-short-term memory (CNN-LSTM) module is used to generate refined spatial-temporal temperature measurements across batteries, addressing the limitations of physical sensors by optimizing PID parameters and predicting temperature sequences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If physical thermal sensors are used to monitor battery temperature, then temperature measurements can be obtained, but the measurement precision and coverage are limited due to the number and distribution of sensors
Solution Approach 1:
The patent creates a virtual copy of the physical temperature field using a neural network model. The CNN-LSTM virtual sensor learns from limited physical sensor readings to generate a complete spatial-temporal temperature distribution map, effectively copying the temperature field information that would require numerous physical sensors to obtain directly.
Solution Approach 2:
The patent introduces a CNN-LSTM neural network as an intermediary between the limited physical sensor measurements and the desired comprehensive temperature field information. This intermediary processes the sparse sensor data to infer the complete temperature distribution, avoiding the need for dense sensor deployment.
2Productivity
If simple physical models are used to estimate temperature distribution, then computational complexity is reduced, but the accuracy of temperature predictions decreases
Solution Approach 1:
The patent transitions from traditional physical model parameters to data-driven neural network parameters. The CNN-LSTM model learns optimal parameters from training data, enabling it to capture complex nonlinear thermal behaviors that simple physical models miss, while maintaining computational efficiency through the learned parameter representations.
Solution Approach 2:
The patent replaces traditional mechanical/physical modeling approaches with a data-driven neural network approach. Instead of relying on simplified physical equations, the system uses learned patterns from training data to predict temperature distributions, substituting the mechanical modeling paradigm with an intelligent computing paradigm.
3Loss of information
If a large number of physical sensors are deployed to improve measurement coverage, then more complete temperature data is obtained, but the cost and system complexity increase significantly
Solution Approach 1:
The virtual sensor creates a complete digital replica of the temperature field using only a few physical sensor inputs. The CNN-LSTM model generates full spatial-temporal temperature sequences that copy the information content of a dense sensor array while using minimal physical hardware.
Solution Approach 2:
The patent makes a single virtual sensor system perform the function of multiple physical sensors. The CNN-LSTM model processes inputs from limited physical sensors to produce comprehensive temperature field information, making the system universally applicable regardless of the specific sensor configuration used.
4Productivity
If traditional virtual sensing methods are used, then computational requirements are reduced, but the accuracy in capturing nonlinear temperature variations is insufficient
Solution Approach 1:
The patent changes from fixed physical model parameters to adaptive neural network parameters that evolve during training. The CNN-LSTM model learns to represent nonlinear temperature variations through its learned weights and biases, capturing complex thermal dynamics that traditional methods miss while maintaining computational efficiency.
Solution Approach 2:
The patent introduces dynamic adaptability to the virtual sensing system. The CNN-LSTM model adapts its parameters during training to match the specific thermal characteristics of the battery system, enabling it to dynamically capture nonlinear temperature variations that static physical models cannot represent.
Data Source
AI summary
This document describes a virtual sensor module for generating spatial-temporal temperature measurements of a battery and a method of using the virtual sensor module to generate the above-mentioned measurements. The virtual sensor module utilizes a Proportional-Integral-Derivative (PID) controller module communicatively coupled to a trained convolutional neural network—long-short-term memory (CNN-LSTM) module to trigger the trained CNN-LSTM module to generate a refined predicted series of real temperature sequences across the battery, based on a set of optimized PID parameters (K*p, K*i, K*d), predicted error sequences, and measured temperature sequences of the battery.