Storage Battery Diagnosis Using Pseudo-Random Pulses and Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for diagnosing storage battery deterioration require extensive Fourier analysis and fitting calculations, which are time-consuming and labor-intensive, making them inefficient for rapid assessment.
Innovation Solution
A diagnosis apparatus using a pseudo-random pulse signal to measure current and voltage changes, combined with a machine learning model to estimate the battery's internal state directly from these measurements, reducing the need for Fourier analysis and fitting calculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Fourier analysis and fitting calculations are performed to estimate battery internal state, then measurement precision is improved, but loss of time increases
Solution Approach 1:
The patent replaces the traditional mechanical calculation system (Fourier analysis and fitting calculations) with a neural network-based system. The neural network is trained offline using impedance frequency characteristics and equivalent circuit model calculations, then deployed for rapid online estimation of battery internal states, eliminating the need for time-consuming real-time mathematical computations while maintaining high accuracy
Solution Approach 2:
The patent performs preliminary training of the neural network using equivalent circuit model calculations and impedance frequency characteristics before actual battery diagnosis. This pre-computation phase creates a trained model that can rapidly estimate internal states during operation, avoiding the need to perform Fourier analysis and fitting calculations in real-time
2Measurement precision
If traditional impedance measurement methods are used, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent replaces the complex analytical system consisting of Fourier analysis modules and fitting calculation algorithms with a pre-trained neural network. This substitution maintains the ability to accurately estimate internal states from impedance measurements while dramatically simplifying the device architecture and reducing computational complexity
3Measurement precision
If Fourier analysis is performed on temporal changes in current and voltage, then measurement precision is improved, but productivity decreases
Solution Approach 1:
The patent substitutes the time-consuming Fourier analysis process with a neural network that has been pre-trained to directly map current and voltage temporal changes to internal state estimates. This enables rapid diagnosis without sacrificing measurement precision
Solution Approach 2:
The patent performs the computationally intensive Fourier analysis and equivalent circuit model fitting in advance during the neural network training phase. The trained network then uses these pre-computed relationships to rapidly estimate internal states during actual battery operation, significantly improving diagnosis speed
Data Source
AI summary
In an embodiment, a diagnosis apparatus includes a processing circuit, and the processing circuit measures temporal changes in current and voltage of a storage battery where a pseudo-random pulse signal of current is input to the storage battery. By using a machine learning model which outputs an internal state of the storage battery in response to input of electrical characteristic data based on current time series data and voltage time series data of the storage battery, the processing circuit inputs target electrical characteristic data based on a measurement result regarding the temporal changes in the current and the voltage as the electrical characteristic data to the machine learning model, and estimates the internal state of the storage battery based on a result of outputting from the machine learning model in response to the input of the target electrical characteristic data.