Battery Control via Neural Network Internal State Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for controlling battery input/output operations are inefficient due to inaccuracies in determining internal battery states, leading to suboptimal performance and lifespan management.
Innovation Solution
A device equipped with a trained artificial neural network that senses input/output parameters to predict internal characteristic parameters, allowing for precise control of charging and discharging based on acquired data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional indirect inference methods are used to determine battery internal state, then the measurement process is simple, but the measurement precision is poor leading to large errors
Solution Approach 1:
The patent introduces an artificial neural network as an intermediary computational model that maps easily measurable external parameters (voltage, current, temperature) to difficult-to-measure internal parameters (state of charge, state of health, capacity). This neural network intermediary enables high-precision indirect measurement without requiring complex internal sensors or destructive testing.
Solution Approach 2:
The patent replaces traditional mechanical/chemical measurement methods (which require physical access to battery internals or destructive testing) with an information-processing approach using neural networks. The system substitutes physical measurement complexity with computational intelligence, achieving high precision through data-driven modeling rather than physical intervention.
2Productivity
If control is performed considering maximum error for stability, then the reliability is improved, but the productivity is reduced as battery performance cannot be maximized
Solution Approach 1:
The patent implements a feedback control system where the neural network continuously predicts battery internal state based on real-time measurements, and this predicted state feeds back to optimize charging/discharging control parameters. This closed-loop feedback enables dynamic adjustment that simultaneously maximizes performance and ensures stability by adapting to actual battery conditions rather than relying on conservative fixed parameters.
Solution Approach 2:
The patent transitions from static, conservative control parameters designed for worst-case scenarios to dynamic control parameters that adapt in real-time based on neural network predictions of actual battery state. This dynamic approach allows the system to operate near optimal performance boundaries while maintaining reliability through continuous state monitoring and adjustment.
3Measurement precision
If the battery is installed in a product, then the adaptability is improved, but the measurement precision of internal state deteriorates as it becomes impossible to grasp internal parameters without destroying the battery
Solution Approach 1:
The patent uses the neural network as an intermediary that translates easily accessible external measurements (voltage, current, temperature) into accurate estimates of internal parameters (capacity, state of charge, state of health). This intermediary computational model enables precise internal state measurement through the battery's normal operational interfaces without requiring physical access to internal components or destructive disassembly.
Data Source
AI summary
A method of controlling a battery is disclosed. The method includes training an artificial neural network to calculate an internal characteristic parameter value of the battery corresponding to a sensed input/output parameter value using training data, sensing the input/output parameter value of the battery, acquiring the characteristic parameter value corresponding to the sensed input/output parameter value using the trained artificial neural network, and controlling charging or discharging of the battery based on the acquired characteristic parameter value.