Battery State Estimation with Accuracy Decline Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing secondary battery life prediction systems using neural networks often produce less accurate estimate values, leading to inconveniences when controlling or announcing the state of the object based on these estimates.
Innovation Solution
A state estimation system that includes a processor to generate a learned model receiving measurement values of a first state variable and outputting an estimate value of a second state variable, with additional learning processes triggered by estimation accuracy decline information, preventing processing based on less accurate estimates by outputting estimation accuracy decline information when sudden changes or out-of-range inputs occur.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a learned model using neural network is used to estimate secondary battery state, then the prediction capability is provided, but the estimation accuracy declines in some cases
Solution Approach 1:
The system monitors estimation results and compares them against expected ranges. When estimation accuracy decline is detected (through sudden changes or out-of-range inputs), the system provides feedback to prohibit processing based on inaccurate estimates and triggers additional learning to improve the model
Solution Approach 2:
The system performs preliminary checks on measurement values before processing. It determines whether input values are within supposed ranges based on teaching data, and prohibits processing before inaccurate estimates can cause problems
2Loss of information
If processing is performed based on estimate values from the learned model, then the state information is obtained, but errors occur when estimation accuracy is low
Solution Approach 1:
The system performs preliminary validation of measurement values against supposed ranges before processing. This preliminary action prevents unreliable processing by identifying potential accuracy issues in advance
Solution Approach 2:
The system implements feedback mechanisms that monitor estimation quality and prohibit processing when accuracy decline is detected, ensuring that only reliable estimates are used for state information
3Productivity
If the learned model processes all measurement values, then continuous state estimation is provided, but inaccurate estimates are generated when inputs are out of range
Solution Approach 1:
The system performs preliminary range validation on all measurement values before they are processed by the learned model. This ensures that only within-range inputs with expected accuracy are processed continuously
Solution Approach 2:
The system dynamically adjusts processing based on input validation results. When out-of-range values are detected, processing is prohibited; when within-range values are detected, continuous estimation proceeds normally
Data Source
AI summary
A state estimation system includes a first processor, and the first processor executes: a first state variable measurement process of measuring a first state variable of a monitored object; a second state variable estimation repetition process of repeating a second state variable estimation process of inputting a measurement value of the first state variable into a learned model and acquiring an estimate value of a second state variable outputted from the learned model; and an estimation accuracy decline information output process of outputting estimation accuracy decline information when a predetermined estimate value sudden change determination condition is met with respect to a first estimate value of the second state variable acquired in response to input of a first measurement value of the first state variable.