Battery Spike Power Prediction Using Segmented Interval Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Batteries are often unable to accurately predict their spike power capability, leading to unexpected power off or brownout conditions, especially in dynamic and non-linear power demand scenarios, which can harm user experience and the load.
Innovation Solution
A battery management unit (BMU) employs iterative processing steps and a battery model to predict battery spike power capability up to 4-5 minutes in advance, accounting for time-variable and non-linear electrical characteristics by segmenting the prediction interval into subsections and correcting for model errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional battery monitoring methods are used, then the system is simple, but the battery cannot accurately predict spike power capability leading to unexpected power off or brownout
Solution Approach 1:
The prediction interval is segmented into multiple subsections (first prediction interval subsection, second prediction interval subsection, etc.), allowing the system to iteratively predict electrical characteristics at different time points. This segmentation enables accurate long-term prediction by breaking down the complex prediction task into manageable steps, resolving the contradiction between prediction accuracy and system complexity.
Solution Approach 2:
The system performs preliminary predictions of electrical characteristics at various time points within the prediction interval before the actual spike power capability prediction is needed. By pre-calculating these intermediate values, the system prepares the necessary data in advance, improving the accuracy of the final spike power capability prediction without requiring complex real-time calculations.
2Adaptability or versatility
If the prediction interval is treated as a single unit, then the calculation is simple, but the system cannot account for time-variable and non-linear electrical characteristics
Solution Approach 1:
The prediction interval is divided into multiple subsections, each handling specific time-variable characteristics. This allows the system to adapt to non-linear electrical behavior at different time points while maintaining a structured prediction process, resolving the contradiction between adaptability and process complexity.
Solution Approach 2:
The system dynamically adjusts predictions for each subsection of the prediction interval, allowing electrical characteristics to vary over time according to actual battery behavior. This dynamic approach enables the system to handle time-variable and non-linear characteristics effectively, improving adaptability without requiring an entirely complex redesign.
Data Source
AI summary
A battery system includes a battery configured to power a load, and a processing system comprising one or more processors. The processing system is configured to determine an electrical characteristic of the battery at a start of a prediction interval, predict a first predicted electrical characteristic of the battery in a first subsection of the prediction interval based at least in part on the electrical characteristic, predict a second predicted electrical characteristic of the battery in a second subsection of the prediction interval based at least in part on the electrical characteristic, the first predicted electrical characteristic, or both, and predict a spike power capability that the battery can support after an end of the prediction interval based at least in part on the second predicted electrical characteristic.