Distance Until Charge Prediction for PHEVs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Plug-in hybrid electric vehicles (PHEVs) face challenges in optimizing energy management due to the difficulty in predicting the distance until the next charge, which affects fuel economy and battery depletion strategies, as existing methods do not accurately account for historical driving patterns and variability.
Innovation Solution
A controller-based system that predicts the distance until charge by filtering historical data based on time, day of week, and day type, using recursive least squares estimation to select reliable distance estimates and generate an accurate prediction for energy management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If historical data is filtered based on time and day of week to improve prediction accuracy, then measurement precision of distance until charge is improved, but device complexity increases due to multiple filtering criteria and recursive least squares estimation
Solution Approach 1:
The historical data is segmented into multiple filtered sets based on different criteria (time of day, day of week, day type). Each filtered set contains distance estimates for specific conditions, allowing the system to select the most relevant data for prediction. This segmentation improves measurement precision by comparing predictions against condition-matched historical data rather than all historical data.
Solution Approach 2:
The system performs preliminary filtering of historical data into multiple sorted sets before making predictions. By pre-organizing historical data according to time, day of week, and day type criteria, the system prepares the data structure in advance, reducing the computational complexity during actual prediction operations.
2Reliability
If multiple filtered sets of historical data are used to improve prediction reliability, then reliability of distance until charge prediction is improved, but loss of information increases due to selective filtering and variance thresholding
Solution Approach 1:
The system applies partial filtering by creating multiple filtered sets with different criteria rather than applying a single comprehensive filter. Each filtered set captures a portion of the historical data relevant to specific conditions. The variance thresholding selectively includes or excludes data points based on statistical criteria, maintaining reliability by focusing on consistent patterns while reducing information loss through conditional rather than absolute filtering.
Solution Approach 2:
The system changes parameters of the historical data by organizing it into multiple dimensions (time, day of week, day type) and applying variance thresholds. This parameter-based organization allows the system to maintain reliable predictions by selecting data sets with appropriate statistical properties while preserving relevant information through multi-criteria classification.
Data Source
AI summary
A vehicle includes a traction battery and a controller programmed to, in response to a first ignition cycle after a charging of the traction battery, output a distance until charge prediction based on one or more selected distance estimates from filtered sets of historical data that are derived by filtering historical data based on a time and a day of the first ignition cycle. The selected distance estimates may be from filtered sets having more than a predetermined number of members. The selected distance estimates may be from filtered sets having a variance that is less than an overall historic variance. The filtered sets may include members of an unfiltered set based on a time of day and day of week.