Stoaty
Well-known member
I made an experimental version 0.95f of the Battery Aging Model:
https://dl.dropboxusercontent.com/u/48149991/Leaf%20Battery%20Degradation%20Model%20Version%20095f.ods" onclick="window.open(this.href);return false;
After studying the table on the calibration table and doing some graphing, plus Nissan's focus on how many miles the Phoenix drivers that lost bars had driven, I decided that the assumption of 1.5% capacity loss per 10,000 miles might be too low. In version 0.95f I changed that assumption to 2.0% (on the Degradation Model tab) to see what effect that would have. Unfortunately, I still have to calculate the Corrected Predicted Total Loss for each entry, although it only takes about 10 seconds per entry since it is semi-automated. I haven't figured out a way to automate the correction.
Mean of (Actual Loss - Predicted Loss) = 0.29% (was 2.01% before the parameter was changed)
Standard Deviation of (Actual Loss - Predicted Loss) = 1.53% (was 1.93% before the parameter was changed)
(Maximum of (Actual - Predicted)) - (Minimum of (Actual - Predicted)) = 4.85% (was 6.30%)
Comment: Mean of differences is now close to zero, Standard Deviation is less, range between highest and lowest difference is less. For now I am going to track both models with additional entries (from both Leafs that haven't been reported yet, and repeat reports every 3 months). Another 6-12 months should give enough info to see if the model can accurately predict loss for new entries.
https://dl.dropboxusercontent.com/u/48149991/Leaf%20Battery%20Degradation%20Model%20Version%20095f.ods" onclick="window.open(this.href);return false;
After studying the table on the calibration table and doing some graphing, plus Nissan's focus on how many miles the Phoenix drivers that lost bars had driven, I decided that the assumption of 1.5% capacity loss per 10,000 miles might be too low. In version 0.95f I changed that assumption to 2.0% (on the Degradation Model tab) to see what effect that would have. Unfortunately, I still have to calculate the Corrected Predicted Total Loss for each entry, although it only takes about 10 seconds per entry since it is semi-automated. I haven't figured out a way to automate the correction.
Mean of (Actual Loss - Predicted Loss) = 0.29% (was 2.01% before the parameter was changed)
Standard Deviation of (Actual Loss - Predicted Loss) = 1.53% (was 1.93% before the parameter was changed)
(Maximum of (Actual - Predicted)) - (Minimum of (Actual - Predicted)) = 4.85% (was 6.30%)
Comment: Mean of differences is now close to zero, Standard Deviation is less, range between highest and lowest difference is less. For now I am going to track both models with additional entries (from both Leafs that haven't been reported yet, and repeat reports every 3 months). Another 6-12 months should give enough info to see if the model can accurately predict loss for new entries.