Method for anomaly prediction of battery parasitic load
View Patent ↗Anomaly prediction of battery parasitic load includes processing input data related to a state of charge for a battery and a durational factor utilizing a machine learning algorithm and generating a predicted start-up state of charge. Warnings are issued if the predicted start-up state of charge drops below a threshold level within an operational time.
1. A method for anomaly prediction of parasitic load on a battery comprising:
processing input data related to a state of charge for the battery and a durational factor, wherein said processing comprises a machine learning algorithm and is operative to generate a predicted start-up state of charge; and
indicating a warning if said predicted start-up state of charge is below a threshold level within an operational time.
2. The method of claim 1 , wherein said machine learning algorithm comprises fuzzy logic.
3. The method of claim 2 , wherein said fuzzy logic updates on the basis of said input data, said predicted start-up state of charge, and a prediction error.
4. The method of claim 3 , wherein said fuzzy logic correlates a temporal indication to said input data, such that said processing indicates temporal trends in the input data.
5. The method of claim 2 , wherein said machine learning algorithm further comprises a neural network.
6. The method of claim 1 , wherein said machine learning algorithm comprises a neural network.
7. The method of claim 6 , wherein said neural network updates on the basis of said input data, said predicted start-up state of charge, and a prediction error.
8. The method of claim 1 , wherein said input data relating to said state of charge comprises a running state of charge.
9. The method of claim 1 , wherein said input data relating to said state of charge comprises a battery state of health.
10. The method of claim 1 , wherein said input data relating to said durational factor comprises a key-off time.
11. The method of claim 1 , wherein said threshold level is a predetermined critical level, said critical level being a calculated lowest charge that will support a start-up event.
12. The method of claim 1 , wherein said threshold level is a predetermined critical level, said critical level being a predetermined measure higher than a calculated lowest charge that will support a start-up event.
13. The method of claim 1 , wherein said threshold level is a threshold function based upon said durational factor.
14. The method of claim 1 , further comprising:
measuring an actual state of charge of said battery;
comparing said actual state of charge to said predicted start-up state of charge; and
initiating a warning if said comparison indicates that said actual state of charge has dropped more than a threshold drop below said predicted start-up state of charge.
15. The method of claim 1 , wherein said warning comprises an alert by means of a communications network.
16. The method of claim 1 , wherein said warning comprises reduced functionality of devices utilizing said battery.
17. The method of claim 1 , wherein said processing further comprises heuristic techniques for the purpose of overcoming inconsistencies in said input data.
18. The method of claim 1 , wherein said processing further comprises trend analysis whereby a battery state of health for said battery is developed.
19. A method for anomaly prediction of parasitic load on a battery in a motor vehicle comprising:
receiving input data related to a running state of charge for the battery and a key-off time;
processing said input data to create a predicted start-up state of charge for said battery, wherein said processing comprises a machine learning algorithm; and
initiating a warning if said predicted start-up state of charge is below a threshold level before a threshold time.
20. The method of claim 19 , wherein said machine learning algorithm comprises fuzzy logic operable to analyze and identify said input data into a plurality of fuzzy groups indicating a degree to which said input data conforms to historical input data sets, said historical input data sets being correlated to resulting prediction errors.
21. The method of claim 20 , wherein said machine learning algorithm comprises a layered feed-forward neural network operative to adapt to multiple iterations of said input data and related prediction errors.