Vehicle energy management system using prognostics
First Claim
Patent Images
1. An energy management system for controlling an electrical system, comprising:
- data collection components providing quantified variables for forming an instantaneous state vector;
a time series predictor comprising an artificial neural network for generating an estimated future vector value in response to said instantaneous state vector, wherein said time series predictor further comprises a memory buffer coupled to said artificial neural network for time sampling contents of said artificial neural network and providing said time sampled contents to said artificial neural network when generating a subsequent estimated future vector value;
a probability calculator for generating a probability value in response to comparing said time sampled contents with a predetermined set of training vectors; and
an electrical system manager including predetermined decision rules invoked in response to said estimated future vector value and said probability value to adapt said electrical system to expected electrical conditions.
12 Assignments
0 Petitions
Accused Products
Abstract
An energy management system for controlling an electrical system comprises data collection components, such as sensors, for providing quantified variables for forming an instantaneous state vector. A time series predictor generates an estimated future vector value in response to the instantaneous state vector. An electrical system manager includes predetermined decision rules invoked in response to the estimated future vector value to adapt the electrical system to expected electrical conditions.
43 Citations
13 Claims
-
1. An energy management system for controlling an electrical system, comprising:
-
data collection components providing quantified variables for forming an instantaneous state vector; a time series predictor comprising an artificial neural network for generating an estimated future vector value in response to said instantaneous state vector, wherein said time series predictor further comprises a memory buffer coupled to said artificial neural network for time sampling contents of said artificial neural network and providing said time sampled contents to said artificial neural network when generating a subsequent estimated future vector value; a probability calculator for generating a probability value in response to comparing said time sampled contents with a predetermined set of training vectors; and an electrical system manager including predetermined decision rules invoked in response to said estimated future vector value and said probability value to adapt said electrical system to expected electrical conditions. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9)
-
-
10. A method of managing electrical energy within an electrical system, said method comprising the steps of:
-
collecting a plurality of quantified variables corresponding to said electrical system; forming an instantaneous state vector; inputting said instantaneous state vector into an input layer of an artificial neural network of a time series predictor; processing outputs of said input layer together with outputs of a buffer memory in a hidden layer of said artificial neural network; time sampling predetermined contents of said artificial neural network in said buffer memory for processing in said hidden layer with a subsequent time sample of said quantified variables; processing outputs of said hidden layer in an output layer of said artificial neural network to produce an estimated future vector of said quantified variables; comparing said time sampled predetermined contents with a predetermined set of training vectors to identify a probability value; and adjusting said electrical system in response to said estimated future vector and said probability value. - View Dependent Claims (11, 12, 13)
-
Specification