| Abstract |
Lead-acid batteries are used in telecommunications power systems that require strict conditions of
power source to provide un-interrupted power for telephone and data transmission service. The
information concerning the remaining battery discharge capacity is very important for the
maintenance and the drive of power systems to provide satisfactory backup energy. Especially it has
great effect on the delivery problem of the mobile power source in emergency states that are supplied
just only by the battery. The battery is also used for the electric vehicle that has to take one charge at
one run. So it is expected that we can predict the remaining battery discharge capacity easily and
quickly.
The battery discharge capacity is influenced by the discharge current, temperature, newness of the
batteries and other nonlinear parameters. Various estimation methods for residual capacity of the
sealed lead-acid battery have been proposed. However, no method can accurately predict the residual
capacity.
This paper describes a new method for predicting remaining battery discharge capacity by using
the backpropagation algorithm for the neural network (NN). The input layer of NN has two input
terminals corresponding to the battery terminal voltage and discharge current, respectively. The NNs
learn the discharge characteristics after getting tuned synaptic weights. And the results of learning are
worked into the display circuits of the residual capacity without calculating process. The temperature
effect on capacity is not considered, but by getting the data of the dependency of capacity on
temperature, it can be simulated easily by adding one input parameter to the NN. The results of the
simulation can be implemented with hardware using standard circuit elements. |
