Neural network for signaling abnormal readings in an electricity consumption application

In the electricity billing Application in Iraq, one problem faced is how to filter abnormal readings that are either exceptionally low or exceptionally high, so as to verify these readings before billing the customer. Previously, various statistical methods were used to differentiate such readings. However, these methods were not considered satisfactory because of the associated rate of incorrect classifications obtained. This problem was addressed using a three-layer Back-Propagation neural network This model succeeded in screening abnormal readings, and proved to outperform the statistical techniques.