Neuro-Genetic Method for Intrusive Appliance Load Monitoring

Abstract

Power utilities must understand how customers use their products. This is essential to formulate any long range planning，forecasting or marketing strategies. However，only customer’s service level loads are available. Monitoring/survey analysis of customer usage patterns is the best way to understand how customers use electricity. The key problem is that collection and analysis of customer usage data is very time consuming and very costly. In this　research，we propose a method to estimate the power consumption of appliances used in a household from service level load by Artificial Neural Network （ANN）. However，as selection of optimum parameters of stand-alone ANN and learning weights in ANN is difficult，in this research，we apply Genetic Algorithm （GA） to ANN called as Neuro-Genetic for automatic finding of optimal ANN parameters such as learning rate，momentum，etc . and also optimal weights for better training. As the result，we obtained some findings that are at least as same of that with the trial-and-error methods. Ultimately，we could greatly shorten the learning time. This method is required for monitoring.