Evolutionary intelligent systems for pattern classification and price based electric load forecasting applications

In this dissertation, evolutionary intelligent system technologies of Fuzzy Logic (FL), Genetic Algorithms (GAs) and Artificial Neural Networks (ANNs) are used to develop two novel systems for pattern classification and short-term electric load forecasting in a price-sensitive environment. The core of both systems consists of FL modules that are developed using GAs. One of the main novelties of this work is in proposing new representation schemes for fuzzy membership functions and fuzzy rules so that all the required parameters of the FL modules can be computed from the training data using GA optimization. A so-called "Effectiveness Measure (EF)" is proposed for systematic addition or deletion of fuzzy rules during GA optimization.; The developed GA-based fuzzy classifier provides an alternative approach to supervised pattern recognition. The only parameters that need to be specified a priori are the maximum numbers of partitions (membership functions) for each of the variables and the general shapes of the membership functions to be used. The approach is applied to two real-world databases referred to as Iris and Wine databases and a simulated Gaussian database. The resulting classifiers are all quite compact with a small number of rules and perform very well. The classification accuracy is compared to other classifiers with favorable results.; The second application is motivated due to the need for a short-term hourly electric load forecaster in a deregulated and price-sensitive environment. A real-time pricing type scenario is envisioned where energy prices could change on an hourly basis with the consumer having the ability to react to the price signal through shifting his electricity usage from expensive hours to other times when possible. The load profile under this scenario would have different characteristics compared to that of the regulated, fixed-price era. Consequently, short-term load forecasting models customized on price-insensitive (PIS) historical data of regulated era would no longer be able to perform well. In this dissertation, a price-sensitive (PS) load forecaster is developed. This forecaster consists of two stages, an ANN based PIS load forecaster followed by a FL system that transforms the PIS load forecasts of the first stage into PS forecasts. The first stage forecaster is a widely used forecaster known as ANNSTLF. The FL system of the second stage is developed using a similar approach to the one used for the fuzzy classifier. Again through utilization of new representation schemes, all the required parameters are obtained from the training data using GA optimization. Since the electricity markets have not completely transitioned to a PS environment yet, actual PS load data is not available. Consequently, another FL system is developed to simulate PS load data from the PIS historical data of a utility. This new PS load forecaster termed NFSTLF (Neuro-Fuzzy Short Term Load Forecaster) is tested on three simulated PS load database and it is shown that it produces superior results to the PIS ANNSTLF load forecaster.