Load forecasting is an important function in economic power generation, allocation between plants (Unit Commitment Scheduling), maintenance scheduling, and for system security applications such as peak shaving by power interchange with interconnected utilities. In this thesis the problem of fuzzy short term load forecasting is formulated and solved. The thesis starts with a discussion of conventional algorithms used in short-term load forecasting. These algorithms are based on least error squares and least absolute value. The theory behind each algorithm is explained. Three different models are developed and tested in the first part of the thesis. The first model (A) is a regression model that takes into account the weather parameters in summer and winter seasons. The second model (B) is a harmonics based model, which does not account for weather parameters, but considers the parameters as a function of time. Model (B) can be used where variations in weather parameters are not available. Finally, model (C) is created as a hybrid combination of models A and B. The parameters of the three models are estimated using the two static estimation algorithms and are used later to predict the load for twenty-four hours ahead. The results obtained are discussed and conclusions are drawn for these models. In the second part of the thesis new fuzzy models are developed for crisp load power with fuzzy load parameters and for fuzzy load power with fuzzy load parameters. Three fuzzy models (A), (B) and (C) are developed. The fuzzy load model (A) is a fuzzy linear regression model for summer and winter seasons. Model (B) is a harmonic fuzzy model, which does not account for weather parameters. Finally fuzzy load model (C) is a hybrid combination of fuzzy load models (A) and (B). Estimating the fuzzy parameters for the three models turns out to be one of linear optimization. The fuzzy parameters are obtained for the three models. These parameters are used to predict the load as a fuzzy function for twenty-four hours ahead. Prediction results are obtained and presented using data from Nova Scotia Power and Environment Canada.
展开▼
