Fuzzy inference systems approach for long term hydrothermal scheduling

The optimization of the water resource usage in hydrothermal electric energy systems is crucial to assure an economic and reliable load supply. The long term hydrothermal scheduling is a complex problem mainly due to the randomness of the inflows and the nonlinearity of hydro production and thermal cost functions. Some optimization approaches have been proposed, including the Stochastic Dynamic Programming (SDP), which is one of the most commonly used techniques to deal with this problem. Its computational requirements, however, tend to be heavy and, as a result, its application on real systems is limited. In this paper we proposed the use of an Adaptive Neuro-Fuzzy Inference System in parallel with a deterministic optimization model as a simpler and less complex alternative approach to the hydrothermal scheduling. The information of the optimal operation is processed by the network that produces fuzzy rules describing the optimal decisions. The performance of the proposed approach was compared to other policies, including SDP, by simulation using historical inflows records of Emborcacao, a large Brazilian hydroelectric power plant. Results demonstrated that the Neuro-Fuzzy approach provided similar performance to the more computationally complex and commonly used SDP.