Short-term Prediction of Water Flow Data into Hydro-electric Power Stations Using Local Fuzzy Reconstruction Method

Abstract

For predicting the flow into a hydro-electric power station, complex natural phenomena have to be dealt with, so conventional mathematical models based on hydraulics may not produce satisfactory results. When a neural network is used, its construction cannot be easily determined, and so extra neural networks have to be provided separately in addition to the normal neural network, according to experts' opinions about the problem. To solve these problems, the authors took the standpoint that if the inflow rate time-series data for hydro-electric power stations exhibit deterministic chaos, the status in the near future can be predicted. So the authors have applied the local fuzzy reconstruction method as a deterministic nonlinear short-term prediction method to data for the flow of water into hydro-electric power stations. In this paper, typical outflow analysis method using conventional mathematical models are first described briefly. Next, the “Local Fuzzy Reconstruction Method” is described. Third, chaotic behavior of water flow data into hydro-electric power stations are illustrated. Finally, the results of applying the method to the prediction of the flow into hydro-electric power stations are presented.