Chaotic Short-Term Prediction to Water Flow into Hydroelectric Power Stations

We applied local fuzzy reconstruction as deterministic nonlinear short-term prediction to data for water flow into hydroelectric power stations. Such prediction involves complex natural phenomena, and conventional hydraulics-based mathematical models do not produce satisfactory results. When a neural network is used, its construction cannot be easily determined, so extra neural networks must also be provided separately, based on experts' opinions. To solve these problems, we held that if time-series data of the inflow rate for hydroelectric power stations exhibits deterministic chaos, the status in the near future is predicted. Typical outflow analysis using conventional mathematical models is described briefly, followed by local fuzzy reconstruction, then results are given from applying this to water flow prediction.