Value of Stochasticity in Hydropower Planning Optimization

• Main Author: Marko, Vistica
• Format: Others
• Language: English
• Published: KTH, Elektriska energisystem 2012
• Online Access: http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-101702

With respect to market liberalization, efficient use of resources is becoming more important for players in the market. In order to achieve that different optimization techniques were developed which enable better operational efficiency. These techniques can be segmented into two different categories, depending on their time horizon: • Yearly time horizon – mid-term hydropower scheduling • Daily time horizon – short-term hydropower scheduling These two time horizons account for two case studies presented in this thesis. In the first case study (mid-term planning), the focus is on determining power plant’s optimal operating strategy, while taking into account the uncertainty in inflows and prices. Stochastic dynamic programming has been chosen as mid-term optimization technique. Since stochastic dynamic programming calls for a discretization of control and state variables, it may fall under the curse of dimensionality and therefore, the modeling of stochastic variables is important. By implementing a randomized search heuristic, a genetic algorithm, into the existing stochastic dynamic programming schema, the optimal way of using the stochasticity tries to be found. Two price models are compared based on the economic quality of the result. The results give support to the idea of using search heuristics to determine the optimal stochasticity setup, however, some deviations from the expected results occur. Second case study deals with short-term hydropower planning, with a focus on satisfying the predefined demand schedule while obtaining maximum profit. With short-term hydropower planning being a nonlinear and nonconvex problem, the main focus is on the linearization of unit performance curves, as well as satisfying technical constraints from the power plant perspective. This optimization techniques also includes the water value in the solution. The problem has been solved by means of mixed integer linear programming. The results from the second case study are fully in line with the expectations and it is shown that mixed integer linear programming approach gives good results with good computational time. Suggested improvements to the model and potential for future work can be found in the final chapter of this thesis.