On stochastic optimization for short-term hydropower planning

Detta är en avhandling från Stockholm : KTH Royal Institute of Technology

Sammanfattning: Renewable generation is the fastest growing energy resources in the past decade. Renewable energy sources, particularly wind power, provide clean and environmentally friendly energy to meet the system demand, meanwhile introducing huge levels of uncertainty in the system. On the one hand the deregulated electric power industry and on the other hand the intermittent nature of renewable energy sources cause highly volatile and uncertain electricity prices in different market places. This will create challenges for economical operation and planning of the flexible energy sources, particularly hydropower, which being a flexible energy source is the best option to balance wind power variation.The main purpose of this work is to develop optimal short-term planning models for price taker hydropower producer working in the existing environment. Those models have to deal with the huge level of uncertainties the wind power introduces into the power system.An optimization tool known as stochastic optimization is used to plan hydropower production under uncertainties.The first model, which is used to make sensitivity analysis, is a twostage stochastic linear programming problem. The uncertainties are handled by generating scenarios based on historical data. Profound sensitivity analysis is provided, in terms of volatility in day-ahead market prices and water inflow level as well as in terms of water opportunity cost and initial volume of the reservoir. Based on the comparison of the stochastic and corresponding deterministic problems, the result aims to show the impact of modeling the uncertainties explicitly. The results show that for the short-term hydropower planning problems the effect of considering price uncertainty in the stochastic model is higher compared with considering inflow level uncertainty.The second model used in this work is a two-stage stochastic linear programming problem. The model generates optimal bids to day-ahead market considering real-time market price uncertainties. While simultaneously bidding to both markets, the results for most of the hours suggest two actions; either to bid the available amount of energy to upward regulation market or to bid the maximum capacity to day-ahead market and bring back the whole amount making down regulation. To make the bidding strategies more flexible and robust different approaches are modeled and assessed. Finally one of the approaches is suggested as the most applicable one.