The coordinated bidding of a hydropower producer in three-settlement markets with time-dependent risk measure

• Published in: Electric power systems research 2017-10, Vol.151, p.40-58
• Main Authors: Vardanyan, Y., Hesamzadeh, M.R.
• Format: Article
• Language: English
• Subjects: Autoregressive processes; Coordinated bidding; Electricity; Electricity generation; Electricity pricing; Hydroelectric power; Hydropower producer; Linear programming; Markets; Markov chains; Optimization; Real time; Risk levels; Three-settlement market; Time dependence; Time-dependent risk measure; Volatility
• ISSN: 0378-7796; 1873-2046; 1873-2046
• DOI: 10.1016/j.epsr.2017.05.007

•The HW–GARCH model for predicting prices in three-settlement markets is proposed.•The Markov model is proposed for predicting the best price model.•The directions of intra-day and real-time prices are predicted using Markov model.•A stochastic-dynamic model for coordinated bidding of a hydropower plant is derived.•The hourly version of CVaR (T-CVaR) is developed. This paper proposes a stochastic and dynamic mixed-integer linear program (SD-MILP) for optimal coordinated bidding of a risk-averse profit-maximizing hydropower producer. The day-ahead, intra-day, and real-time markets are considered. To model and predict day-ahead, intra-day, and real-time prices, the Holt–Winter (HW) and the Generalized Autoregressive Conditional Heteroscedastic (GARCH) predictors are combined using a proposed Markov switch. The discrete behavior of intra-day and real-time prices is modeled as different Markov states. The proposed Markov-based HW–GARCH model with a standard scenario generation-reduction technique is used to capture the uncertainty in day-ahead, intra-day, and real-time prices. The time-dependent conditional value at risk (T-CVaR) is proposed to model the risk of trading in different considered markets. The convex combination of the expected profit and T-CVaR is used as the objective of SD-MILP. The Markov-based HW–GARCH is modeled in Matlab and the SD-MILP is coded in GAMS. The Markov-based HW–GARCH predictor and the SD-MILP are used to develop the bidding curve of a three-reservoir hydropower producer using the electricity prices from the Nordic power market. To further examine the developed models a seven-reservoir hydropower producer is also studied. For these two cases, the coordinated bidding curves are derived and discussed.