Sizing of a pumped storage power plant in S. Miguel, Azores, using stochastic optimization

• The scheduling model is used for the sizing of a pumped storage power plant.
• All variables are interrelated in inter-temporal constraints, leading to a direct approach.
• The decision vector is separated into past and future decisions.
• The model can be immediately translated into a mathematical program modeling language.
• Forecast errors are simulated as uncertainty cones of correlated time series of growing variance.

A pumped storage power plant is being planned in the island of S. Miguel, Azores, as a way to accommodate an increasing share of renewable generation. With that purpose, an optimal schedule program of the whole system is simulated considering a moving window in a receding horizon, in which the operational rules established by the island's system operator are implemented. Given the stochastic nature of renewable sources and loads, this schedule has to be framed as a stochastic optimization problem with interrelated decision variables. The uncertainty originates from forecasting errors that are superimposed on simulated time series of load and wind generation, and this uncertainty is simulated as stochastic, normally distributed vectors of growing variance. The adequacy of the pumped storage power plant is assessed through statistical measures on lost load, lost reserve, and lost wind.