Optimizing Energy Dispatch Leveraging Uncertainty Information in Short-Term Wind Power Forecasts

Abstract

To adapt to the increasing share of renewable producers in power systems, the energy dispatch has to be redesigned to consider the intermittent nature of renewable energy sources and its predictability. A stochastic dispatch model was tested for optimizing the energy dispatch in the day-ahead market, which implements expected balancing costs using ensemble forecasts in a two-stage stochastic market clearing. It was compared to a conventional dispatch model, which bases its day-ahead market clearing on a deterministic forecast. The impact of the level and spatial distribution of generator flexibility and of link capacities was analyzed for both models. The results of the thesis show that the stochastic dispatch model improves the overall market performance. It decreases total system costs from 2.72 €/MWh in the conventional dispatch model to 2.38 €/MWh equalling yearly total savings of 47.5 Mio €. It decreases curtailment slightly by 4000 MWh from 184.4 × 105 to 184.0 × 105 MWh. It significantly improves shedding by reducing the total amount of shedded load by 317000 MWh from 318034 MWh in the conventional model to 1448 MWh with a total load of ca. 140 × 106 MWh. Increased generator flexibility level reduces the total amount of shedded load significantly for both models, where it is ideal to locate the generator with the highest generator flexibility at the bus with the highest load. Increasing link capacity of links connected from a bus with a high share of wind power to a bus with a high load, decreases overall curtailment. This results in increased system costs for the conventional model due to increased shedding as the deterministic forecast is less accurate in predicting sudden downward ramps in wind power, which trigger shedding. In the stochastic model total system costs are decreased as it utilizes successfully more wind energy with marginal costs of 0 €/MWh using the ensemble forecast.