Optimal unit commitment and dispatch for wind farm operations

The variability of wind poses a challenge for wind farm operators to enable a reliable source of electricity that can be integrated with other generating capabilities. Balancing this situation with the inherent latency in changing the production of electricity generated by conventional fossil-fueled or nuclear facilities adds further complexity to the problem. To begin to address this situation, we outline an approach to optimizing both unit commitment and dispatch.

Conventional generators require time and expense to power up and have limits on how quickly they can change their production levels. Unit commitment concerns the timing of turning such generators on and off while dispatch is determining how much power to produce from each generator. To hedge against unexpected load fluctuations and generator outages, system operators keep a certain amount of spinning reserve, generators that are on but producing at minimal levels that can increase production quickly. Intermittent generators require more spinning reserves, the cost of which detracts from their economic efficiency. Hence, more accurate and timely unit commitment and dispatching of power generation assets can have significant impact on the viability of wind farm operations.

Our approach involves the coupling of an non-hydrostatic numerical weather prediction (NWP) code operating at the meso-gamma scale to optimization software. In particular, the WRF-ARW community model can be configured with a focus on the region of a wind farm with appropriate physics to include sufficient vertical resolution in the boundary layer to capture details in the regime swept out by turbine blades. Data generated by the NWP code can be used to define a range of potential wind forecasts to create a scenario tree to estimate the uncertainty in the prediction. Since this is essentially a linear programming problem with uncertainty in both supply (i.e., wind) and demand, we use stochastic optimization. This leads to building recourse for each wind scenario, which leads to a reduction in the volume of spinning reserves as well as unmet demand and overall cost, regardless of the variance in the wind forecast. We will illustrate this coupled physical-optimization approach with example ramping events, where the predictive optimization has sufficient lead time to improve both dispatch and commitment.