Optimal wind farm energy and reserve scheduling incorporating wake interactions

This paper proposes a novel approach for optimal energy and reserve scheduling of wind farms by explicitly modelling wake interactions to enhance market participation and operational efficiency. Conventional methods often neglect wake effects, relying on power curve estimations that represent an upper limit and reduce market performance. To address this, a two-stage stochastic programming framework is developed, integrating a wake-aware power estimation model within the FLORIS simulation software. Wind and reserve uncertainties are addressed through scenario generation and reduction, enabling wind power producers to optimise participation in day-ahead energy and ancillary services markets, with particular focus on the Frequency Restoration Reserve (FRR). The wake-aware model provides more realistic power output predictions based on site-specific wind and atmospheric conditions, improving scheduling accuracy and reducing imbalance penalties. Wake steering is further employed to mitigate wake-induced losses and increase income through participation in ancillary services. The proposed approach is evaluated through a case study of the London Array offshore wind farm participating in the Great Britain (GB) electricity markets. Results show that conventional methods estimate production 12-13% higher, leading to imbalance penalties and 3% lower revenue compared with the wake-aware approach accounting for wake interactions. Moreover, the steering-enhanced approach yields an additional 1-2% increase in income relative to the wake-aware baseline. These findings underscore the value of accounting for wake interactions in wind farm scheduling and demonstrate the economic and operational benefits of active wake management, offering insights for improving grid stability and profitability as wind penetration continues to rise.

💡 Research Summary

The paper introduces a novel two‑stage stochastic programming framework that explicitly incorporates wind‑farm wake interactions into day‑ahead (DA) energy and ancillary‑service (specifically Frequency Restoration Reserve, FRR) scheduling. Traditional approaches either rely on turbine power‑curve based “Power Curve” models, which ignore wake losses and therefore over‑estimate farm output by 12‑13 %, or on “Baseline” forecasts that include wake losses but do not exploit wake‑management techniques. Both lead to sub‑optimal market bids and potentially large imbalance penalties.

To overcome these shortcomings, the authors embed a wake‑aware power estimation model within the FLORIS aerodynamic simulator. The model uses turbine layout, wind direction, wind speed, and turbulence intensity (TI) to compute wake‑induced speed deficits and turbulence, yielding a realistic power forecast for each scenario. Scenario generation captures wind speed and wake uncertainties; dimensionality reduction and clustering produce a tractable set of representative scenarios for the stochastic program.

In the first stage, the optimizer decides on DA energy bids (Pᵉ) and FRR capacity bids (P↑ᶠʳ) for each hourly interval. The second stage adjusts these commitments through redispatch variables (ΔP, ΔPᶠʳ) once the actual wind state is realized, thereby internalising imbalance settlement costs (energy and FRR imbalance prices). The objective maximises expected total revenue, comprising market energy and reserve payments, availability/utilisation fees, and penalties for deviations.

A key innovation is the inclusion of wake steering as a controllable decision variable. By yawing upstream turbines, the wake is deflected away from downstream units, reducing wake‑induced losses. Although yawing slightly reduces the power of the yawed turbine, the overall farm output increases, translating into higher market revenues. The model optimises yaw angles jointly with energy and reserve bids.

The methodology is validated on the London Array offshore wind farm (≈630 MW, 175 turbines) participating in Great Britain’s electricity markets. Using real wind data, market prices, and imbalance settlement rules, three strategies are compared: (i) Power‑Curve (optimistic, no wake), (ii) Baseline (conservative, wake losses only), and (iii) the proposed wake‑aware, steering‑enhanced approach. Results show that the Power‑Curve method over‑predicts production by 12‑13 %, leading to imbalance penalties that reduce total revenue by roughly 3 % relative to the wake‑aware baseline. The wake‑aware model eliminates most of these penalties, delivering the highest revenue. Adding optimal wake steering yields an additional 1‑2 % revenue increase over the wake‑aware baseline without steering.

The contributions are threefold: (1) a realistic wake‑aware power estimation integrated into market bidding, (2) the joint optimisation of wake steering and market participation, and (3) an extension of the analysis from primary frequency control (FCR) to secondary control (FRR). The work bridges the gap between aerodynamic research and electricity‑market optimisation, demonstrating that accounting for wake interactions—and actively managing them—can materially improve both operational efficiency and profitability of wind farms as their penetration grows.

Limitations include the focus on DA markets only, the exclusion of intra‑day re‑dispatch, and the lack of network constraints or multi‑farm interactions. Future research directions suggested are real‑time (intra‑day) optimisation, simultaneous scheduling of multiple ancillary services, and scaling the framework to clusters of wind farms with grid constraints. Overall, the paper provides a solid foundation for integrating physical wake dynamics with stochastic market optimisation, offering a pathway toward more reliable and economically sustainable wind‑farm operations.