Multi-task neural diffusion processes for uncertainty-quantified wind power prediction

TL;DR

This work addresses uncertainty-aware wind power forecasting using SCADA data by introducing neural diffusion processes (NDPs) and extending them to a multi-task framework (MT-NDP) with a task encoder to exploit cross-turbine correlations. The MT-NDPs provide calibrated predictive distributions and enable few-shot adaptation to unseen turbines, outperforming single-task NDPs and Gaussian processes in calibration and, in many cases, point accuracy. The empirical evaluation on Kelmarsh wind farm data demonstrates scalability to higher-dimensional inputs and robust uncertainty quantification, with the MT-NDP showing the strongest gains for turbines that deviate from fleet norms. These results have practical implications for dispatch, maintenance, and grid operation by delivering sharper, trustworthy predictive intervals without requiring large task-specific datasets.

Abstract

Uncertainty-aware wind power prediction is essential for grid integration and reliable wind farm operation. We apply neural diffusion processes (NDPs)-a recent class of models that learn distributions over functions-and extend them to a multi-task NDP (MT-NDP) framework for wind power prediction. We provide the first empirical evaluation of NDPs in real supervisory control and data acquisition (SCADA) data. We introduce a task encoder within MT-NDPs to capture cross-turbine correlations and enable few-shot adaptation to unseen turbines. The proposed MT-NDP framework outperforms single-task NDPs and GPs in terms of point accuracy and calibration, particularly for wind turbines whose behaviour deviates from the fleet average. In general, NDP-based models deliver calibrated and scalable predictions suitable for operational deployment, offering sharper, yet trustworthy, predictive intervals that can support dispatch and maintenance decisions in modern wind farms.

Figures (13)

• Figure 1: SCADA data for the six turbines from the Kelmarsh wind farm Plumley2022 after removing standbys and warnings using the operational status and events file. (a) Wind power curves showing mean power output (kW) by mean wind speed (m/s). (b) Joint distribution of mean wind speed (m/s) and mean wind direction (°) for turbine 1 (left), and mean power output (kW) by mean wind speed (m/s) and wind direction (°) for turbine 1 (right).
• Figure 2: SCADA data for the six turbines from the Kelmarsh wind farm Plumley2022 after removing standbys and warnings using the operational status and events file. (a) Mean power output (kW) by mean nacelle temperature (° C). (b) Mean power output (kW) by mean transformer temperature (° C).
• Figure 3: Illustration of the forward diffusion process. Gaussian noise is added incrementally at each timestep under a cosine variance schedule, progressively degrading the structure of the data until it converges to white noise. GP= Gaussian process, CI= Confidence interval
• Figure 4: Comparison of unconditional and conditional reverse diffusion sampling. We sampled 10 reverse trajectories per input location, plotting the mean of these. Conditional sampling provides a closer fit to the true data across all timesteps. GP= Gaussian process
• Figure 5: Forward and backward trajectories for a single input point (a) conditional sampling and (b) unconditional sampling. For both (a) and (b) in the forward process (blue), Gaussian noise is gradually added. In the reverse process (red), noise is removed, converging towards the true value with decreasing uncertainty. The shaded area shows one standard deviation.