Our Power Forecasting services allows renewable asset owners to reduce the cost of variability of the power output of their large scale wind- and solar farms. It produces highly accurate and reliable short term (intra-day, day-ahead and week-ahead) power forecasts based on asset telemetry and numerical weather forecasts.
We use deep learning models with probabilistic distributional modelling trained on a variety of weather forecasts and historical power to ensure optimal performance. Our models don’t come out of a formatted pipeline. We handcraft them. We know different customers will have different requirements.
Accurate short term forecasting of the power output from renewable assets is crucial in mitigating the effects of their innate variability. In this example, the deployment of our Power Forecasting reduced the imbalance costs and penalties incurred by a large utility by 15% across the assets on which it was deployed. Higher accuracy leads to a lower Normalized Mean Average Error, and a probabilistic forecast rather than a deterministic one leads to increased confidence and better trade-offs when the power volume is brought to the market.