Electricity producers rely on forecasts of electricity prices for bidding in the markets and power plant scheduling. Markets are changing: A much tighter integration between European markets and a rise in unregulated renewable energy production, especially wind and photo-voltaic, call for joint probabilistic forecasts. Incorporating the transient interplay between productions from renewable sources is critical to power production and financial operations. Multivariate probabilistic forecasts of electricity prices in the short horizon are required.

Appropriately characterising multivariate uncertainty will enable more effective operational decisions to be made.

Conventional power grids add extra generation and distribution capacity. Smart grids actively match energy supply and demand and combine the needs of the markets with the limitations of the grid infrastructure. With the implementation of smart meters and grid sensors, enormous amounts of time series data are generated, with seconds resolution. Our objective is to develop new methods that extract the right information from data to optimise grid control and for real time operation.

Temperature localization for consumption forecasting

A critical component of any market model for the Nordic region is temperature, as Nordic homes are heated with electricity. Throughout Big Insight, we have worked on methodology to improve forecasts of electricity consumption, ultimately replacing a neural network with a more accurate method based on principal component regression, which was completed in 2021. This model operated on Nordic-wide temperature fields and ultimately forecasted the entire system consumption. Our main effort in 2022 was to “locallize” this framework, by looking at the individual regions of the Nordpool market and associating them with temperature sub-fields. We ultimately found that a combination of Nordic level and local level temperature fields gave the best forecast of region-level consumption. This model was subsequently put into production at Hydro.

Medium-term renewable production forecasting

Having an accurate forecast of renewable energy production is an important component of a market model, since high or low levels of production can have outsized impact on electricity prices. In 2021 we developed a ridge regression framework that substantially improved upon previous models for forecasting these quantities based on weather forecasts. After implementation at the industrial partner, it was discovered that performance declined substantially at the 11 to 15 day-ahead level, occasionally leading to unintuitive behaviour. We investigated these outcomes and concluded that weather forecasts at this time horizon must be pooled to offer any acceptable performance. This adjustment to the modelling framework was then implemented at Hydro.

Solar production forecasting in “shoulder” seasons

As noted above, a critical component investigated in the last years of BigInsight was forecasts of renewable energy production. Accurately forecasting solar power production proved particularly difficult in the beginning of spring. This is because the models that had been developed are trained on recent historical data, typically on an hourly basis. When a given hour in the day goes from being dark to having some illumination, the industrial partner noticed unstable behaviour of the solar production model. We investigated this phenomenon and concluded that it was due to instability in the splines used in the generalized additive model that forecasted solar production. We proposed a pooling mechanism which addressed this issue. The updated model was then implemented at Hydro.