Team

Local short-term weather forecasts are an important component in day-ahead energy markets, where renewable energy production has a large role. BCDC Weather will develop local weather forecasting methods to support the planning of the new market mechanisms in the BCDC Market team. The goal is to develop the Numerical Weather Prediction (NWP) model, Harmonie, from the wind and solar energy production point of view.

Sami Niemelä, the head of Meteorological Research Unit, at the Finnish Meteorological Institute (FMI), will lead the BCDC Weather team. The team consists of professor Anders Lindfors, postdoctoral researchers Evgeny Atlaskin and Laura Rontu and doctoral students Herman BöökTuuli Perttula and Karoliina Hämäläinen.

BCDC Weather combines expertise on weather modelling, remote sensing observations and renewable energy in order to support the overall goals of the BCDC Energy project.

Forecasting renewable energy production requires high resolution NWP-models. Such models can simulate and predict small scale weather phenomena, such as flow structures forced by variable terrain or land-sea contrasts. BCDC Weather is focusing on the aspects of weather forecast development that are most useful from the wind and solar energy production points of view.

In the picture: Sami Niemelä. Photo: Kati Leinonen.

Observations

A successful weather forecast requires that initial conditions provided to the forecast model are accurate and consistent with available observations. From the point of view of short-term (hours to days) energy production forecasts, high-resolution remote sensing data are currently not used in an optimal way. This task will focus on developing methods for using wind and cloud information from weather radars and satellites in order to improve the forecasting of wind and solar energy production.

FMI’s radar network provides data on both precipitation areas and their movement with high spatial and temporal accuracy. In addition, radial wind information can also be obtained thanks to the Doppler feature of the radars. The new data is expected to improve the wind and cloud forecasts in day-ahead time scales.

In non-precipitating cases, satellite information can be used for extracting wind information. Atmospheric Motion Vectors (AMVs) are derived by tracking subsequent features from satellite imagery. AMVs from geostationary satellites are widely used in tropical and mid-latitude areas; however, their quality in the high-latitudes is poor.

BCDC Weather will explore how AMV-data from polar orbiting satellites can be utilize to improve the initial conditions. Furthermore, the most critical component for prediction of solar energy production is loud forecasting. Although the current models are using satellite derived temperature and humidity profile information in the data assimilation process, the actual cloud information is not widely used. BCDC Weather aims to improve the data on initial moisture fields by assimilating cloud mask data from polar orbiting satellites in a thermodynamically consistent way. This is expected to improve cloud forecasts during the first 24 hours.

Modelling methods

Wind and solar energy forecasting have special characteristics that require further development of the physical parameterization of the NWP model. Furthermore, the northern location of Finland poses challenges both to energy production and weather forecasting. Production losses due to ice formation should be taken into account in the wind energy production forecast. BCDC Weather will assess the icing risk to wind turbines by integrating the in-cloud icing methodology with the Harmonie model and develop methods for short term forecasting for predicting wind energy production loss.

The second most important component in solar energy production forecasting is the effect of atmospheric aerosols. Current operational NWP models do not directly take into account the weather-dependent aerosol related attenuation of solar radiation. The aim of the BCDC Weather team is to evaluate the direct radiative effect of aerosols on solar energy production forecasts in Finland. We will use already existing aerosol optical depth data in the Harmonie radiation parameterization scheme.

Cloud computing algorithms, which will drive the energy market system, require solar and wind energy production forecasts locally. BCDC Weather will develop automatic conversion methods for casting the weather information in energy terms.

Forecast uncertainty

The estimate of forecast uncertainty is as important as the forecast itself. Energy system models use energy production forecasts by giving them weight according to statistical uncertainty estimates. In reality the forecast uncertainty depends on the weather itself. BCDC Weather will develop flow dependent methods for forecast uncertainty. Furthermore, in cooperation with our partners we will explore the best practices to utilise uncertainty estimates in cloud computing algorithms to control local power flow in the most efficient way.

The principal of probabilistic forecasts

Animation/Figure: The red line represents the traditional deterministic forecast, where initial conditions are estimated as accurately as possible and single forecast realization is performed. However, both observation process and modelling methods always contain approximations leading to forecast errors. The effect of observation and model uncertainty can be estimated by running an ensemble of model realizations where the observation and model errors are simulated by perturbing the initial conditions and model integration, respectively.

Integration of new methods

Weather information is a critical data source for energy markets with intermittent energy production and related cloud computing algorithms. Energy markets also need near real time power production information from the production units. Such information can be used to further improve weather and power production forecasts.

BCDC Weather will develop forecast calibration methods, which can use real wind and solar energy production data. Furthermore, very short-range (0-6h) forecasting will be improved by combining the real-time production statistics from nearby micro-grids. In this way, solar panels from individual households and wind parks may form new measurement networks also providing weather related information.