The WindFor base module consists of the WindFor engine; a time controlled system with automatic retries which controls the internal and external data flows. The WindFor base module also includes the well proven WindFor power curve and dynamical models, which are some of the most accurate forecast models using standard meteorological forecast input.
WindFor Optimal Forecast Combination
Optimally combine MET forecasts from several providers in order to increase forecast performance and operational robustness. The module continuously measures forecast performance and correlation of forecast errors for the individual meteorological forecasts and based on this information an optimal weighting is derived and used to generate the final forecast.
WindFor High Resolution Forecasting
The module produces forecasts with high resolution in time, e.g. forecasts for time intervals of 5 minutes or even less. On such short time intervals it has been observed that in some situations the wind power production is quite stable over time, whereas in other situations the wind power production is quite variable.
This variability can be measured and used to characterize the current situation. Based on the current variability of the wind power production and the meteorological (interpolated) wind speed forecast, the model derives the optimal weighting between a simple (interpolated) wind power forecast and the latest observed wind power production.
WindFor Quantile Regression
Using regular meteorological forecasts the WindFor quantile regression module produces forecasts of the future situation specific uncertainty as full probabilistic forecasts communicated as quantile forecasts. As an example, the 20% quantile forecast is the level for which the actual production is lower in only 20% of all cases and the 80% quantile forecast is the level for which the actual production is higher in only 20% of all cases.
Opposed to ensemble based quantile forecasts, the quantile forecasts based on regular meteorological forecasts are preferable within the first few days where the ensemble spread is known to be unrealistic small.
WindFor Ensemble Forecasting
Using meteorological ensemble forecasts the WindFor ensemble forecasting module produces forecasts of the future situation specific uncertainty as full probabilistic forecasts communicated as quantile forecasts. Opposed to forecasts based on quantile regression, the ensemble based quantile forecasts are preferable for longer horizons. Due to a number of factors related to scale and the particular methods behind meteorological ensemble forecasting, the quantiles cannot be derived from the meteorological ensemble forecasts in a simple way. Instead statistical methods are applied in order to ensure the reliability of the quantile forecasts.
This module is under development. However, the models and meteorology are fully developed. Only outstanding issue is full integration into the WindFor framework.
WindFor Scenario Generation
Based on quantile forecasts this module produces scenarios of the future development in the wind power production which obey the quantiles and for which the correlation between time steps and farms/regions are based actual observations. The correlation is measured on an appropriately transformed scale.
Given adequate quantile forecasts, the method results in scenarios, which are correct in the sense that the spread and the development over time and between regions are realistic.
The module allows for thousands of independent scenarios to be generated on-the-fly. Optionally, scenario reduction techniques can be applies in order to represent the original scenarios using a few scenarios with a correct probability label.
WindFor SCADA Up-scaling
The SCADA up-scaling (SCUP) module is used for estimating the amount of production which is lost due to active curtailment. Curtailment for a wind farm occurs when using a set point which is lower than its rated power.
The SCUP module provides an estimate of a normalized power curve for a wind farm based on the most recently observed data. The module also gives a decision of whether the available data is of sufficient quality to be used for the up-scaling algorithm. The module requires measurements of nacelle wind speed, power production, curtailment level and preferably also availability.
The power curve estimation part of the module is normally run daily providing an updated estimate of a normalized power curve. This estimate is used for on-line upscaling of actual production to potential production during active curtailment. The on-line part provides estimates of potential production (estimated production with no curtailment given current availability) and down regulated production (difference between actual and estimated potential production).
WindFor Shut-down Probability Module
The shut-down probability module can be used in order to warn operators / traders about the possibility of shut-down events. This is quantified by specifying the probability of such events for each hour in the forecast horizon. Besides the standard WindFor inputs (installed capacity, availability, actual production, and met. forecasts) the module needs the manufacturer’s power curve as input.
Note however that measurements of actual wind speed are not required. The module internally adjusts for local systematic errors of the meteorological forecast; calculates the uncertainty of the meteorological forecast and use this to calculate the shut-down probability. As actual shut-down events can be rare, observations of actual shut-down events are not required.
Complete forecasting software solutions for the energy sector
Wind power prediction for wind farms
Solar power prediction for PV plants
Electricity load prediction of electrical power systems
Load forecasting, temperature control, and optimization of district heating networks