Horizontally propagating wind features have been found to be a major cause of large rapid changes in wind power for wind farms. These features are generally predicted well by numerical weather prediction systems, but can be displaced by 50-100 km relative to their eventual true position. The Wind Forecasting Software fields animations (Figure-1) allows a forecast user to visualize the propagating wind features predicted in the region of single or multiple wind farms, assess likely scenarios in wind power production from displacing these features and make decisions accordingly.
The Wind Forecasting Software demonstrated below uses numerical weather prediction system forecasts from the Global Weather Forecast Provider. The conversion of wind speed forecasts to available wind power was devised using publicly available historical wind power observation data. The demo forecasts using real-time wind power observations and turbine availability data provides the improved forecasts. The large rapid change alerts can be tuned to meet specific forecast user needs.
The risk of incurring Unscheduled Interchange (UI) Charges can be mitigated through accurate wind power forecasts using a Physical and Statistical approach – refining weather model inputs from a number of approved suppliers and enhancing the data.
This is achieved using High Resolution Meso-Scale weather models, Micro-Scale wind-farm flow models and via statistical routines that can adapt to the measured conditions. This means that wind forecasts can be delivered any time for Energy Forecasting and Scheduling.
Together with our Forecasting MNC, we have processed our trial run for testing wind forecasts on real time basis.The figures given below shows our study from the data from trial run, available from one of the Wind Farms. Shown below is the daily data inputs recorded which plots the Actual Generation with respect to the Forecast parameters.
The quality of the wind power forecast is very much depended on the wind speed, direction and the portfolio size of the wind Farm. In India, we have the challenge for the wind power forecast, which is highly uncertain and complex. For a single plant and portfolio, we face the errors in wind power forecasting as shown in the figure above for a time series in intra day forecast on a particular day.
At times there are quite high variations in the production due to local winds. However the predictability and the challenges are depending on the area and with the possibility that the re-nomination have a bigger positive impact on the wind as the changes are still on a larger time scale. The Indian wind power forecast quality shall be very much dependant on season and area.
One revision for each time slot of 3 hours starting from 00:00 hours of a particular day subject to maximum of 8 revisions during the day seems to be the best possible way to achieve that the errors are not higher than +/-30%. We have been working intensively over the quarters to improve the short term forecast (0-6 hours ahead).The system needs online data, ideally every 15 minutes, in order to optimization results. Combined with accurate wind forecast the system will go on issuing a new forecast every hour.
The meteorological generation trace is what the Wind Software will use to train the parameters for transforming the wind speed forecasts to wind power. After the meteorological wind power is predicted, it will be scaled down with a prediction of how much capacity is expected to be available.
As per our experience the best mix are models using the two leading global weather models as boundary for Regional Meso-scale models. In order to predict the local phenomena we always use two Meso-scales together.