Optimized Wind and Solar Power Forecasts For India

2nd Int'l Conference on Large-Scale Grid Integration of Renewable Energy in India| New Delhi, India | 4-6 Sep 2019
Optimized Wind and Solar Power Forecasts for India
Hans-Peter Igor Waldl Detlev Heinemann

Overspeed GmbH University of Oldenburg
Balaraman Kannan Giridhar, G.

NIWE, India NIWE, India
Contact: igor@overspeed.de
Abstract— India has very ambitious plans of increase of wind spread out over all states, so a proper analysis of the clear
and solar generation for the upcoming years. In order to sky behavior in different climatic regions support model
maintain or even improve the stability of the electrical grid, development, and the final tuning of these models. (see also
accurate forecasts of wind and solar power are an important [1,2])
building block for a sustainable future. Currently, many
approaches are transferred from countries like Germany,
Denmark, and the US in order to offer prediction services for
Indian wind and solar power farms.
Without doubt, this is a good starting point. Nevertheless, a lot

of conditions are quite different from the situation in the
mentioned countries. Climatological conditions like rainy
season and high and low wind seasons must be taken
into account during tuning of prediction models. Influences
like haze, aerosols and soiling demand for extended solar
power forecasting models. Quite strong thermal gradients in
the atmosphere must be modelled differently from conditions
in the moderate latitudes for predicting the power output from
wind farms.
Figure 1: Figure 1: Representation of Binned irradiation bias as
Finally, we will shed some light on the current Indian function of clear-sky index and cosine of zenith angle compared to
regulations with respect to forecasting and will propose some SRRA data. [3]
potential developments for the future.
Another big impact on PV production is the soiling of
Keywords- Solar forecasting, Wind forecasting, Day ahead, PV panels by dust and dirt. Model development and
physical, statistical,intra-day, India improvement is ongoing in at least two studies carried out
by NIWE and the German company Suntrace, partially
I. INTRODUCTION
funded by the German giz.
II. SOLAR POWER PREDICTIONS With respect to very-short term predictions (30 minutes
In India, the influence of haze and aerosols is much to 4 hours), an approach based on Indian weather satellite
bigger than in regions like Europe and the US. This data is under development at NIWE and overspeed. (Figure
naturally has a significant influence on the irradiation which 2 and Figure 3)
can be used by PV modules. In addition, the day-to-day
variations of the so-called clear sky irradiation are much
bigger than in countries with less aerosol impact. In order to
gain excellent solar power forecasts for India in general, the
tuning of clear sky models must be adapted. In addition,
existing forecasts of aerosols are tested in order to improve
the day-to-day forecasts quality even further.
A big advantage in India is the availability of an
irradiation measuring network with high quality
measurement stations (SRRA), which is operated by
NIWE. More than 100 irradiation measurement stations are
In general, forecast deviations of a single farm do no

harm to the power system (of course, depending on the size
of the farm; this doesn’t hold for Gigawatt farms). For that
reason, in most countries with high penetration from wind
and solar, there is the possibility to deliver forecasts for a
complete pool of wind and solar farm. The power
production of the portfolio(s) is summed up by specialized
companies called aggregators, similar to aggregators for
pooling stations in some Indian states. Prediction products
are then delivered on the aggregated levels.
This opens the opportunity to ask for higher prediction
accuracy (due to smoothing effects), stimulates investments
in better forecasts (like better FSPs and higher quality of
measurement data, and is closer to the needs of the energy
system operations.
Last, but not least, a good quality of the measured data
from wind and solar farms is crucial for an excellent
prediction process. The occurrence of mixed power
Figure 2: Insat 3D satellite. (source: isro.gov.in)
readings in some states, combining wind, solar and
load, leads to the application of quite complicated correction
and upscaling methods with limited accuracy. One proposal
III. WIND POWER PREDICTIONS for future regulations could be an incentive for delivering
reliable data to the dispatch centers/REMCs for every single
With respect to wind power predictions, there are mainly farm, including signals of current power production,
two points which must be taken into account in India. downregulation set-points, and turbine or inverter
Compared to countries at moderate latitudes, the availability.
influence of the thermal stratification of the atmosphere (i.e.
the difference in temperature at different heights) is much
bigger. This must be taken into account in the model chain Thanks are due to the wind and solar forecasting teams at
by a more accurate model approach for these effects. As a NIWE, Oldenburg University and overspeed, German giz,
follow up, this dependency on the atmospheric conditions SGS India, and suntrace.
leads to higher amplitude of daily patterns and extreme
events/ramps of power production. Part of the presented results are gained within a R&D and
Today, most prediction systems work on an adaptive capacity building project together with the National
basis. This means that the tuning of models is updated Wind Energy Institute NIWE, and funded by the German giz.
permanently with the latest measurement data. This helps to
take into account slow changes in conditions, like the VI REFERENCES
transition from summer to fall to winter. In areas with
Monsoon influences, this approach may lead to big forecast [1] Sahana, L., et al.: Impact of Soiling on Energy Yield of
deviations at the beginning of the rainy season. The models Solar PV Power Plant And Developing Soiling Correction
then have been trained adaptively on low-wind data for up Factor for Solar PV Power Forecasting.
to 9 month, and the training for the high wind season may be This conference.
poor. For that reason, the model training approaches must be
optimized for the situation of high- and low-wind seasons. [2] Ramanan, A., et al.: Integrated regional solar PV
(Similar arguments hold for the training of solar power forecasting method by hybrid method. Case Study of an
prediction models. Indian State.
IV. PROPOSALS FOR FUTURE REGULATIONS This conference.
Currently, the Indian regulations for the power sector [3] Roy, A., et al.: Day Ahead Solar Pv Power Forecasting
demand a forecast for each single wind farm. This is not a Based On A Combination Of Statistical And Physical
bad approach, but the combination with penalties for
ModellingUtilizing Nwp Data For Solar Parks In India.
forecast deviations on a single farm level limit the
possibility of improving the prediction quality. This conference.
EL S 6
Consecutive satellite images
Motion Vector Field Cloud
Figure 3: Principle of deriving solar power prediction from consecutive satellite images, producing cloud motion vectors, from which the
irradiation forecast is derived.
Figure 4: Soiling analysis at NIWE [1].

Optimized Wind and Solar Power Forecasts For India

Uploaded by

Copyright:

Available Formats

Optimized Wind and Solar Power Forecasts For India

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Optimized Wind and Solar Power Forecasts For India

Uploaded by

Copyright:

Available Formats

2nd Int'l Conference on Large-Scale Grid Integration of Renewable Energy in India| New Delhi, India | 4-6 Sep 2019

Optimized Wind and Solar Power Forecasts for India

Hans-Peter Igor Waldl Detlev Heinemann

Balaraman Kannan Giridhar, G.

Without doubt, this is a good starting point. Nevertheless, a lot

In general, forecast deviations of a single farm do no

Motion Vector Field Cloud

Figure 4: Soiling analysis at NIWE [1].

You might also like