Conference programme

Back to the programme printer.gif Print

Delegates are invited to meet and discuss with the poster presenters during the poster presentation sessions between 10:30-11:30 and 16:00-17:00 on Thursday, 19 November 2015.

Lead Session Chair:
Stephan Barth, ForWind - Center for Wind Energy Research, Germany
Siniša Knežević Energy Institute Hrvoje Pozar, Croatia
Hrvoje Keko (1) F Nikola Karadza (1) Laszlo Horvath (1) Sinisa Knezevic (1) Kristian Horvath (2) Alica Bajic (2) Josip Zivkovic (3) Bojan Rescec (3)
(1) Energy Institute Hrvoje Pozar, Zagreb, Croatia (2) Croatian Meteorological and Hydrological Service, Zagreb, Croatia (3) RP Global Croatia, Zagreb, Croatia

Share this poster on:

Printer friendly version: printer.gif Print

Presenter's biography

Biographies are supplied directly by presenters at EWEA 2015 and are published here unedited

Siniša Knežević is currently a senior researcher at Energy institute Hrvoje Požar in Croatia. His main occupations are wind studies. His prior employment was in Croatian electricity company on medium voltage grids.


Poster Download poster (11.51 MB)


Wind Power Forecasting Implementation Challenges: A Croatian Job


In Croatia, over the recent years the wind power integration has been a success story: with a feed-in tariff regime in place the total installed wind power reached 346 MW, with further 81 MW operational by 2016. The deployment of advanced forecasting algorithms has not been the story of similar success due to feed-in regime stipulations. Previous wind integration studies show that, without no actions taken and 400 MW of wind power online, the Croatian grid already operates close to safe-side limits. Thus unless the grid integration challenges are tackled, considerable resources would remain untapped. Well-functioning forecasting enables further integration in a cost-effective manner, however it is challenging: the Croatian region is prone to severe downslope bora windstorms, with hurricane force winds, high variability, frequent ramps and high turbulence. The most resourceful region has very difficult terrain with local specifics not well captured by mesoscale models. Most of Croatian wind farms are exposed to fairly similar conditions. This work describes the experiences gathered in a collaborative effort during the implementation of a short-term wind power forecasting algorithm.


The work presented here is a collaborative effort and an extension of the work in the WILL4WIND project. The efforts in this work extend the scope towards wind power forecasting, i.e. the WILL4WIND NWP model chain is extended with a customized wind-to-power statistical forecasting tool. The actual data collected from a 43 MW power plant in Šibenik region in Dalmatia are utilized, and the paper delivers the implementation experiences.

Main body of abstract

This work is based on wind-to-power principle: taking the relevant outputs of the NWP forecasting system as inputs, it delivers the forecasted values of wind power. As a statistical system, it models the relation between NWP predictions, the current measurements and forecasted wind power. As with all statistical methods, the physical phenomena are not considered directly. To increase the training robustness, information theoretic learning (ITL) principles are used. This means no assumption of the wind power forecast error distribution being Gaussian is made. Thus the mean square error criterion is not used since it requires normality of forecast error distribution as an underlying assumption to operate optimally. The authors’ experience is that this approach delivers robust training and less susceptibility to performance degradation when facing outliers. The system is designed as an (adaptively trained) neural network, with custom filtering modules. The implementation details will be described in full paper. The system has been tested in several configurations. Given the difficult terrain and the typical modelling of mesoscale terrain in NWP models, further efforts have been devised towards including the physical characteristics through spatial refinement. This means, although the main method used in this work is a statistical W2P method, a part of the presented effort is directed towards including the physical characteristics of the local terrain. The final part of the paper evaluates the obtained results, concludes with assessment of the forecasting system performance and delivers the most relevant directions for the further research on advanced wind power forecasts for the Croatian wind power plants.


In this work, experiences from a collaborative effort on implementation of a short-term wind power forecasting system in Croatia are presented, in the context of Croatian difficult terrain and complex wind regime. The work describes the challenges and indicates the most promising directions to improve the performance of current systems, which is especially valuable in light of the further expected developments in Croatian wind power sector.

Learning objectives
In this work, the implementation of a custom statistical W2P tool for a wind power plant in Croatia is presented. The work extends the previous research efforts. Considering the current situation of wind power in Croatia, it is reasonable to expect the value of forecasts in Croatia will increase, both for the developers and the system operator.