\
logo2

ES | EN | PT

CLIMATE AND ENERGY IN BELGIUM

We can predict the generation of wind energy in Belgium. By 2020, 20% of the energy consumed in Europe should come from renewable sources (World Energy Council). Particularly, in the case of Belgium, this country should contribute in an order of 13% of its energy consumption in order to achieve the common european goal for renewable sources of energy (European Wind Energy Association y Renewable Energy Directive - European Commission). Can we predict the generation of eolic energy in Belgium based on weather data reliably? The answer is yes, based on the Big Data and models based on algorithms of Artificial Intelligence (AI) in BerecoLabs we have succeeded in doing so with less than 2% errors.

In 2012 the share of renewable energy consumed was 6.8%, showing a steady increase since 2005 when only 2.3% responded to this type of energy. Much of this increase is due to the increased production of wind power, both onshore and offshore.

Figure 1. Location of wind farms in Belgium.

 

The challenges of eolic energy

The generation of eolic energy is uncertain and difficult to store the amount of excess of energy for future. For this reason, it is desirable to have reliable predictions, to be able to guarantee the satisfaction of consumer demand. In addition, to allow the growth expected of the generation according to the objectives of each regional level.

Methodology

In BerecoLabs we are developing and applying approaches related with Big Data and the use of models based on algorithms of Artificial Intelligence (AI). After the capture of data on wind farms in Belgium and the climate (historical data and forecasts), we have identified optimal configurations of parameters for models that are manifested in greater efficiency to "learn" from the data and improve forecasts. In Figure 2 you can see a comparison between the measurements of the results obtained by models and wind power generation. It is worth mentioning that the results that we are generating consistently show errors less than 2%.

Figure 2. Comparison of measurements of wind power generation and the results of the forecast models.

 

Analysis of results

The Figure 3 shows the variation of both parameters for twelve of the models studied. It is so model that performs the best forecast is the one with two hidden layers using the hyperbolic tangent function on the first layer and the sigmoid function in the second one. This configuration of the network gets the minimum value of RMSE and maximum value of the Index of Agreement of all the models.

Figure 3. RMSE and index of agreement for some of the models studied configurations.

The paradigm of renewable energy is large and is in full development within a framework of sustainability. The guideline work presented here, exemplifies one of its many applications. There could be more of them, that share similar characteristics, being the production of energy from solar farms just one example, where the dominant variable winds but not the incidence of sunlight.

In this context, an improved prognosis allows: 1) reduce the costs of operation of the network of eolic energy supply, 2) plan the distribution of the same in order to satisfy a growing stably and 3) thus help migrate towards the renewable energy consumption. In BerecoLabs look for innovate to develop new and creative approaches to using the greater abundance of data (Big Data) and the possibility of generating capacity and most intelligent models of 'learning' to deal with some of the most important challenges of this time.