Artificial Neural Networks (ANNs), certain forms of computer logic and architecture inspired by biological neural networks like ones we have in our brain, are a tool of vast potential of real-world applications. ANNs are definitely not limited in their use only as a medium to investigate theoretical concept of artificial intelligence – these computational structures are being successfully applied to the real-world tasks such as automated decision making in finances or medicine, pattern recognition in various biometric systems, robotics, internet data mining and even e-mail spam filtering. So it’s probably not a big surprise to see artificial networks in the field of electrical energy production and distribution.
In a paper recently published on arXiv.org, a team of scientists investigated a possibility to apply the artificial neural network approach to forecast the consumption of electric energy. The authors or the study argue that ANNs have already demonstrated their potential to predict the state of future events and optimize available resources. “For effective planning and operation of power systems, optimal forecasting tools are needed for energy operators to maximize profit and also to provide maximum satisfaction to energy consumers”, the scientists say. Financial aspects aside, the energy production and distribution system security is another important aspect to predict and includes load forecasting and in-advance detection of system vulnerabilities.
How difficult could it be to forecast the amount of electrical energy consumed? This quantity and its dynamics may seem quite obvious, as the season-related causes are the first things that come to mind when thinking about our energy bills. However, such forecasting is greatly associated with several additional factors: historical grid load, mean relative air humidity, population living within a particular geographical area, even GDP per capita and probably even more. In this light the forecasting becomes much more difficult.
The study presented by the team was based on monthly data for electric energy consumed in the Gaza strip (Palestine) collected from year 1994 to 2013. The energy demands in this region are often barely met, and because of this reason the Gaza strip served as a perfect ground to investigate efficient utilization of limited energy resources.
A feedforward type of artificial neural networks was selected to implement a particular forecasting model. The proposed model was validated using 2-Fold and K-Fold cross validation techniques and several error criteria were applied to evaluate the accuracy of ANN-based prediction. The authors say that after testing the developed model with the real-world data the results “had good accuracy and showed that the proposed ANN model can be used to predict future trends of electric energy consumption”. The paper does not present many technical details of how the forecasting model was implemented and this is a little bit disappointing, but let’s hope we will see this development applied in actual electrical grids.
Written by Alius Noreika