Interest in sports predictions as well as the public availability of large amounts of structured and unstructured data are increasing every day. As sporting events are not completely independent events, but characterized by the influence of the human factor, the adequate selection of the analysis process is very important. In this paper, seven different classification machine learning algorithms are used and validated with two validation methods: Train&Test and cross-validation. Validation methods were analyzed and critically reviewed. The obtained results are analyzed and compared. Analyzing the results of the used machine learning algorithms, the best average prediction results were obtained by using the nearest neighbors algorithm and the worst prediction results were obtained by using decision trees. The cross-validation method obtained better results than the Train&Test validation method. The prediction results of the Train&Test validation method by using disjoint datasets and up-to-date data were also compared. Better results were obtained by using up-to-date data. In addition, directions for future research are also explained.

Thermoelectric modules are used for heating or cooling if an electric current flows through them. If one side of such a module is heated and the other side is cooled, the outputs show a potential difference that creates the current flow through an enclosed circuit. The paper describes briefly the physical basis of the operation of semiconductor thermoelectric modules and their application for generating electricity as thermoelectric generators. The experimental measurements were made with different available modules of thermoelectric generators. The analysis of obtained results has provided the conclusions about the possibility of their application as sources of electricity. Although the efficiency of thermoelectric modules is still quite low, the latest research reveals the possibility of the use of materials that make modules more efficient. Considering the fact that they do not contain moving parts, they are characterized by reliable, long-lasting work and simple operation and maintenance.

This paper proposes forecasting methods for Wind Power Plants` (WPPs`) generation based on Machine Learning Algorithms. In order to increase the precision of generation forecasts from WPPs further, the methodology is introduced of organising WPP owners in a so-called "ECO balance group". The described theoretical bases have been applied to the WPPs in the Croatian transmission power system. Deviation calculations were made for the forecasted and realised generation of WPPs in 2015, when the proposed methods were not effective, and for 2019 when these methods had already been applied.

This paper proposes several probability distribution models for wind- and net-load-power forecast error, i.e., a generalized normal model (GN) and extended skew-generalized-normal models (ESGN, SGN and SN). They are all compared to different models known from the literature (beta, Weibull and gamma). The evaluation of the discussed probability distribution models shows that the ESGN model presents better fit for probability and cumulative distribution functions.

This paper proposes an approach to schedule and allocate regulating reserve in systems with a large share of generation from wind power plants (WPPs). The method for scheduling regulating power is based on the correction of day-ahead schedules planned by the known variations of the load- and wind-power from the previous hour. The correction term incorporates the tendency of changes for previous hours. In this way, more of the regulating power can be reserved during sudden changes in the intensity of the wind. Hence, available existing operational data are required, instead of historical data. The method for allocation of regulating power to regulating units is based on minimization of the transmission losses with optional correction according to economic criteria. The qualitative analysis of the relevant economic features is introduced, as well as the possibilities for applying advanced information and communications technology. The methods discussed were tested for the standard test network RTS 96 and for the Croatian transmission network. Test results showed considerable reduction of the unnecessary positive and negative regulating reserve with increased safety compared to the classic method. Furthermore, optimal allocation of regulating reserve reduced transmission losses for all the cases discussed. Reductions were more significant when the layouts of WPPs and regulating capacities were more geographically dispersed.