| Abstract [eng] |
At the final project of the master thesis the forecast of the wind speed in a small windmill park was researched. In this paper the region of Lithuania is defined as a small windmill park. Some of the major cities represent different power plant. Thus windspeed in major cities is treated as major windspeed in separate power plants. The importance of this final project comes from the object of the research which is relevant to the energy problem. New ways to harness renewable energy is searched constantly. Wind power parks are expanding. The major aim is to generate as much electricity as possible. So wind speed should be forecasted more accurate. Thus, the aim of the work was to create a novel mathematical model and compare it with one of the classical models of wind speed prediction. In the final work, a new stochastic matrix mathematical model was created, which was applied to the data of wind speed. The results obtained are compared to one of the classic methods of forecasting. In this case, the ARIMA model was chosen. After the completion of the stochastic matrix model for time series predictions, we get the values of the forecasts and the corresponding errors. Then, we found the optimal joint model of the prediction problem. For this purpose the PSO algorithm was employed. It enabled to find the optimal parameters for the joint model of wind speed prediction with the smallest error. After the calculations and the corrections made to the model, we got a short-term forecasting model wich resulted in the errors of the speed of wind through all park objects ranging from 0,005% to 72,53%. (the lowest value was received: 0,000289 and the biggest – 4,35202 in one area in Biržai). On the one hand, by looking at the percentage values of predictions errors, they have not verry high values, but knowing the characteristics of the wind data, which is chaotic in nature, the features of the wind speed and the fact that these data are hard to forecast due to their unpredictable dynamics, it can be argued that such a result is pretty good . Also if you look at the numerical values of this data, it shows us that the errors of the model is only up to 4 m/s. The objective of the research was achieved by developing a mathematical model for forecasting of the wind speed. The model developed resulted in being more accurate than the classical time series model employed (with the exception of the 2 regions). This was due to the fact that the matix model resulted in prediction erros lower than classical approach. |
