Research On Full Electric Propulsion Ship Short-time Power Load Forecasting

As the development of the full-electric propulsion ship power station's automation, the ship power load forecasting has become a research hotspot. More and more artificial intelligence methods have been applied to power load forecasting, whick make the power load forecasting become more accurate than other methods. A full-electric propulsion ship is chosen as a research object, and an improved SVM method is used to forecast its power load.Firstly, the common used short-time load forecasting methods are introduced. By analyzing domestic and foreign methods of ship short-time load forecasting, it shows the advantages of ship short-term load forecasting based on SVM. After that, this paper introduces the power system of full-electric propulsion ship, and some other technical foundations which are associated with the forecasting, such as SVM and Ant Colony Algorithms.Moreover, the basic methods of SVM, and the conception of the Kernel are introduced. Because short-time power load forecasting need get the results quickly, Least Squares method has been applied into SVM, to reduce the complexity of the problem and increase the speed. By analyzing the disadvantages of the existing methods to set the parameters, Ant Colony Algorithms is proposed to optimize the two parameters in LS-SVM.Finally, the historical data is pre-processed. After the initialization of the data, a prediction model is established to calculate the punishment parameter and Kernel parameter. Then, the short-time power load of the target full-electric propulsion ship is forecasted by ACA-LS-SVM method. After comparing the experimental result to the results get by ANN and traditional SVM methods, it demonstrates that ACA-LS-SVM method is faster and more accurate then other methods in the field of short-time power load forecasting of full-electric propulsion ship.