M.K. Gill, D. Moon, B. Lee, and D. Elliott (USA)
Evolutionary computing, Particle Swarm Optimization, Machine learning, Support Vector Machines, Wind Power, Optimization
The training of learning algorithms poses a difficult task for high dimensional problems encountered in system operations. It is not only time-consuming, but also in some cases can result in sub-optimal models. The current paper describes an efficient and effective method for training a widely used learning algorithm, the support vector machine (SVM). The proposed methodology employs an evolutionary computing strategy based on swarm paradigm called “Particle Swarm Optimization (PSO)” to devise an intelligent predictor. The use of an automatic optimization scheme allows picking a specific objective criterion during the training process and can accordingly redirect the search in the “region of interest”. The PSO-SVM predictor is applied and tested on two different datasets within North America, for predictions of Wind Power using National Center for Environmental Prediction (NCEP)’s global reanalysis gridded dataset.
Important Links:
Go Back
