28 فروردین 1403

سجاد احمدیان

مرتبه علمی: استادیار
نشانی: دانشگاه صنعتی کرمانشاه
تحصیلات: دکترای تخصصی / مهندسی کامپیوتر
تلفن: 09188339565
دانشکده: دانشکده فناوری اطلاعات

مشخصات پژوهش

عنوان
An advanced short-term wind power forecasting framework based on the optimized deep neural network models
نوع پژوهش مقاله چاپ شده
کلیدواژه‌ها
Deep neural networks; Evolutionary computation; Neuroevolution; Optimization; Wind power forecasting
پژوهشگران سید محمد جعفر جلالی (نفر اول)، سجاد احمدیان (نفر دوم)، مهدی خدایار (نفر سوم)، عباس خسروی (نفر چهارم)، میعادرضا شفیعی خواه (نفر پنجم)، سعید نهاوندی (نفر ششم به بعد)، ژائو کاتالائو (نفر ششم به بعد)

چکیده

With the continued growth of wind power penetration into conventional power grid systems, wind power forecasting plays an increasingly competitive role in organizing and deploying electrical and energy systems. The wind power time series, though, often present non-linear and non-stationary characteristics, allowing them quite challenging to estimate precisely. The aim of this paper is in proposing a novel hybrid model named Evol-CNN in order to predict the short-term wind power at 10-min interval up to 3-hr based on deep convolutional neural network (CNN) and evolutionary search optimizer. Specifically, we develop an improved version of Grey Wolf Optimization (GWO) algorithm by incorporating two effective modifications in its original structure. The proposed GWO algorithm is more effective than the original version due to performing in a faster way and the ability to escape from local optima. The proposed GWO algorithm is utilized to find the optimal values of hyperparameters for deep CNN model. Moreover, the optimal CNN model is employed to predict wind power time series. The main advantage of the proposed Evol-CNN model is to enhance the capability of time series forecasting models in obtaining more accurate predictions. Several forecasting benchmarks are compared with the Evol-CNN model to address its effectiveness. The simulation results indicate that the Evol-CNN has a significant advantage over the competitive benchmarks and also, has the minimum error regarding of 10-min, 1-hr and 3-hr ahead forecasting.