As is well known, there are two main sources of energy, which are renewable energy sources like fossil fuels and non-renewable energy sources like solar and wind energy. From available data renewable energy sources contribution to global energy needs is more than that from non-renewable energy sources. Of the renewable energy sources the most promising area in terms of energy revolution is solar energy. It is also believed that the global energy demand could be met many times over through direct utilization of the suns radiation energy. Furthermore solar energy is considered as a reliable alternative to fossil fuels because of its lower environmental pollution.
Over the past years, several studies have focused on designing and modeling of solar energy systems. Of the various methods the machine learning techniques such as artificial neural networks (ANNs), extreme learning machine (ELM) and support vector machines (SVM) have been extensively used to simulate the energy demand, solar radiation and many other real-world problems. The focus of majority of the exiting solar radiation research is on the prediction of monthly or annual solar radiation.
Training of artificial neural networks with optimization algorithms has proved to be an effective approach to find optimal solutions. Simulated annealing (SA) has been found to be an efficient tool in different scientific domains. In a combined method of artificial neural networks and simulated annealing, simulated annealing plays a key role in the training of the artificial neural networks algorithm. The hybrid artificial neural networks and simulated annealing method have been successfully applied in some area like improving response surface methodology, predicting the accuracy of components produced by wire electrical discharge machining etc. While this hybrid method has a remarkable prediction performance there are almost no studies on its application in the field of energy conversion and management up to now.
Dr. Pengcheng Jiao at the University of Pennsylvania in collaboration with Dr. Seyyed Mohammad Mousavi at Azad University and Dr. Elham Mostafavi at Isfahan University of Technology proposed a new hybrid approach to estimate the daily solar radiation on horizontal surface. The role of simulated annealing in this hybrid algorithm is to provide sufficient initial values for the artificial neural networks weights. The hybrid model was transformed into a functional representation this was done so as to facilitate the use of the optimal form. Mashhad city in Iran was chosen as test site and daily solar radiations measurements in this city which were stored in a database were used. The authors also went further to compare the results with those provided by the regular artificial neural networks and other existing models. Their findings were published in the peer reviewed journal Energy Conversion and Management.
The authors found that using simulated annealing to adjust the network’s weights has improved the performance of single artificial neural networks. Based on the conducted comparative study, the artificial neural networks/simulated annealing model outperforms the artificial neural networks and support vector machine models, and different solutions presented in the literature.
The research team was able to show that introducing the simulated annealing strategy into the artificial neural networks modeling process improves the prediction performance. They were also able to show that the hybrid model outperforms the artificial neural networks and support vector machine models. Their findings indicate that the artificial neural networks/simulated annealing method and the interpretation process is not applicable only to the current problem but can be extended to a variety of energy-related problems.
Seyyed Mohammad Mousavi, Elham S. Mostafavi, Pengcheng Jiao. Next generation prediction model for daily solar radiation on horizontal surface using a hybrid neural network and simulated annealing method. Energy conversion and management, Volume 153, Issue 1, December 2017, Pages 671-682
Go To Energy conversion and management