A novel hybrid approach to improve the accuracy of load forecasting by IOWA

Load balancing is one of the most challenging goals in smart grid systems. Obviously, a selfish user's behavior necessitates the use of incentive compatible mechanisms in order to regulate supply and demand. Dynamic pricing is one of the best mechanisms in which the price is being adjusted dynamically such that to make a balance between supply and demand. In the balance, the consumer's demand for energy through financial incentives is adjusted. To determine and announce the appropriate electricity price, there should be a precise forecast for energy usage. In this paper, in order to forecast energy usage two neural networks for each influential factors based on the situation such as weather related or historical loads criteria are developed. Afterwards, the outputs of neural networks are aggregated with the use of Induced Ordered Weighted Operator (IOWA). The argument ordering process is guided by mean absolute percentage error. Also the cuckoo optimization algorithm is applied on artificial neural networks to improve the accuracy of them. The experimental result show that the precision of aggregated load forecasting based upon IOWA operator is improved significantly.