Prediction using long short‐term memory networks in the service of designing a novel pricing policy for smart grid

Abstract

Abstract Dynamic pricing is one of the most effective solutions for controlling and managing power consumption in the electricity markets. Two challenging issues to achieve this goal are to design a comprehensive policy, which can determine optimal prices for every party, and define an accurate simulator of the real environment, which can express the complexity of the satisfaction. To overcome these problems, we innovatively design an applicable and reliable policy that can determine the optimal price from every electricity market parties' point of view. Moreover, in this policy, satisfaction and profit coefficients are defined for the retailer and the customer, and then a step‐by‐step simulation is presented based on them. The proposed algorithm simulates interactions between the customer and retailer to reach an optimal point with improved performance. Furthermore, to increase flexibility and accuracy of the results, which make the system commercially operational, two long short‐term memory networks predict wholesale price and power demand will be utilised by the pricing section. For efficiency evaluation, the proposed method is compared with similar work to prove better precision and performance.

Keywords

• Artificial intelligence and data analytics
• optimal control
• pricing
• predictive control
• power markets
• power distribution reliability