The country’s power market is still in the process of reform, and the agency purchase of electricity by power companies plays a decisive role in further exerting the market’s role in resource allocation. To ensure the stable development of the agency power purchase mechanism, it is necessary to clarify the market-based power purchase scale of power companies. At present, the amount of electricity purchased by agents is mainly predicted based on the electricity consumption of industrial and commercial users and typical load curves. There is a lack of a complete system, and it is difficult to predict accurately, resulting in a lack of reasonable planning. Therefore, a set of intelligent algorithms combining the similar daily and monthly prediction algorithms with hybrid time-series monthly prediction algorithms is proposed for power prediction and evaluation of agency power purchase business. The power consumption situation in Jiangsu Province in 2022 is forecasted and analyzed to help power companies accurately predict the overall electricity sales scale and plan power purchase plans reasonably.