Indoor environment is under the influence of building operating circumstances and outdoor environment which are continuously changing in operational phase and difficult to predict. It is therefore hard to achieve optimal control of indoor environment. To cope with this problem, we developed optimal control method based on learning ability. Learning ability is a notable feature of life and it may achieve optimal determination autonomously. This paper reports the concept of the optimal control method and a feasibility study on HVAC system in intermediate season. Results of experiment indicate that developed control system reduces considerable amount of air conditioning energy consumption without worsening indoor thermal environment.