In order to improve the efficiency of parking resource allocation in urban core areas, this paper proposes a dynamic parking pricing model based on reinforcement learning algorithm. Firstly, based on revealed preference (RP) survey data, a multinomial logit model (MNL) is constructed to quantify the effects of various influencing factors on parking decisions, providing a basis for pricing response modeling of parking choice behavior. On this basis, with the optimization objective of balancing the parking lot utilizations, a discriminated dynamic pricing model is constructed, and solved by a Q-learning reinforcement learning algorithm. Then, to achieve continuous optimization and dynamic adjustment of pricing strategies, an adaptive reinforcement learning framework is established by defining the system state as a combination of utilizations and charge rates for various types of parking lots, taking price decisions as actions, and constructing a reward function based on the degree of occupancy dispersion. Finally, to verify the effectiveness of this method in optimizing parking resource allocation, a simulation experiment is conducted on the area of the Second Affiliated Hospital of Nanchang University in Donghu District, Nanchang City, China.