The invention discloses a traffic signal lamp control method and system based on reinforcement learning, and the method comprises the steps: firstly building a signal lamp control Agent model, then building a road and road intersection model, enabling the road traffic state information to be represented as a speed, position and current signal lamp state matrix, then, on the basis of traditional Q-Learning, establishing a traffic signal lamp control algorithm based on Deep Q Network (DQN) according to the road environment information; and finally, obtaining real-time road condition information through interaction between the Agent and the environment, performing search self-learning in a behavior space, estimating Q values for executing all possible actions in the current state, and selecting the action with the larger Q to be executed by utilizing an epsilon-greedy strategy. An existing traffic signal lamp control method is improved, the waiting time of vehicles at the intersection is minimized, the effective green light time of a signal control period is maximized, the vehicles are assisted to rapidly pass through the intersection, the passing time is shortest, meanwhile, the maximum traffic flow exists at the intersection, and thus the purpose of relieving traffic jam is achieved, and adaptive control of the traffic signal lamp is realized.