In response to the low efficiency and high cost of smart contract execution in the current digital economy environment, this article conducts algorithm optimization research based on blockchain smart contracts. Using Convolutional Neural Network (CNN) method, key features are extracted from massive data, and their historical running records are sequentially modeled. The behavior rules of contracts are analyzed to identify potential vulnerabilities and abnormal behaviors. Adopting the reinforcement learning algorithm of Deep Q-Network (DQN), the optimal decision is made based on the characteristics and behavior of smart contracts. This article compares the operating costs of two algorithms under the same contract, compares the amount of gas consumed by the two methods under the same contract, and their impact on system operating costs. It also compares them with existing dynamic scheduling methods to achieve the goal of reducing operating costs. The research results indicate that the optimization method proposed in this article can effectively reduce network operating costs and provide support for the construction of the national digital economy.