The scheduling of software-defined radio (SDR) sensors in wireless sensor networks is challenging due to their reconfigurable nature across time and frequency. Unlike traditional sensors, SDR sensors face complex constraints including time slots, frequency selection, and activation limits. These challenges are critical in smart city monitoring and spectrum surveillance. This paper models the problem as a large-scale multi-constraint combinatorial optimization task and proposes an improved evolutionary algorithm to address it efficiently. The proposed method integrates prior knowledge and information gathered during the evolution process using Bayesian inference, dynamically updating the posterior probabilities to effectively guide the search process of the genetic operators. Furthermore, a repair scheme based on the posterior probabilities is designed to enhance the likelihood of finding high-quality solutions in a complex solution space. To evaluate the effectiveness of the proposed algorithm, extensive experiments were conducted on six datasets. The experimental results show that, compared with existing evolutionary algorithms, the proposed method exhibits superior performance.