【摘要】 针对牧草产量预测存在的多因素耦合难题，以朔州市辖区为研究对象，整合2005—2022年国家基本气象观测站数据(包括气温、降水、日照及10 cm地温数据),耦合该地区近10年牧草年产量数据及其生长期气候条件需求特征，建立了基于BP神经网络的基础预测模型，并引入标准粒子群算法(PSO)及改进微粒群算法(IPSO)进行模型参数优化。仿真对比研究表明，经IPSO优化的BP神经网络预测性能显著提升，其平均绝对误差较BP模型与PSO-BP模型更低。算法改进后的模型也展现出更强的泛化性能，IPSO算法较基础PSO的全局寻优效率提高，从而验证了改进算法在复杂气象条件下牧草产量预测中的应用价值。更多还原

【Abstract】 To address the multi-factor coupling challenges in forage yield prediction, this study focuses on the Shuozhou municipal area as the research subject. By integrating data(including temperature, precipitation, sunshine duration, and 10 cm ground temperature data) from six national basic meteorological stations from 2005 to 2022, and coupling with the region’s annual forage yield data and growth-stage climate requirement characteristics over the past decade, established a BP neural network-based foundational prediction model. The model parameters were optimized through the introduction of standard particle swarm optimization(PSO) and improved particle swarm optimization(IPSO) algorithms. Simulation comparative studies demonstrate that the IPSO-optimized BP neural network achieves significant performance enhancement in prediction accuracy, exhibiting lower Mean Absolute Error compared with both the basic BP model and PSO-BP model. The algorithm-improved model also demonstrates stronger generalization capabilities, with IPSO showing higher global optimization efficiency than basic PSO. These findings validate the application value of the improved algorithm for forage yield prediction under complex meteorological conditions.更多还原