为了提高光伏发电系统短期输出功率的预测精度,建立了基于改进鸡群算法优化支持向量机(ICSO-SVM)的预测模型,在鸡群算法中引入动态惯性权重和自适应因子加强算法的寻优能力。通过计算得到对光伏发电影响较大的因素为太阳辐射强度、大气温度和相对湿度;计算出待预测日期和历史日期之间的关联度,确定预测所需要的训练样本并对模型进行训练;利用训练好的预测模型对预测地区秋季平稳天气和突变天气的光伏阵列输出功率分别进行预测。仿真实验表明：该模型的平均绝对百分比误差和均方误差与改进前相比分别降低5.547%和0.080,与基于粒子群优化算法模型相比分别降低8.255%和0.202,该模型使预测精度得到有效提高。

To improve the short-term output power prediction accuracy of photovoltaic power generation, a prediction model based on improved chicken swarm optimization-support vector machine (ICSO-SVM) was established. The dynamic inertia weights and adaptive factors were introduced into the chicken swarm algorithm to strengthen the algorithm’s optimization ability. The factors greatly impacting the photovoltaic power generation were obtained through calculations, including solar radiation intensity, atmospheric temperature and relative humidity. The correlation degree between the predicted date and the historical date was obtained, the training samples needed for the prediction were determined, and the prediction model was trained with the training samples. The output powers of PV array in the stable and abrupt weather in autumn were predicted by using the trained prediction model. The simulation experiment results show that the mean absolute percentage error and mean square error of the model reduce by 5.547% and 0.080 compared with those before the improvement, and by 8.255% and 0.202 respectively compared with those of the model based on particle swarm optimization algorithm. The method can effectively improve the prediction accuracy.