CAJ | 학술논문

以光伏阵列为研究对象，分析了辐照强度、温度以及日类型对光伏阵列出力的影响。建立了光伏短期功率预测最小二乘支持向量机LS-SVM模型，依据实验数据对模型进行了验证计算，并与BP神经网络模型做了比较，其中LS-SVM模型最大相对误差值为10.54%，平均绝对百分比误差（MAPE）为8.18%，绝对误差平方和平均值的均方根（RMSE）为0.4884，表明模型预测值离散化程度较小，所有预测点均与实际值非常接近，模型具有较好的拟合效果和泛化能力，可以有效地预测短期光伏发电功率。
이광복진렬위연구대상，분석료복조강도、온도이급일류형대광복진렬출력적영향。건립료광복단기공솔예측최소이승지지향량궤LS-SVM모형，의거실험수거대모형진행료험증계산，병여BP신경망락모형주료비교，기중LS-SVM모형최대상대오차치위10.54%，평균절대백분비오차（MAPE）위8.18%，절대오차평방화평균치적균방근（RMSE）위0.4884，표명모형예측치리산화정도교소，소유예측점균여실제치비상접근，모형구유교호적의합효과화범화능력，가이유효지예측단기광복발전공솔。
The photovoltaic array is used as the object of study to analyze the effect of irradiation intensity and temperature on the output of photovoltaic array. The least squares support vector machine (LS-SVM) model for short-term photovoltaic power prediction is established ,verified and calculated on the basis of experimental data,and compared with BP neural network model. The maximum relative error of LS-SVM model is 10.54%, the mean absolute percentage error (MAPE) is 8.18%and the root-mean-square error (RMSE) of mean value of absolute error sum of squares is 0.4884,which means that the discretization level of predicted value of model is relatively low,all predicted points are close to their actual value and the model has a good imitative effect and generalization ability and can predict the short-term photovoltaic power effectively.