基于SSA-FIG-LSTM模型的天然气负荷区间预测OA

Natural gas load data is characterized by significant fluctuation,high noise level,and redundancy.The prediction effect is poor using single model prediction method,and point forecast results fail to reflect the randomness of load fluctuation.Therefore,a natu-ral gas load interval forecasting method based on singular spectrum analysis(SSA),fuzzy information granulation(FIG),and long short-term memory(LSTM)is proposed.Using SSA,the raw data is decomposed and reconstructed into trend component,periodic com-ponent,and noise component;The noise component is processed by FIG,effective information is extracted from high-frequency data,and the maximum,minimum,and average values of window fluctuations are extracted to replace the original noise component;The feature se-lection is performed according to the characteristics of each component,and the prediction is performed by combining feature sequence with component sequence and inputting into the LSTM model;The predicted results of various components are effectively integrated u-sing the idea of"point+interval"to form final prediction interval.Compared with other traditional models,the proposed method has higher prediction accuracy and narrower prediction interval,and the prediction results can objectively reflect the uncertainty of load fluc-tuations.