(d) The MAPEs of LSTM with the MOP algorithm
The number of multi-steps	MAPEs of MOP of three-layer LSTM (%)	MAPEs of MOP of four-layer LSTM (%)	MAPEs of MOP of five-layer LSTM (%)	MAPEs of MOP of six-layer LSTM (%)	MAPEs of MOP of 10-layer of LSTM (%)	Average MAPE of MOP of three-, four-, five-, six-, 10-layer LSTM (%)
2	11.766	10.163	9.879	10.345	11.581	10.747
3	13.043	13.609	11.728	12.451	12.871	12.740
4	18.158	16.142	13.528	15.252	16.663	15.949
5	19.728	17.442	14.985	19.175	15.292	17.324
6	19.253	18.237	14.882	18.512	24.393	19.055
7	21.221	20.625	16.826	18.690	20.154	19.503
8	20.350	20.560	17.933	20.814	20.864	20.104
9	24.043	24.871	20.444	18.006	23.792	22.231
10	22.487	24.197	21.250	21.708	20.632	22.055
11	22.679	19.278	21.047	22.339	27.194	22.507
12	23.095	24.249	22.869	23.707	26.795	24.143
13	24.532	24.571	29.554	25.307	27.979	26.389
Average MAPE of MOP of all multi-steps (%)	20.030	19.495	17.910	18.859	20.684

MAPEs, mean absolute percentage errors; MOP, multiple-output prediction; LSTM, long short-term memory.