TY - JOUR
T1 - A multi-timescale smart grid energy management system based on adaptive dynamic programming and Multi-NN Fusion prediction method
AU - Yuan, Jun
AU - Zhang, Guidong
AU - Yu, Samson S.
AU - Chen, Zhe
AU - Li, Zhong
AU - Zhang, Yun
N1 - Publisher Copyright:
© 2022 Elsevier B.V.
PY - 2022/4/6
Y1 - 2022/4/6
N2 - The complexity of power grids, the intermittent renewable energy generation and the uncertainty of load consumption bring great challenges to modern energy management systems (EMSs). To solve the energy optimization problem in the time-varying smart grid, this paper proposes a multi-timescale EMS based on the adaptive dynamic programming (ADP) algorithm and multi-neural-network fusion (MNNF) prediction technology. In detail, according to different power consumption characteristics, this paper uses fuzzy C-means (FCM) clustering algorithm to classify power users into industrial users, commercial users and residential users. Based on the classification results, an MNNF prediction method is proposed that can integrate different influencing factors to predict load consumption and renewable energy generation. Then a multi-timescale ADP optimization algorithm is proposed to maximize the utilization of renewable energy on daily, intra-day and real-time (i.e., three timescales) of energy behavior. The convergence of the multi-timescale ADP algorithm is proved mathematically when the initial value is a random semi-positive definite function. Finally, the proposed ADP with MNNF energy management system is verified on a hardware-in-the-loop (HIL) platform.
AB - The complexity of power grids, the intermittent renewable energy generation and the uncertainty of load consumption bring great challenges to modern energy management systems (EMSs). To solve the energy optimization problem in the time-varying smart grid, this paper proposes a multi-timescale EMS based on the adaptive dynamic programming (ADP) algorithm and multi-neural-network fusion (MNNF) prediction technology. In detail, according to different power consumption characteristics, this paper uses fuzzy C-means (FCM) clustering algorithm to classify power users into industrial users, commercial users and residential users. Based on the classification results, an MNNF prediction method is proposed that can integrate different influencing factors to predict load consumption and renewable energy generation. Then a multi-timescale ADP optimization algorithm is proposed to maximize the utilization of renewable energy on daily, intra-day and real-time (i.e., three timescales) of energy behavior. The convergence of the multi-timescale ADP algorithm is proved mathematically when the initial value is a random semi-positive definite function. Finally, the proposed ADP with MNNF energy management system is verified on a hardware-in-the-loop (HIL) platform.
KW - Adaptive dynamic programming
KW - Energy management system
KW - Hardware-in-loop
KW - Multi-neural network fusion prediction algorithm
UR - http://www.scopus.com/inward/record.url?scp=85124076691&partnerID=8YFLogxK
U2 - 10.1016/j.knosys.2022.108284
DO - 10.1016/j.knosys.2022.108284
M3 - Journal article
AN - SCOPUS:85124076691
SN - 0950-7051
VL - 241
JO - Knowledge-Based Systems
JF - Knowledge-Based Systems
M1 - 108284
ER -