Deep learning-based scheduling of virtual energy hubs with plug-in hybrid compressed natural gas-electric vehicles

The virtual energy hub (VEH), a combination of virtual power plant and energy hub concepts, faces many uncertainties due to its constituent distributed energy resources. This paper presents the deep learning-based scheduling of VEH for participation in electrical and thermal markets using bidirectional long short-term memory (BLSTM) network, which offers excellent accuracy in forecasting uncertain parameters by concurrent using past and future dependencies. In addition to applying learning methods, energy storage systems can also influence the optimal management of uncertainties. To provide the required electrical storage equipment, the VEH employs plug-in hybrid CNG-electric vehicles (PHGEVs) that can use both electrical energy and compressed natural gas (CNG) to fulfill their energy needs. The alternative fuel can tackle the limitations of prolonged charging of electric vehicles and excess load caused by these vehicles at peak hours. To supply the secondary fuel of PHGEVs, the modeled VEH includes a CNG station, which compresses the natural gas imported from the natural gas grid before delivering it to the vehicles. Furthermore, phase change material-based thermal energy storage (PCMTES) is considered in the VEH configuration, which unlike other common thermal energy storage systems, operates at a constant temperature during the charging and discharging period. Lastly, the simulation of the developed system illustrates that PHGEVs can reduce the imposed cost in unforeseen situations by up to 26 percent and increase the system’s flexibility. •Self-scheduling of virtual energy hub for participating in energy markets.•Deep learning-based uncertainty forecasting.•Bidirectional long short-term memory network.•Energy management of Plug-in hybrid CNG-electric vehicles.•CNG station for supplying required CNG for vehicles.•Energy management of phase-change material based thermal energy storage.