A reinforced learning-based urban rail transit ground type super-capacitor energy storage system energy management method

Abstract

The invention relates to a reinforced learning-based urban rail transit ground type super-capacitor energy storage system energy management method. The method comprises two parts which are policy network initialization and online learning. The policy network initialization is to build a multi-vehicle operation scene model by using known line and vehicle information, a pre-compiled train working diagram and practically collected historical vehicle data in urban rail transit, and combine the multi-vehicle operation scene model, a no-load voltage prediction model, a direct current supply load flow calculation algorithm and an approximate dynamic programming algorithm to obtain a policy network which serves as an initial value of an online learning module; the online learning module employs amodel-free reinforced learning algorithm and performs charging and discharging threshold value online adjustment through a super-capacitor intelligent agent trial and error method. The method can perform online learning on super-capacitor energy storage system control policies in urban rail traction power networks, thereby optimizing an energy-saving effect and a voltage stabilizing effect.

Description

technical field [0001] The invention relates to rail transit control and energy-saving technology, in particular to an energy management method for an urban rail transit ground-type supercapacitor energy storage system based on reinforcement learning. Background technique [0002] In the traction power supply system of urban rail transit, the traction substation usually adopts 24-pulse diode rectification to convert 10kV / 35kVAC AC power into 750V / 1500V DC power to provide traction energy for line trains. Due to the unidirectionality of diode rectification, when the train brakes, the braking energy is transmitted to the traction network. If there is no traction train nearby to absorb it, the voltage of the traction network will rise rapidly, causing the start-up of the braking resistor and the occurrence of regenerative failure. In order to fully recover the regenerative energy of trains, reduce regenerative failures and voltage fluctuations in the traction network, supercapa...

AI Technical Summary

Problems solved by technology

[0006] The technical problem to be solved by the present invention is to provide an energy management method for ground-based supercapacitor energy storage systems of urban rail transit based on reinforcement learning, which can be used for complex and time-varying urban rail traction power supply networks and difficult to model accurately. The control strategy of the supercapacitor energy storage system is learned online to realize the optimization of energy saving effect and voltage regulation effect; an energy management method of the energy storage system based on reinforcement learning is proposed as an innovative method for optimizing the effect of energy saving and voltage regulation

Images