Multi-level converter control method based on reinforcement learning

Abstract

The invention discloses a multi-level converter control method based on reinforcement learning, and belongs to the technical field of direct current transmission. The control method comprises the following steps: constructing a multi-input multi-output neural network by taking processed key electrical quantity as input and taking the probability of occurrence of actions adopted by each effective switch of a multi-level converter as output; using a neural network to control the multi-level converter; and meanwhile, introducing a reward value, and updating internal parameters of the neural network through a reinforcement learning algorithm. Therefore, the reinforcement learning algorithm is applied to the control of the multi-level converter with a relatively complex circuit, a complex circuit model does not need to be known, the parameters of the neural network can be directly updated according to the measured input and output data to realize control, and a plurality of links such as PI control, modulation and voltage-sharing control in the traditional control of the multi-level converter are omitted.

Description

technical field [0001] The invention belongs to the technical field of direct current transmission, and more specifically relates to a control method of a multilevel converter based on reinforcement learning. Background technique [0002] In recent years, with the successful application of power electronic technology in motor drive, reactive power compensation, active power filter and distributed power generation, the research and development of high-voltage and large-capacity power electronic equipment has attracted increasing attention. Restricted by the development level of power electronic devices, the traditional two-level converter topology is difficult to meet the requirements of high-voltage and large-capacity power electronic conversion. The emergence of multi-level converters has changed this situation. Considering the limited withstand voltage of a single device, multi-level converters use multiple devices to share high voltage, and clamp the terminal voltage of p...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the defects and improvement needs of the prior art, the present invention provides a control method of a multilevel converter based on reinforcement learning, the purpose of which is to use the reinforcement learning algorithm directly according to the measured Input and output data are used for parameter training to achieve control, omitting many links in the traditional control of multi-level converters, such as PI control, modulation, voltage equalization control, etc.

Images