Bacteria Foraging Reinforcement Learning for Risk-Based Economic Dispatch via Knowledge Transfer

Abstract

This paper proposes a novel bacteria foraging reinforcement learning with knowledge transfer method for risk-based economic dispatch, in which the economic dispatch is integrated with risk assessment theory to represent the uncertainties of active power demand and contingencies during power system operations. Moreover, a multi-agent collaboration is employed to accelerate the convergence of knowledge matrix, which is decomposed into several lower dimension sub-matrices via a knowledge extension, thus the curse of dimension can be effectively avoided. Besides, the convergence rate of bacteria foraging reinforcement learning is increased dramatically through a knowledge transfer after obtaining the optimal knowledge matrices of source tasks in pre-learning. The performance of bacteria foraging reinforcement learning has been thoroughly evaluated on IEEE RTS-79 system. Simulation results demonstrate that it can outperform conventional artificial intelligence algorithms in terms of global convergence and convergence rate.

A new generation of AI: A review and perspective on machine learning technologies applied to smart energy and electric power systems

“…In terms of risk dispatching, the RL algorithm can be used to solve risk assessment issues for economic dispatch (ED). For example, Han et al demonstrated an RL‐based transfer BFO (TBFO) knowledge model, which can address the fast risk‐based ED optimization issues in large‐scale complex power grids. Concretely speaking, in this knowledge model, a kind of heuristic BFO algorithm was combined with the try‐error mechanism of Q‐learning.…”

“…The methods to solve such issues not only include conventional optimization methods such as the interior‐point algorithm but also include the probabilities achieved search‐based heuristic SI algorithms such as GA, PSO, ABC, and QGA . In the solution process, TL, a learning framework that has been a growing concern and has been widely investigated in recent years, aims to apply experience (ie, knowledge or strategy) and the results learned previously in the auxiliary field to a similar but different target domain based on task similarity and to reuse the existing experience to accelerate the learning speed of the new tasks and thus improve the efficiency with which the new tasks can be learned …”

“…In the field of SG, TL has been preliminarily used in scenarios such as reactive power optimization (ie, AVC), short‐term power load forecasting, short‐term wind speed prediction, decentralized carbon‐energy combined‐flow optimization of large‐scale power systems, risk‐based ED, and generation‐consumption coordinated frequency control of an islanded microgrid . In addition, TL has been integrated with PSO for regional VQC of cluster wind power farms grid .…”

“…TL has also achieved some good results in the field of EI. For instance, scholars have combined TL with various heuristic SI algorithms and RL algorithms to form more advanced TL algorithms and applied them to multiple EI scenarios such as optimal joint dispatch of multienergy systems, real‐time supply‐demand interactive dispatching of power systems, decentralized optimal carbon‐energy combined‐flow power system optimization, dynamic dispatching of AGC power commands involving wind, photovoltaic, and EVs, and risk‐based ED . In these scenarios, TL has achieved good effects.…”

Online Area Load Modeling in Power Systems Using Enhanced Reinforcement Learning

“…In applications involving the online control or dispatch of power systems, reinforcement learning (RL) methods have exhibited remarkable capabilities for performing online calculations [21][22][23]. RL is based on the concept of trial and error and explicitly considers the problem of a goal-directed agent interacting with an uncertain environment [24].…”