Research On Human Learning Optimization Algorithm With Reasoning Mechanism And Its Application

Human Learning Optimization(HLO)is a novel meta-heuristic algorithm inspired by the learning mechanisms of humans.Many cognitive activities of humans contain an element of reasoning,and with reasoning,humans can gain deeper information on problems to boost learning performance.However,the standard HLO does not include the learning strategy of reasoning,which is invented based on a simplified human learning model.To improve the optimization performance of the standard HLO and continuous HLO,this thesis introduces the reasoning mechanism into the standard HLO and continuous HLO,and proposes binary-coding HLO with reasoning,real-coding HLO with reasoning and hybrid-coding HLO with reasoning.Then the improved algorithms are applied to optimize the recognition model of furnace flame and the prediction model of furnace temperature.The main research work of this thesis is as follows:(1)Inspired by intuitive reasoning and logical reasoning of humans,the intuitive reasoning learning operator(IRLO)and the logical reasoning learning operator(LRLO)are designed,then the HLO with intuitive reasoning learning(HLOIRL)and the HLO with logical reasoning learning(HLOLRL)are proposed.The numerical experiments analyze the role and function of the designed IRLO/LRLO,and the reasons why the designed IRLO/LRLO can effectively enhance the exploration and exploitation of algorithms are provided.Besides,the convergence of HLOIRL and HLOLRL is analyzed to prove that they both converge to the global optimal with probability 1.(2)The synergy reasoning learning model is designed based on the in-depth analysis of the interaction mechanism between intuition reasoning and logical reasoning,and the HLO with synergy reasoning learning(HLOSRL)is proposed.The numerical experiments analyze that the designed synergy reasoning learning model can help HLOSRL utilize the learning ability of intuitive reasoning and logical reasoning more efficiently.Based on this,the synergy reasoning learning model with enhanced exploration-exploitation is further designed based on the difference of utilization efficiency between reasoning learning operator and social learning operator,and the enhanced HLO with synergy reasoning learning(EHLOSRL)is proposed.The numerical experiments analyze that the designed learning model can help EHLOSRL to achieve a practically ideal trade-off between exploration and exploitation,and therefore the optimization ability of the algorithm is significantly enhanced.Besides,the convergence of HLOSRL and EHLOSRL is analyzed to prove that they both converge to the global optimal with probability 1.(3)For high-dimensional continuous problems,the performance of the binary HLO will significantly decrease due to the "dimensionality disaster".Inspired by the metaheuristic-based intuitive reasoning,the linear-based reasoning learning operator(RLO-L)and the levy-flights-based reasoning learning operator(RLO-F)are designed,then the linear-based continuous HLO with reasoning learning(CHLORL-L)and the levy-flights-based continuous HLO with reasoning learning(CHLORL-F)are proposed.The numerical experiments analyze that the designed RLO-L/RLO-F can help the algorithm to perform the exploration ability more effectively,and work with other learning operators to better search for optimal information in the solution space.At the same time,the convergence of CHLORL-L and CHLORL-F is analyzed to prove that they converge to the global optimal with probability 1.Besides,a hybrid-coded human learning optimization with reasoning learning(Hc HLORL)is proposed.(4)The flame distribution and temperature changes are the most direct and effective features to reflect the combustion status of coal-fired boilers.The double threshold color recognition model(DTCRM)is designed to improve the recognition accuracy of furnace flame under normal operating conditions,and the recognition method for furnace flame based on DTCRM with EHLOSRL is proposed.Then the adaptive color recognition model with mixed variable(ACRMM)is designed to improve the recognition accuracy of furnace flame under complex operating conditions,and the recognition method for furnace flame based on ACRMM with Hc HLORL is proposed.Besides,an optimized kernel extreme learning machine(OKELM)is used to model the furnace temperature,and the prediction method for furnace temperature based on OKELM with Hc HLORL is proposed to improve the prediction accuracy of furnace temperature.Therefore,the proposed methods are beneficial to monitor the combustion state of the furnace more accurately,which further improves the combustion efficiency of the furnace and lays the foundation for ensuring its safe operation.