鱼群算法与神经网络结合的节能减排效果评价

从污染物减排率、单位工业增加值减排量、治理工业污染投资总额、GDP相关指标、能耗下降率5个方面建立节能减排效果评价指标体系,分析BP神经网络与鱼群算法结合的可行性,探讨鱼群算法优化神经网络的步骤。最后对7个地区2006~2009年节能减排效果评价指标,在专家打分测评的基础上,运用神经网络及鱼群算法优化神经网络方法进行节能减排效果评价。研究结果表明:在收敛过程中,运用神经网络所得实际输出值与专家评分的误差长时间停留在0.7左右,而运用鱼群算法优化神经网络方法能够以较大的斜率快速收敛到期望误差;在误差为0.001时,前者经过202次训练后能够达到目标,而后者只需要75次训练就能达到目标,这表明鱼群算法优化神经网络具有准确、快捷、简易等优点,此方法用于节能减排效果评价行之有效。

Comprehensive effect evaluation of energy saving and emission reduction based on fish-swarm algorithm optimizing neural network

Based on five factors,i.e.pollutant emission reduction rate,emission reduction of unit industrial added value,total investment in industrial pollution control,related indicators of GDP and energy consumption reduction rate,an index system to comprehensively evaluate energy saving and emission reduction effect was established.Then the feasibility of optimizing BP neural network by fish-swarm algorithm was analyzed,and the procedures for the optimization of BP neural network by fish swarm algorithm were researched.According to the energy-saving data in seven regions from 2006 to 2009 and based on the assessment of expert scoring,energy-saving effects were evaluated by neural network and fish swarm algorithm-optimized neural network respectively.The results show that during the convergence,the error local optima is about 0.7 for a long time using neural network algorithm-optimized neural network while it reaches the target error figure quickly using fish swarm algorithm-optimized neural network.When error is 0.001,the former reaches the target after 202 times of training,and the latter only 75 times.The results indicate that the method of fish swarm algorithm-optimized neural network is accurate,fast,simple and easy,and this method is effective for the evaluation on effect of energy saving and emission reduction.