【作者】 徐梅；

【导师】 刘宝；

【作者基本信息】 中国石油大学(华东) ， 控制科学与工程， 2023， 硕士

【摘要】 水资源是人类不可缺少的自然资源之一,对人类基本的生产生活和社会经济发展都有很大影响。随着社会经济的不断发展,人们对水资源的需求急剧增加。而水资源的时空分布极不均匀,在季风气候地区,夏季和冬季降雨量差距很大,当降雨量大时,河流径流量骤增,易引起洪涝灾害;当降雨量小时,径流量很小甚至断流,既满足不了人们基本的生活需求也满足不了航运要求。对流域流量进行合理调控,既能够保障人们基本的生产生活也能降低洪涝灾害的发生,对保障经济和民生具有重要意义。因此,建立精确的流量预测模型并且对流量进行合理分配成为当前急需解决的问题。为提高澳大利亚墨累河多支流流域流量预测方法的精确性和流量规划决策方案的有效性,本文基于生物协同机制对墨累河系统建模与规划算法进行研究。(1)针对多支流流域流量预测模型预测精度较低的问题,建立了基于细胞协同机制的优化算法优化Elman神经网络的预测模型(CCOA-Elman)。根据人体免疫系统中免疫细胞协同作用来消灭入侵病毒的过程,提出了一种基于细胞协同机制的优化算法(CCOA),并用其对Elman网络参数的训练过程进行优化,从而改善了梯度下降方法的局限性。本文利用CCOA-Elman神经网络建立多支流流域系统模型,该模型预测效果优于Elman网络模型。(2)为进一步提高CCOA-Elman预测模型的预测精度,基于门控循环单元(GRU)提出了CCOA-Elman-GRU融合预测模型(以下简称CEG)。该模型由CCOA-Elman预测模型和GRU偏差修正模型两部分构成。偏差修正模型的加入,使得融合模型的拟合精度有所提升。CEG融合预测模型在CCOA-Elman预测模型的基础上提升了流量预测效果。(3)为解决复杂多支流流量规划决策算法存在寻优精度较低和效率低下等问题,提出了一种将神经内分泌免疫系统机制(NEI)与粒子群算法相结合的混合优化算法(NEI-PSO),并将该算法对多支流流域流量进行规划决策。根据人体内NEI机制,设计了神经指引单元、免疫定向单元和内分泌精调单元。在三种单元的协同作用下对粒子速度的更新方式进行改进,使粒子群算法的优化效果有所提升,从而减少算法收敛时间,提高其全局寻优能力。将NEI-PSO算法应用于河流规划决策中,获得了较好的规划决策效果。最后在MATLAB平台中对本文提出的CCOA-Elman预测模型、CEG融合预测模型以及NEI-PSO优化算法进行仿真验证。实验结果表明:CCOA-Elman预测模型的预测精度可保持在95.2%以上,模型性能优于传统预测方法;CEG融合预测模型的预测精度可保持在98%以上;NEI-PSO规划算法在模型建立的基础上对阀门开度和蓄水池个数等决策变量进行规划,找到决策变量的最优配置,NEI-PSO算法寻优能力明显优于PSO算法、LDWPSO算法和PSO-CF算法。上述三种算法在流量预测及规划决策方面具有一定的优越性,为未来多支流流域系统的优化调度提供了新的研究思路。更多还原

【Abstract】 Water resource is one of the indispensable natural resources,which has a profound effect on human basic production life and socio-economic development.As the economy continues to grow,the demand of people for water resources has increased sharply.However,the spatial and temporal distribution of water resources is extremely uneven.In most areas,when the rainfall is high,the river runoff increases sharply,which is easy to cause floods.When the rainfall is small,the runoff is small or even interrupted,which can neither meet basic of people living needs nor meet the requirements of shipping.Reasonable regulation of the flow of the basin can not only protect people’s basic production and livelihood but also reduce the occurrence of flooding,which is of great significance to protect the economy and people’s livelihood.Therefore,it is urgent to establish an accurate traffic prediction model and allocate traffic reasonably.To improve the accuracy of current flow prediction methods and the rationality of flow planning decision scheme in multi-tributary Murray River basin in Australia,this thesis investigates the modeling and planning algorithm of water conservancy system based on biological cooperation mechanism.(1)To improve the effectiveness and precision of the runoff prediction model for multi-tributary basins,an intelligent optimization algorithm based on the cell coordination mechanism to optimize the prediction model of Elman neural network(CCOA-Elman)is proposed.An optimization algorithm based on cell communication mechanism(CCOA)is proposed based on the coordination action of immune cells within the human immune system to destroy invading viruses.This algorithm is also used to improve the training process of Elman neural network model parameters,which improves the limitations of the gradient descent method and increases the fitting precision of model.The CCOA-Elman neural network is used to model the multi-tributary watershed system and predict river discharge.(2)To further improve the prediction precision of CCOA-Elman prediction model,the gated recurrent unit(GRU)is used to improve it,and a CCOA-Elman-GRU fusion prediction model(CEG)is proposed.The model consists of two parts,the CCOA-Elman prediction model and the GRU deviation correction model.The addition of the bias correction model improves the fitting precision of the fusion model.The CEG fusion model effectively improves the traffic prediction based on the original improved model.(3)A hybrid optimization approach by combining PSO with mechanisms in neuro-endocrine-immune systems(NEI-PSO)is proposed to solve the problem of low accuracy and low efficiency in the traditional flow planning decision planning method.It is also used to make planning decisions for the flow of multi-tributary basins.Inspired by the NEI mechanism in humans,a neural guidance unit,an immune targeting unit and an endocrine regulation unit are designed in NEI-PSO.The particle velocity update method is improved by the coordination effect of the three units to enhance the optimization effect of the particle swarm algorithm,thus the convergence time of NEI-PSO is reduced and the global optimization ability is improved.The NEI-PSO is applied to a river planning decision model and achieves better planning decision results.Finally,the CCOA-Elman prediction model,CEG fusion prediction model and NEI-PSO algorithm proposed in this thesis are simulated and validated in MATLAB.The experimental findings demonstrate that the prediction accuracy of the CCOA-Elman prediction model can be maintained above 95.2%,which is a substantial enhancement in performance compared with the traditional prediction models.The prediction precision of the CEG can be maintained above 99%.The NEI-PSO algorithm plans for decision variables such as the valve opening degree and the number of reservoirs,and finds the optimal configuration of the decision variables.The optimization ability of NEI-PSO algorithm is obviously better than that of PSO algorithm,LDWPSO algorithm and PSO-CF algorithm.The above three algorithms have certain advantages in flow prediction and planning decision,and provide new research ideas for the optimal scheduling of multi-tributary watershed systems in the future.更多还原