| 摘要: |
| 【目的】采用优化的寻优算法进行高效PI控制参数寻优的尝试。【方法】首先采用复合梯度法和复合粒子群法进行单一渠池PI控制参数寻优，对寻优过程中存在的初值不稳定等问题进行了探讨和针对性优化，提出了一种适用于渠道控制的分步寻优算法；随后针对多渠池的多变量问题提出了随机多渠池多变量寻优算法与固定权重的多渠池多变量寻优算法。【结果】单渠池工况下：复合梯度法和复合粒子群法能较好地解决寻优初值不稳定及寻优效率低下的问题。在控制效果相当的情况下，2种算法较均匀网格法的计算效率均得到显著提升，最长寻优耗时仅为网格法的5%左右。复合梯度法所用平均寻优耗时仅仅为复合粒子群算法的35.76%，复合梯度法较复合粒子群算法更为高效。多渠池工况下，随机多渠池多变量寻优算法与固定权重的多渠池多变量寻优算法在显著提高寻优效率的同时保证了较好的控制性能。然而，随机多渠池多变量寻优的控制效果较固定权重的多渠池多变量寻优有较大的改善，随机法较固定权重法的流量偏差小90.48%、闸门开度偏差小77.57%，系统稳定时间小17.05%，水位偏差小50.42%。但随机法所付出的时间成本也较高，寻优耗时较固定权重法多63.19%。【结论】随机多渠池多变量寻优算法的寻优耗时更长但控制性能更佳，使用者可根据其对于控制性能或耗时的要求有针对性地选择。 |
| 关键词: 渠道控制； PID算法； 参数寻优；网格法； 梯度算法； 粒子群算法； 多渠池参数寻优 |
| DOI： |
| 分类号:TV133 |
| 基金项目:国家自然科学基金(51979202,51439006,51009108);“十三五”国家重点研发项目(2016YFC0401810),国家自然科学基金项目（面上项目，重点项目，重大项目） |
|
| Research on single objective multivariable PI channel control parameter tuning algorithm based on simulation |
|
Zhang Yu-meng, Guan Guang-hua
|
|
State Key Laboratory Of Water Resources And Hydropower Engineering Science,Wuhan University
|
| Abstract: |
| 【Background】 In engineering practice, classical control method is widely used, especially the PID algorithm with simple principle and high robustness. When designing controller parameters, the algorithm has two main methods: the first is the traditional classical design method, mainly including the construction of control model, and the combination of the optimal control theory to optimize the PID parameters; the second is to carry out the computer simulation and do trial and error, using the digital simulation script combined with the optimization program to determine the optimal PID parameters. However, because the classical controller design theory does not take into account the limitations of the real channel, the nonlinearity and the communication delay of the system, the performance are often very poor, and even shoe unstable behavior. Currently for channel control system, PI control parameters optimization mostly adopts the Grid Search method as main way to tune the controller. 【Objective】But due to the characteristics of the canal system, such as nonlinear and coupling, the initial value of the trial calculation and the setting of the grid density are highly dependent on experience and the calculation time can be huge, which is more difficult to implement on the large canal system with complex multi-pools. 【Method】 Therefore, this paper intends to adopt the optimized optimization algorithm to carry out the efficient PI control parameter tuning attempt by simulation. In this paper, the compound gradient algorithm and the compound particle swarm optimization algorithm are adopted to tune the PI control parameters of channel with single reach, and the initial value instability in the optimization process is discussed and optimized. It puts forward a kind of applicable to channel control step by step optimization algorithm. 【Result】 Under the condition of single reach: the composite gradient algorithm and composite particle swarm optimization algorithm proposed in this paper can solve the problems of unstable initial value avoid low efficiency of optimization. When the control effect was similar, the efficiency of the two algorithms was greatly improved compared with that of the Grid Search method in terms of time, and the maximum time consumption was only about 5% of that of the Grid Search method. But in general, the average time used by the composite gradient algorithm was only 35.76% of that of the composite particle swarm optimization (PSO). When applied to multi-reach channel, the random multi-variable optimization algorithm and the fixed-weight multi-variable optimization algorithm was proposed. They both can significantly improve the optimization efficiency and guarantee the better control performance. However, the control performance of the random multi-reach multi-variable optimization algorithm was much better than that of the fixed-weight multi-variable optimization algorithm. The flow deviation of the random method was 90.48% less than that of fixed-weight method, the gate opening deviation was 77.57% less, the system stability time was 17.05% less, and the water level deviation was 50.42% less. However, the calculation time consumption of the random method was also larger, which was 63.19% more than the other. Therefore, the random multi-reach multi-variable optimization algorithm took longer time but had better control performance. 【Conclusion】The algorithm proposed in this paper can be used in the design of new water delivery control system and the upgrade of existing channel control system. It has essential value for the modernization transformation of water distribution system in a large number of irrigated areas in China. |
| Key words: channel control; PID algorithm parameter optimization; grid search method; gradient algorithm; particle swarm optimization; multi-channel pool parameter optimization |
