灌溉渠系输配水过程控制对水资源利用效率影响显著，手动控制模式下水量调控不精准、优化调度困难、运行管理成本高。因此，灌溉渠系的自动化控制，是提升农业灌溉效率的重要手段。水联网实时感知与过程跟踪的技术特征，与灌溉渠系自动化控制技术相契合。在水资源短缺形势日益严峻、灌区精细化管理日益迫切的背景下，研究基于水联网的灌区水资源管理及输配水过程自动控制，对于加快现代化灌区建设进程，提高水资源利用效率与效能具有重要意义。本研究全面梳理了灌区水资源管理中的水联网技术、明渠水力学、自动控制理论、渠系联控算法等理论基础，推导了比例积分微分控制、线性二次型最优控制与模型预测控制三种算法，并对比分析了各自的有效性、适用性与优劣性。针对大型灌区长渠池、多取水等特性，建立了控制结构调控动作、各取水口流量时滞干扰与非回水区蓄量转移等因素驱动的系统动态方程，提出了分段积分时滞模型。而后，以分段积分时滞模型为系统控制模型，以模型预测控制为控制算法框架，以差分进化算法在线辨识系统控制模型、径向基神经网络在线修正水位预测方程、并对权重矩阵进行自适应整定，提出了预测控制的自适应优化方法。基于鲁棒控制理论，分析灌区不确定性因素对渠系控制的影响机制，建立闸门流量滤波器减小流量观测的误差、建立反馈控制器减小过流公式反算开度的误差，综合以上，形成一整套可应对各类不确定性的自适应预测控制算法。以宁夏西干渠、秦汉渠两个大型灌区为案例，对本研究所建立的模型与算法进行综合应用与验证分析。分水位目标不变和水位目标增加两类情景，对比分析了自适应预测控制较线性二次型控制、模型预测控制的性能提升。并以秦汉渠为案例，分析了观测噪声、参数误差及动态干扰对水位控制性能的影响，验证了自适应预测控制算法对不确定性因素的应对能力，并识别出灌区硬件中制约自动控制应用的主要因素。本研究揭示了明渠输水中多取水扰动对水位变化的影响机制，分段积分时滞模型的水位拟合误差可从原有模型的1.69%降低至0.33%。研究提出了预测控制的自适应优化方法，使得灌溉过程中水位波动更小、调控过程更平稳，以最大绝对水位偏差、绝对水位偏差积分为主要评价指标，水位调控性能提升30%~60%。模拟结果显示自适应预测控制算法对多种不确定性因素均具有较好的应对能力，并识别出对水位调控影响最大的因素为流量观测精度、开度控制准度和渠系实时数据覆盖度，在灌区的自动化性能提升及现代化改造中应引起重视。
The control of water transmission and distribution processes in irrigation canal system has a significant impact on efficiency of water utilization. The manual control mode has imprecise water quantity regulation, difficulty in optimal scheduling, and high operation and management costs. Therefore, automatic control of irrigation canal system is an important tool to improve the efficiency of agricultural irrigation. The technical characteristics of Internet-of-Water, including real-time sensing and process tracking, fit with the irrigation canal system automation control technology. Under increasingly severe water shortage situation and the urgent need for fine-grained management of irrigation districts, the study of Internet-of-Water based irrigation district water resources management and automatic control of water transmission and distribution process is of great significance to accelerate the construction process of modern irrigation districts and improve the efficiency and effectiveness of water resources utilization. This study comprehensively reviews the theoretical basis of Internet-of-Water technology, open canal hydraulics, automatic control theory, and canal control algorithms in water resource management of irrigation district. Three control algorithms, Proportion-Integral-Derivative control, PID, Linear Quadratic optimal control, LQ and Model Predictive Control, MPC, are deduced, and the effectiveness, applicability, advantages and disadvantages of the three control algorithms are compared and analyzed. Considering the characteristics of large irrigation districts such as long canal pool and multiple water offtakes, the dynamic equations of the system, driven by the control structure regulation action, delayed disturbance of the flow at each offtake and movement of water storage in the non-return area, are established, and Segment Integrator Delay Model, SIDM is proposed. Based on Segment Integrator Delay Model and the framework of Model Predictive Control, adaptive optimization methods for predictive control is proposed, including online identification of system control model through the differential evolution algorithm, online correction of the water level prediction equation by the radial basis neural network, and adaptive adjustment of the weight matrix. Based on the robust control theory, the influence mechanism of various uncertainties in the irrigation district on the control of the canal system is analyzed. Using the gate flow filter to reduce the error of flow observation, establishing a feedback controller to reduce the error of inverse calculation of the over-flow equation, together with methods above, a complete set of adaptive predictive control algorithm that can cope with various uncertainties is established. The models and algorithms developed in this study were comprehensively applied and validated using two large irrigation districts, Ningxia Canal and Qinhan Canal, as case studies. The performance improvement of Adaptive Predictive Control over Linear Quadratic control and Model Predictive Control is verified in two scenarios: constant water level target and increased water level target. And using Qinhan Canal as a case, the effects of observation noise, parameter errors and dynamic disturbances on the water level control performance are analyzed, and the ability of Adaptive Predictive Control algorithm to cope with uncertainty factors is verified. Futhermore, the results indicate the major factors in the irrigation district hardware that constrain the application of automatic control. This study reveals the mechanism of the effect of multiple water offtakes on water level change in open channel water transmission, the percentage error of water level fitting can be reduced from 1.69% to 0.33% for the Segment Integrator Delay Model compared to the original model. The study proposes an adaptive optimization method of predictive control, which makes the water level fluctuation in the irrigation process smaller and the regulation process smoother, with the maximum absolute water level error and integral of absolute magnitude of error as the main evaluation index, and the water level regulation performance is improved by 30%~60%. The simulation results show that adaptive predictive control algorithm has good coping ability for a variety of uncertainties, and identify that the main factors, which have the greatest impact on the water level regulation and control process, are flow observation accuracy, gate opening control accuracy and real-time data coverage of the canal system, which should be paid more attention to in the improvement and modernization of the automatic control performance of the irrigation district.
