Search

This research proposes a novel generic method for irrigation scheduling in a canal network to optimize multiple objectives related to canal scheduling (e.g. maximizing water supply and minimizing imbalance of water distribution) within multiple hierarchical layers (e.g. the layers consisting of the main canal, distributaries) while utilizing traditional canal scheduling methods. It is based on modularizing the optimization process. The method is theoretically capable of optimizing an unlimited number of user-defined objectives within an unlimited number of hierarchical layers and only limited by resource availability (e.g. maximum canal capacity and water limitations) in the network. It allows flexible decision-making through quantification of the mutual effects of optimizing conflicting objectives and is adaptable to available multi-objective evolutionary algorithms. The method’s application is demonstrated using a hypothetical canal network example with six objectives and three hierarchical layers, and a real scenario with four objectives and two layers.

In model-based irrigation control, the root zone soil moisture deficit (RZSMD) is maintained based on the water balance. To predict RZSMD in real-time, effective rainfall, irrigation and crop evapotranspiration need to be calculated online. Estimating the first two variables is more important yet tedious due to practical limitations of knowing the amount of water actually infiltrated into the soil. In order to solve this problem, we propose to apply system identification on water balance data to obtain a linear time series model. We further investigate how to carry out the modelling (i) under saturated conditions, (ii) when there is a rule-based irrigation control, and (iii) under measurement noise in the soil moisture readings. Using synthetic data we obtained a model fit above 80% in all cases. Additionally, we show the model optimality and applicability with an independent dataset, using residual tests. For two sets of field data, we observed model fits of 84% and 63%, and satisfaction in all residual tests. Simplicity in the model reduces calibration efforts whereas its linearity and adequacy recommend it for real-time irrigation control applications. In summary, the results indicate that a first order linear time series model based on system identification can successfully predict RZSMD in a real-time irrigation control system.

Irrigation underpins agricultural productivity. The purpose of irrigation is to match water supply to crop water demand. The effectiveness of irrigation depends on the quality of the timing and duration of watering events, also called irrigation scheduling. Most farmers use heuristic rules to determine irrigation events. This often leads to over-watering which results in lower crop yields and wasted water. By acquiring good estimates of a plant's water demand and local weather, it is possible to use optimization theory to compute an irrigation schedule that matches supply and demand thereby improving crop yields. Previous work has focused on scheduling irrigation over long time frames such as seasonal water allocations. Real-time irrigation scheduling, e.g. hourly or daily, has received little attention. Farmers rely on heuristic approaches implemented using simple spreadsheet tools to help them in this task. This approach cannot deal effectively with operational constraints and thereby results in poor performance. In this paper we develop a Model Predictive Control framework for real-time irrigation scheduling. The proposed formulation can take into account common operational constraints, including limitations on water availability as well as practical limits on the maximum or minimum amount of water that should be applied. We use measured climate data coupled with a simulation model to evaluate the proposed algorithm.

This paper presents a novel approach to real time automatic flood control in a managed river network that is subject to uncertain inflows. The proposed approach uses multiple models to represent inflows ranging from low to high flow. Optimal model selection is achieved in a minimum mean square error sense using a bank of Kalman filters to identify the most likely inflow characteristic. There are no a-priori probabilities assigned to the individual models. Model Predictive Control is used for water level controller design. Our Adaptive Multi Model Predictive Control (AMMPC) method is proposed as an alternative to existing techniques that also use multiple inflow models but with a-priori inflow model probabilities, either weighted or equally likely. The performance of the approach is demonstrated using a simulated river-reservoir model as well as using data collected at the Wivenhoe Dam during the 2011 floods in Queensland, Australia.

In many agricultural countries in the world, water is supplied to the crop fields through canal based distribution systems. Fields would have different crop types with varying water demands. When irrigating multiple fields, capacities of the canal and its outlets also need to be considered. This paper develops an integrated scheduling method which addresses both these concerns, in the context of a single canal based farm focusing on short term irrigation. The proposed methodology is a step towards real time irrigation scheduling in response to future weather and crop demands. At the first stage of the method, model predictive control (MPC) calculates the irrigation demand of the individual fields. Then, particle swarm optimization (PSO) optimizes the allocation of irrigation amounts based on these demands and the demands of neighboring fields. Integer zero programming is used for optimal delivery of the suggested allocations, by opening and closing the outlets. Simulations based on the model Aquacrop show that the method is capable of maintaining the soil moisture levels above wilting point at all times while utilizing the limited infrastructure capacities available.