Hasan Arshad Nasir

Comparison of prediction error methods and subspace identification methods for rivers

Data based modelling is an important tool in the operation and management of rivers. In this paper, we compare Prediction Error Methods (PEM) and Subspace Identification Methods (SIM) for system identification of rivers. PEM can incorporate the available prior information and can accommodate the non-linearities in the model structure with ease. The models obtained by SIM are linear, and it is generally difficult to incorporate prior information, but SIM is well suited to MIMO systems such as rivers. In this paper, we simulate a few typical river scenarios and use the simulated data to obtain PEM and SIM based models. The models are compared using several measures and for the scenarios considered it is found that PEM has an edge over SIM.

System identification of the upper part of Murray river

Data based modelling is an important tool for the efficient planning and management of river systems. In this paper models of the water level in the upper part of Murray River in Australia are derived from observed operational river data by system identification methods. We consider Prediction Error Method (PEM) and Subspace Identification Method (SIM) and present the system identification procedure from collecting prior physical knowledge to model validation. It is shown that first order models capture the main trends in the data and perform well on validation datasets from different years.

A randomised approach to flood control using Value-at-Risk

Flooding is one of the major risks associated with rivers, and a typical operational goal is to reduce the risk of severe floods while at the same time not being overly cautious. In this paper we consider a Stochastic Model Predictive Control (S-MPC) based strategy which is well suited for rivers with uncertain in- and out-flows. In order to reduce the risk of floods, Value-at-Risk (VaR) is used as a risk measure and incorporated as a chance-constraint in the control optimisation problem. A computationally tractable scenario-based iterative optimisation and testing algorithm is proposed for solving the corresponding S-MPC problem, and its usefulness is demonstrated on a simulated river example.

Scenario based stochastic MPC schemes for rivers with feasibility ssurance

Water is a valuable resource, and improved management of rivers by using control techniques is receiving increased attention. Along a river there will typically be inflows from tributaries over which we have no control, but for which forecasts exist. The use of Stochastic Model Predictive Control (S-MPC) or a randomised version of it is a promising control strategy since it can accommodate such forecasts. However, due to uncertainties in the forecasts, the feasibility of the optimisation problem cannot be guaranteed in the presence of constraints. In this paper we consider two schemes for S-MPC of rivers that provide satisfactory results and guarantee feasibility via relaxation of the constraints. Simulation results on a real river show that the schemes perform well.

Control of a river stretch with uncertain inflows

Unregulated flows in tributaries to a river carry uncertainties, and due to the long time delays, these flows need to be forecast. For efficient river operations we need a control strategy that can take such forecasts into account and generate a suitable control action. A randomized approach to Stochastic Model Predictive Control for such problems is proposed in this paper. It involves solving a computationally tractable variant of a finite horizon chance-constrained optimization problem. Results from a simulation example show that the strategy performs well.

A randomised approach to Multiple Chance-Constrained Problems: An application to flood avoidance

One of the major risks associated with rivers is flooding, and a desirable way to manage rivers is to reduce the risk of severe floods without affecting the normal river operations. The flood risks are mainly contributed by uncertain inflows from tributaries. Due to uncertain in- and out-flows, the river control problem is formulated in this paper as a Multiple Chance-Constrained optimisation Problem (M-CCP), within a Stochastic MPC setting. M-CCPs are difficult to solve and this paper proposes an optimisation and testing algorithm to find approximate solutions of such problems. The algorithm is a significantly improved version of our previous proposal in [1]. Each step of the algorithm is supported with rigorous probabilistic bounds, and the usefulness of the algorithm is demonstrated on a simulated river example.

Control of rivers with flood avoidance

River control and decision support system can help improving water resource management. Rivers often have two modes of operations: normal operations, which include maintaining water levels and flows at their reference points, and flood avoidance. This work formulates the river control problem as a Stochastic Model Predictive Control (S-MPC), where a Multiple Chance-Constrained optimisation Problem (M-CCP) is solved at every time step. The M-CCP accommodates the constraints related to normal river operations and flood avoidance as two probabilistic constraints. We use the soft probabilistic constraints, because a hard constraint on the water level that depends on uncertain forecasts of the unregulated flows can cause infeasibility of the optimisation problems. We employ the Optimisation and Testing algorithm, proposed in [1], to solve the M-CCP and present simulation results of the control of the upper part of Murray River in Australia.