Juan M. Grosso

Learning-based tuning of supervisory model predictive control for drinking water networks

This paper presents a constrained Model Predictive Control (MPC) strategy enriched with soft-control techniques as neural networks and fuzzy logic, to incorporate self-tuning capabilities and reliability aspects for the management of drinking water networks (DWNs). The control system architecture consists in a multilayer controller with three hierarchical layers: learning and planning layer, supervision and adaptation layer, and feedback control layer. Results of applyingthe proposedapproach to the Barcelona DWN show that the quasi-explicitnature of the proposedadaptive predictivecontrollerleads to improvethe computationaltime, especially when the complexityof the problemstructure can vary while tuning the receding horizons.

Economic MPC for the management of drinking water networks

This paper addresses the management of drinking water networks (DWNs) regarding a multi-objective cost function by means of economically-oriented model predictive control (EMPC) strategies. Specifically, assuming the water demand and the energy price as periodically time-varying signals, this paper shows that the EMPC framework is flexible to enhance the control of DWNs without relying on hierarchical control schemes that require the use of real-time optimisers (RTO) or steady-state target optimisers (SSTO) in an upper layer. Four different MPC strategies are discussed in this paper: a hierarchical two-layer approach, a standard EMPC where the multi-objective cost function is optimised directly, and two different modifications of the latter, which are meant to overcome possible feasibility losses in the presence of changing operating patterns. The discussed schemes are tested and compared by means of a case study taken from a part of the Barcelona DWN.

A service reliability model predictive control with dynamic safety stocks and actuators health monitoring for drinking water networks

This paper presents a model predictive control strategy to assure reliability in drinking water networks given a customer service level and a forecasting demand. The underlying idea concerns a two-layer hierarchical control structure. The upper layer performs a local steady-state optimization to set up an inventory replenishment policy based on dynamic safety stocks for each tank in the network. At the same stage, actuators health is revised to set up their next maximum allowable degradation in order to efficiently distribute overall control effort and guarantee system availability. In the lower layer, a model predictive control algorithm is implemented to compute optimal control set-points to minimize a multi-objective cost function. Simulation results in the Barcelona drinking water network have shown the effectiveness of the dynamic safety stocks allocation and the actuators health monitoring to assure service reliability and optimizing network operational costs.

Multi-layer Decentralized MPC of Large-scale Networked Systems

In this chapter, a multi-layer decentralized model predictive control (ML-DMPC) approach is proposed and designed for its application to large-scale networked systems (LSNS). This approach is based on the periodic nature of the system disturbance and the availability of both static and dynamic models of the LSNS. Hence, the topology of the controller is structured in two layers. First, an upper layer is in charge of achieving the global objectives from a set \(\mathcal {O}\) of control objectives given for the LSNS. This layer works with a sampling time \(\Delta t _1\), corresponding to the disturbances period. Second, a lower layer, with a sampling time \(\Delta t _2\), \(\Delta t _1 > \Delta t _2\), is in charge of computing the references for the system actuators in order to satisfy the local objectives from the set of control objectives \(\mathcal {O}\). A system partitioning allows to establish a hierarchical flow of information between a set \(\mathcal {C}\) of controllers designed based on model predictive control (MPC). Therefore, the whole proposed ML-DMPC strategy results in a centralized optimization problem for considering the global control objectives, followed of a decentralized scheme for reaching the local control objectives. The proposed approach is applied to a real case study: the water transport network of Barcelona (Spain). Results obtained with selected simulation scenarios show the effectiveness of the proposed ML-DMPC strategy in terms of system modularity, reduced computational burden and, at the same time, the admissible loss of performance with respect to a centralized MPC (CMPC) strategy.

On the assessment of tree-based and chance-constrained predictive control approaches applied to drinking water networks

Abstract Water systems are a challenging problem because of their size and exposure to uncertain influences such as the unknown demands or the meteorological phenomena. In this paper, two different stochastic programming approaches are assessed when controlling a drinking water network: chance-constrained model predictive control (CC-MPC) and tree-based model predictive control (TB-MPC). Under the former approach, the disturbances are modelled as stochastic variables with non-stationary uncertainty description, unbounded support and quasi-concave probabilistic distribution. A deterministic equivalent of the related stochastic problem is formulated using Booles inequality and a uniform allocation of risk. In the latter approach, water demand is modelled as a disturbance rooted tree where branches are formed by the most probable evolutions of the demand. In both approaches, a model predictive controller is used to optimise the expectation of the operational cost of the disturbed system.

Reliability-Based Economic Model Predictive Control for Generalised Flow-Based Networks including Actuators' Health-Aware Capabilities

Abstract This paper proposes a reliability-based economic model predictive control (MPC) strategy for the management of generalised flow-based networks, integrating some ideas on network service reliability, dynamic safety stock planning, and degradation of equipment health. The proposed strategy is based on a single-layer economic optimisation problem with dynamic constraints, which includes two enhancements with respect to existing approaches. The first enhancement considers chance-constraint programming to compute an optimal inventory replenishment policy based on a desired risk acceptability level, leading to dynamical allocation of safety stocks in flow-based networks to satisfy non-stationary flow demands. The second enhancement computes a smart distribution of the control effort and maximises actuators’ availability by estimating their degradation and reliability. The proposed approach is illustrated with an application of water transport networks using the Barcelona network as the case study considered.

A distributed predictive control approach for periodic flow-based networks: application to drinking water systems

ABSTRACT This paper proposes a distributed model predictive control approach designed to work in a cooperative manner for controlling flow-based networks showing periodic behaviours. Under this distributed approach, local controllers cooperate in order to enhance the performance of the whole flow network avoiding the use of a coordination layer. Alternatively, controllers use both the monolithic model of the network and the given global cost function to optimise the control inputs of the local controllers but taking into account the effect of their decisions over the remainder subsystems conforming the entire network. In this sense, a global (all-to-all) communication strategy is considered. Although the Pareto optimality cannot be reached due to the existence of non-sparse coupling constraints, the asymptotic convergence to a Nash equilibrium is guaranteed. The resultant strategy is tested and its effectiveness is shown when applied to a large-scale complex flow-based network: the Barcelona drinking water supply system.

Adaptive multilevel neuro-fuzzy Model Predictive Control for Drinking Water Networks

This paper presents a constrained Model Predictive Control (MPC) strategy enriched with soft-control techniques as neural networks and fuzzy logic, to incorporate self-tuning capabilities and reliability aspects for drinking water transport network management. The system is a multilevel controller with three hierarchical layers: neural level, fuzzy level, MPC level. Results in the Barcelona Water Network have shown that the quasi-explicit nature of the proposed predictive controller leads to improve the computational time, especially when the complexity of the problem structure can vary while tuning the receding horizons.

Non-centralized predictive control for drinking-water supply systems

This chapter discusses the application of non-centralized MPC (NCMPC) approaches to DWTNs. The aim of DMPC is to reduce the computational burden and to increase the scalability and modularity with respect to the centralized counterpart, but still maintaining a convenient level of suboptimality with respect to the desired control objectives. Moreover, the advantage of NCMPC approach is the simplicity of its implementation given the absence of negotiations among controllers, which allows for a simple implementation.

Stochastic model predictive control for water transport networks with demand forecast uncertainty

Two formulations of the stochastic model predictive control (SMPC) problem for the control of large-scale drinking-water networks are presented in this chapter. The first approach, named chance-constrained MPC, makes use of the assumption that the uncertain future water demands follow some known continuous probability distribution while at the same time, certain risk (probability) for the state constraints to be violated is allocated. The second approach, named tree-based MPC, does not require any assumptions on the probability distribution of the demand estimates, but brings about a complexity that is harder to handle by conventional computational tools and calls for more elaborate algorithms and the possible utilization of sophisticated devices.