Salvador Carlos-Hernandez

A NEW NEURAL OBSERVER FOR AN ANAEROBIC BIOREACTOR

In this paper, a recurrent high order neural observer (RHONO) for anaerobic processes is proposed. The main objective is to estimate variables of methanogenesis: biomass, substrate and inorganic carbon in a completely stirred tank reactor (CSTR). The recurrent high order neural network (RHONN) structure is based on the hyperbolic tangent as activation function. The learning algorithm is based on an extended Kalman filter (EKF). The applicability of the proposed scheme is illustrated via simulation. A validation using real data from a lab scale process is included. Thus, this observer can be successfully implemented for control purposes.

Modelling and Analysis of the Anaerobic Digestion Process

Abstract In this contribution, the anaerobic digestion is modeled and analyzed. A detailed description of this biological process in two different structures is presented: completely stirred tank reactor and fluidized bed reactor. The modeling method is presented and a general mathematical model is deduced: the influence of biological and hydrodynamic phenomena in the model structure is shown. Perspectives to control and supervision are introduced and some results are presented.

Hierarchical fuzzy control for a wastewater anaerobic treatment plant

A hierarchical fuzzy control for a wastewater treatment anaerobic plant is presented. On the basis of a previously developed integrated control, which switches two controllers, two new variables are considered (COJ/X/sub 2/ and /spl Delta/QCH/sub 4/); next, the two controllers are transformed to fuzzy proportional-integral (PI) variables; and finally, Takagi-Sugeno fuzzy supervisor is developed in order to smooth the switching. Hence, a hierarchical fuzzy control structure is built: the low level is constituted by the fuzzy PI and the higher one by the fuzzy supervisor. This new strategy allows an increase in methane production. Applicability of the proposed structure is illustrated via simulations.

A neural observer with time-varying learning rate: analysis and applications.

In this paper, a reduced order neural observer (RONO) with a time-varying learning rate is proposed. The proposed scheme is based on a discrete-time recurrent high order neural network (RHONN) trained with an extended Kalman filter (EKF)-based algorithm. A time-varying learning rate is designed in order to improve the learning of the neuronal network in presence of disturbances and parameter variations. This work includes the stability proof of the time-varying learning. The applicability of the developed observer is illustrated via simulations for a nonlinear anaerobic digestion process.

Hybrid Intelligent Control Scheme for an Anaerobic Wastewater Treatment Process

A control strategy, composed by a neural observer and a fuzzy supervisor, for an anaerobic process is proposed in this paper. A recurrent high order neural observer (RHONO) is developed to estimate variables difficult to measure (biomass and substrate) in a completely stirred tank reactor (CSTR).The recurrent high order neural network (RHONN) structure is trained by an extended Kalman filter. The fuzzy supervisor uses estimations of biomass and methane production to detect biological activity inside the reactor and to apply an L/A (logarithm/anti-logarithm) control action if required in order to avoid washout. The applicability of the proposed scheme is illustrated via simulation.

Fuzzy Observer for the Anaerobic Digestion Process

Abstract In this paper a Takagi-Sugeno fuzzy observer for an anaerobic wastewater treatment plant is introduced. First, anaerobic digestion process is briefly explained. After that, Takagi-Sugeno systems are described. Next, observer design is presented and its performance is illustrated via simulations. Finally relevant conclusions are stated.

INTELLIGENT CONTROL STRATEGY FOR AN ANAEROBIC FLUIDIZED BED REACTOR

Abstract This paper deals with the control of the anaerobic digestion process in a fluidized bed reactor. The main idea is to develop a supervision mechanism which selects the most appropriate control action in function of the process state and the operating conditions. The supervisor is built on the basis of the Takagi-Sugeno algorithm and the control actions are implemented as fuzzy L/A PI controllers. The empirical knowledge is considered to build the fuzzy rules of the control strategy.

Neural network identification for biomass gasification kinetic model

This paper presents a neural network application to identify a kinetic model for the char reduction zone of a solid fuel gasification process. The considered model consists of six differential equations which represent the production of six components (carbon, hydrogen, carbon monoxide, water, carbon dioxide and methane) and are obtained from reaction rate equations of the four main reactions in the char reduction zone of a fluidized bed gasifier. On the other hand, the identification presented in this work is based on a discrete-time high order neural network (RHONN), which is trained with an extended Kalman filter (EKF) algorithm. The objective is to reproduce with the neural network the different components production under various operating conditions. The neural identifier performance is illustrated via simulation.

Neuro-fuzzy control strategy for methane production in an anaerobic process

In this paper, a neuro-fuzzy control strategy composed by a neural observer and fuzzy supervisors for an anaerobic digestion process is proposed in order to maximize methane production. A nonlinear discrete-time recurrent high order neural observer (RHONO) is used to estimate biomass concentration and substrate degradation in a continuous stirred tank reactor. A Takagi-Sugeno supervisor controller based on the estimation of biomass, selects and applies the most adequate control action, allowing a smooth switching between open loop and closed loop. The control law calculates dilution rate and bicarbonate rate based on speed-gradient inverse optimal neural control. Finally, Takagi-Sugeno supervisors calculate reference trajectories for the system states, and gain scheduling for the dilution rate control law at different operating points of the process. The applicability of the proposed scheme is illustrated via simulations.

Neurofuzzy Control Strategy for an Abattoir Wastewater Treatment Process

Abstract In this paper a neurofuzzy control strategy, composed by a neural observer and a fuzzy supervisor, for an anaerobic wastewater treatment process is proposed. The neural observer is based on a recurrent high order neural network (RHONN) which is trained by an extended Kalman filter. The main objective of the observer is to estimate methanogenic biomass, which is employed by a fuzzy supervisor. The tasks of this supervisor are: to detect the biological activity inside the bioreactor, to select and to apply an adequate control action depending on the operating conditions in order to avoid washout. The applicability of the proposed scheme is illustrated via simulations considering the model of a prototype bioreactor which is used to treat effluents collected from an abattoir.