Anoop K. Mathur

Parameter Estimation for Process Control with Neural Networks

Abstract Neural networks are applied to the problem of parameter estimation for process systems. Neural network parameter estimators for a given parametrized model structure can be developed by supervised learning. Training examples can be dynamically generated by using a process simulation, resulting in trained networks that are capable of high generalization. This approach can be used for a variety of parameter estimation applications. A proof-of-concept open-loop delay estimator is described, and extensive simulation results are detailed. Some results of other parameter estimation networks are also given. Extensions to recursive and closed-loop identification and application to higher-order processes are discussed.

Parameter estimation for process control with neural networks

An application of neural networks to the problem of parameter estimation for process systems is described. Neural network parameter estimators for a given parametrized model structure can be developed by supervised learning. Training examples can be dynamically generated using a process simulation, resulting in trained networks that are capable of high generalization. This approach can be used for a variety of parameter estimation applications. A proof-of-concept open-loop delay estimator is described, and extensive simulation results detailed. Some results of other parameter estimation networks are also given. Extensions to recursive and closed-loop identification and application to higher-order processes are discussed.

Automation in the food processing industry: distributed control systems

Abstract The demand for high-quality product, the flexibility to share equipment to manufacture several products and other factors have moved the food industries towards increased automation. Control system vendors have responded to these needs by providing appropriate hardware and modular software capability so the process engineer can concentrate on the process control strategy rather than the control system design. Also, the development of sensors that measure product quality and subjective properties such as taste, smell, etc., is providing new vistas in automation. This paper reviews the trends in sensors and control systems for the food processing industry. The advantages offered by the distributed control system (DCS) so far enjoyed by large, continuous process plants are now available to small users and to batch processing industries. Examples of DCS applications in the food industry are described.

REAL TIME ENERGY MANAGEMENT : CUTTING THE CARBON FOOTPRINT AND ENERGY COSTS VIA HEDGING, LOCAL SOURCES AND ACTIVE CONTROL

This article provides an analysis of the effect on the overall energy bill of a commercial facility of active energy management. We first show the benefits of pure hedging, hedging when the facility has its own power source—we consider the use of co-generation in winter and the use of solar power in summer. We next show how active control of facility temperature set points augments the benefits of the hedging and use of local power. Our studies are based on real consumption data of a large commercial facility, the corresponding real time prices of grid power, prices of natural gas, intensity of solar radiation, and temperature history of the period under consideration. We show that the combination of hedging, local power generation and active control can reduce facility energy bills by up to 30%, and bill variance by up to 80%. Thus, we have a scenario where consumers save significantly while using power sources with a smaller carbon footprint.Copyright

Control-Oriented Concentrated Solar Power Plant Model

We model the dynamics of solar thermal plants-the first model covering all processes between market demand through power output at millisecond resolution-for the purpose of control design. Our model integrates solar reflectors, power tower, salt tank, boiler, turbine, generator, piping, and pumps along with the flows of energy and information between them. We show how our detailed model permits control designs that can potentially enhance the life of the power tower, and thereby aid in reducing generation costs of solar power to a level competitive to current electricity price. Our modeling also opens up several problems of control and optimization of concentrated solar power (CSP) plants that can quickly bring the maturity of CSPs near that of traditional coal-fired plants and aid their integration with the other power sources. We show some of this potential via simulations.

Will the Smart Grid Be Stable: Approaches for Supply-Demand Imbalances

Electric utilities have been reluctant to embrace realtime pricing and the integration of renewables because both demand and supply will become uncertain. We quantify these risks as a function of solar/renewable penetration, and provide engineering analysis here to show it is possible to preserve the difference between supply and demand at levels manageable with current grid equipment and within the current deregulated market structure in the USA. First, we show how simple futures contracts and real-time energy management can motivate widespread acceptance of real-time pricing (RTP). Our illustrative calculations use data from a large commercial facility in New York. In the other parts of our analysis, we show how spatially distributed supplies of solar energy connected to the grid reduce unpredictability of power production. We also show a natural stabilizing effect - the correlation between local solar production and local consumption of air-conditioners.