Action-Dependent Adaptive Critic Design Based Neurocontroller for Cement Precalciner Kiln

Full Text (PDF, 178KB), PP.60-67

Views: 0 Downloads: 0

Author(s)

Baosheng Yang 1,* Deguang Cao 2

1. Suzhou Univertity/ Department of computer science and technology, Suzhou, China

2. Guangxi University/ School of Chemistry & Chemical Engineering, Nanning, China

* Corresponding author.

DOI: https://doi.org/10.5815/ijcnis.2009.01.08

Received: 1 Apr. 2009 / Revised: 10 Jun. 2009 / Accepted: 16 Aug. 2009 / Published: 8 Oct. 2009

Index Terms

Adaptive critic design, ADACD, neural network, controller, precalciner kiln system

Abstract

There are many factors that can affect the calciner process of cement production, such as highly nonlinearity and time-lag, making it very difficult to establish an accurate model of the cement precalciner kiln (PCK) system. In order to reduce transport energy consumption and to ensure the quality of cement clinker burning, one needs to explore different control methods from the traditional way. Adaptive Critic Design (ACD) integrated neural network, reinforcement learning and dynamic programming techniques, is a new optimal method. As the PCK system parameters change frequently with high real-time property, ADACD (Action-Dependant ACD) algorithm is used in PCK system to control the temperature of furnace export and oxygen content of exhaust. ADACD does not depend on the system model, it may use historical data to train a controller offline, and then adapt online. Also the BP network of artificial neural network is used to accomplish the network modeling, and action and critic modules of the algorithm. The results of simulation show that, after the fluctuations in the early control period, the controlled parameters tend to be stabilized guaranteeing the quality of cement clinker calcining.

Cite This Paper

Baosheng Yang, Deguang Cao, "Action-Dependent Adaptive Critic Design Based Neurocontroller for Cement Precalciner Kiln", International Journal of Computer Network and Information Security(IJCNIS), vol.1, no.1, pp.60-67, 2009. DOI:10.5815/ijcnis.2009.01.08

Reference

[1] R.Howard. Dynamic Programming and Markhov Processes[M]. MIT PTess, 1960.

[2] D.Bertsekas. Dynamic Programming and Optimal Control[M]. Belmont, MA: Athena Scientific, 1995.

[3] P. Werbos. “Approximate dynamic programming for realtime control and neural modeling” in Handbook of Intelligent Control: Neural, Fuzzy, and Adaptive Approaches (Chapter 13)[M]. New York, NY: Van Nostrand Reinhold, 1992.

[4] P.J., Werbos, Handbook of Intelligent Control. Van Nostrand Reinhold, New York, 1992.

[5] J. J., Murray, Cox C J, Lendaris G G, and Saeks R, “Adaptive Critic Design,” IEEE Trans. On Syst., Man, and Cybern, vol. 32, 2002, pp.140-153.

[6] T., Kobayashi, A., Yokoyama, An adaptive neuro-control system of synchronous generator for power system stabilization. IEEE Transactions on Energy Conversion, vol. 11, 2006, pp. 621–630.

[7] D.C. Hughes, D.B. Sugden, D. Jaglin, D. Mucha. Calcination of Roman cement: A pilot study using cementstones from Whitby, Construction and Building Materials, vol. 22, 2008, pp.1446-1455.

[8] G. Genona, E. Briziob. Perspectives and limits for cement kilns as a destination for RDF, Waste Management, vol. 28, 2008, pp.2375–2385.

[9] K. H., Karstensen. Formation, release and control of dioxins in cement kilns, Chemosphere, vol. 70, 2008, pp. 543-560.

[10] D., Liu, X., Xiong, Y., Zhang, Action-dependent Adaptive Critic Designs. In: Proceedings of the INNS-IEEE International Joint Conference on Neural Networks, Washington, DC, vol. 7, 2001, pp. 990–995.

[11] Kim, B.H. A study on the convergency property of the auxiliary problem principle. Journal of Electrical Engineering & Technology, vol. 1(4) , 2006, pp. 455–460.

[12] Si, J., Wang, Y.T. On-line learning control by association and reinforcement. IEEE Transactions on Neural Networks, vol. 12, 2001, pp. 264–276.