Papers by Author: John T. Agee

Abstract: Sensorless solar tracking involves the low speed tracking of the direction of sunrays across the sky. At such low speeds, frictional effects in the electromechanical drive system of the tracker become amplified, and can significantly affect the accuracy of tracking. This paper models the frictional effects in low speed solar tracking and shows that the frictional effects lead to significant positional error during tracking. Further results show that, a class of controllers derived from strictly positive real (SPR) transfer functions, could be used to eliminate the positional tracking error due to frictional effects.

Abstract: Predicting congestion in a telephone network has become part of an efficient network planning operation. The excellent capability of neural network (NN) to learn complex nonlinear systems makes it suitable for identifying the relationship between traffic congestion and the variables responsible for its occurrence in a time-varying traffic situation. This paper presents an artificial NN model for predicting traffic congestion in a telephone network. The design strategy uses a multilayered feedforward NN with backpropagation algorithm to model the telephone traffic situation. Matlab was used as a platform for all simulations. Regression analysis between predicted traffic congestion volumes and corresponding actual volumes gave a correlation coefficient of 87% which clearly shows the utility and effectiveness of Neural Networks in traffic prediction and control.

Abstract: Exponential nonlinearities in the thermistor limit the accuracy of the device for temperature measurement. This paper presents the design and implementation of a novel conditioning circuit for enhancing the accuracy of the sensor. Simulations of the conditioning circuit using MULTISIM were compared with the laboratory implementation. Results show that the proposed circuit could sufficiently linearise the thermistor characteristics and improve device sensitivity, throughout the range of sensor operation.

Abstract: Tracking solar-power devices often employ photovoltaic position sensors to detect the angle of misalignment between the axis of mounted solar panels and that of sunlight. The nonlinear input-output characteristics of this type of sensors tend to complicate controller design in such systems. This paper presents a nonlinear mathematical model of the photovoltaic position sensor. A three-layer feedforward neural network was trained to linearise the characteristics of the sensor. The MATLAB neural network tool (nntool) was used for neural network training. A final error of was obtained after training. Simulation of the neural network showed that linear sensor characteristics could be reproduced throughout the domain of sensor operation.

Abstract: Solar energy is increasingly becoming a significant component in the energy profiles of several tropical nations. This paper discusses trends in solar tracking technologies: analyzing the cost of acquisition, domains of application, maintenance costs and efficiency improvements. The paper concludes that hydraulic-based tracking systems are suitable for low capacity installations with low pay loads while polar axis tracking systems offer a performance nearly equal to that of two-axis tracking systems, at the cost of single axis trackers.