Search

This paper presents a solution framework for the student project allocation (SPA) problem which is based on evolutionary algorithms (EAs), a biologically inspired stochastic, population-based search paradigm. Project-based assessment is a common component of engineering courses that are conducted in universities around the world. In their final year of study, a list of projects is made available by the academic staff and students are required to select a specific number of options from this list. The department then assigns a suitable project to each student such that preferred projects can be allocated to as many students as possible. While student interest is the primary criteria, several additional factors need to be considered such as project prerequisites, load balancing of staff commitments, and other specific university requirements. The allocation problem can therefore be seen as a complex multiobjective problem with multiple constraints. The EA-based project allocation system was recently developed and implemented in a large university department to automate this process, and to improve the matching of students to their desired projects. The solution which provides the highest level of satisfaction in meeting the varied objectives is then used to allocate projects to students. This new automated system is not only able to achieve a very high level of user satisfaction, but is also able to do so in a very short time, resulting in significant time savings. Index Terms--Evolutionary algorithm (EA), hybrid fuzzy evolutionary algorithm, student project allocation (SPA) problem.

The purpose of a voltage controller for uninterruptible-power-supply inverters is to produce stable output voltage with low distortion under all loading conditions, particularly under nonlinear loads and load transients. Since the output impedance of the inverter is the source for tracking error caused by load current, this paper proposes a simple digital feedback controller that focuses on reducing the output impedance of inverters by a feedback of the load current. The proposed control strategy ensures high-quality steady-state and dynamic responses from the inverter system. The design of the proposed digital controller is simple and requires only a reasonably accurate knowledge of the output L-C filter parameters. In addition, only the output voltage and the load current are sensed if the dc link voltage is constant. The results of simulations and experiments show that the proposed controller can achieve very low total harmonic distortion and fast dynamic response under varied loads including nonlinear loads with low switching frequency. Index Terms--Digital control, pole-placement, pulse-widthmodulation (PWM) inverter, uninterruptible power supply (UPS).

This paper proposes a controller for uninterruptible power supply inverters based on a particular type of onlinetrained neural network, which is called the B-spline network (BSN). Due to its linear nature and local weight updating, the BSN controller is more suitable for real-time implementation than conventional muitilayer feedforward neural controllers. Based on a frequency-domain stability analysis, a design methodology for determining the two main parameters of the BSN are presented. The model is found to be similar to that of an iterative learning control (ILC) scheme. However, unlike ILC, which requires a complex digital filter design that involves both causal and noncausal parts, the design procedure of the proposed BSN controller is straightforward and simple. Experimental results under various conditions show that the proposed controller can achieve excellent performance, comparable to that of a high-performance ILC scheme developed earlier. The proposed controller is an attractive alternative to both the multilayer feedforward neural controller and iterative learning controller in this and similar applications. Index Terms--Digital control, inverters, iterative methods, neurocontrollers, spline functions, uninterruptible power systems.

In this paper, two iterative-learning-based control methods for uninterruptible power system (UPS) inverters, viz., direct iterative learning controller (direct ILC) and hybrid iterative learning controller (hybrid ILC), are proposed. In both methods, a zero-phase filter that is designed in the frequency domain is applied in order to provide compensation for the resonant peak in the system model to ensure error convergence. Furthermore, a 'forgetting factor' is introduced in both control algorithms to increase the robustness of the scheme against measurement noise, initialization error, and/or variation of system dynamics due to any parameter drift. In the direct ILC method, ILC is combined with the feedforward of the reference to improve dynamic performance. This method is shown to be capable of achieving very high steady-state performance although its dynamic response was not very good. In the hybrid ILC method, ILC is combined with a proportional-derivative controller aside from the reference feedforward to improve the dynamic response further. Experimental results show that both proposed controllers can achieve very low total harmonic distortion and fast error convergence under different loads while using only one sensor. The proposed direct ILC is an effective solution for UPS products where high-quality steady-state output voltage is more important than fast dynamic response, while hybrid ILC can result in excellent steady-state performance with a more improved dynamic response than direct ILC. Index Terms--Digital control, inverters, iterative methods, uninterruptible power systems (UPSs).

This paper proposes a new hybrid neural network (NN) model that employs a multistage online learning process to solve the distributed control problem with an infinite horizon. Various techniques such as reinforcement learning and evolutionary algorithm are used to design the multistage online learning process. For this paper, the infinite horizon distributed control problem is implemented in the form of real-time distributed traffic signal control for intersections in a large-scale traffic network. The hybrid neural network model is used to design each of the local traffic signal controllers at the respective intersections. As the state of the traffic network changes due to random fluctuation of traffic volumes, the NN-based local controllers will need to adapt to the changing dynamics in order to provide effective traffic signal control and to prevent the traffic network from becoming overcongested. Such a problem is especially challenging if the local controllers are used for an infinite horizon problem where online learning has to take place continuously once the controllers are implemented into the traffic network. A comprehensive simulation model of a section of the Central Business District (CBD) of Singapore has been developed using PARAMICS microscopic simulation program. As the complexity of the simulation increases, results show that the hybrid NN model provides significant improvement in traffic conditions when evaluated against an existing traffic signal control algorithm as well as a new, continuously updated simultaneous perturbation stochastic approximation-based neural network (SPSA-NN). Using the hybrid NN model, the total mean delay of each vehicle has been reduced by 78% and the total mean stoppage time of each vehicle has been reduced by 84% compared to the existing traffic signal control algorithm. This shows the efficacy of the hybrid NN model in solving large-scale traffic signal control problem in a distributed manner. Also, it indicates the possibility of using the hybrid NN model for other applications that are similar in nature as the infinite horizon distributed control problem. Index Terms--Distributed control, hybrid model, neural control, online learning, traffic signal control.

Recently, there has been an increase in the application of fuzzy logic controllers (FLCs) for control of power electronic converters. Due to the FLCs' complex algorithm, their realization often calls for a compromise between cost and performance. In this paper, it is shown that the rule table of most of the two-input FLCs used with power converters can be approximated into a single nonlinearity. This allows the controller to be easily realized using simple, fast, and inexpensive analog circuits. The simplified 'nonlinear function controller (NLFC)' developed in this manner is shown to be equivalent in performance to the original FLC through simulations. The NLFC concept is then applied to PI-FLC, a type of FLC popular in power converter control applications. This results in the PI-FLC being replaced by a simple 'nonlinear PI controller (NPIC).' Using this simplification, the design of NPICs to obtain good dynamic performance in power converters is explained. An example design of NPIC for controlling a dc-dc boost power converter is presented. Experimental results are also presented to demonstrate the superior dynamic performance of the converter with NPIC versus that of a linear-PI controller. Index Terms--Dynamic response, fuzzy control, fuzzy logic, pulsewidth-modulated (PWM) power converters, switched-mode power supplies, Topelitz matrices.

There are many learning algorithms available in the field of pattern classification and people are still discovering new algorithms that they hope will work better. Any new learning algorithm, beside its theoretical foundation, needs to be justified in many aspects including accuracy and efficiency when applied to real life problems. In this paper, we report the empirical comparison of a recent algorithm RM, its new extensions and three classical classifiers in different aspects including classification accuracy, computational time and storage requirement. The comparison is performed in a standardized way and we believe that this would give a good insight into the algorithm RM and its extension. The experiments also show that nominal attributes do have an impact on the performance of those compared learning algorithms. Index Terms--Hyperbolic functions, machine learning, parameter estimation, pattern classification, polynomials.

A tri-state boost converter with an additional boost-inductor free-wheeling interval was proposed earlier to eliminate the right-half-plane (RHP) zero that occurs in the control-to-output transfer function of a classical boost converter under continuous-conduction mode of operation. A 'constant-Do' control scheme, in which the 'capacitor-charging' interval of the converter was kept constant, was employed. This resulted in a relatively large inductor current, especially under high line and load conditions, thereby causing high circuit losses. This paper proposes two variations of a novel dual-mode control (DMC) scheme that vary both the 'boost' and the 'capacitor-charging' intervals. Control analysis, design, and limitations of the proposed DMC schemes are presented. Through computer simulations and hardware experiments the performance of the tri-state boost converter with DMC schemes are compared with those of the tri-state boost converter with 'constant-Do' control scheme, and the classical PI-controlled boost converter. The DMC schemes achieve a significant (about 10%) improvement in converter's efficiency for a wide load range over the 'constant-Do' control scheme. Index Terms--Boost converter, 'constant-Do' control, dc-dc converter, dual-mode control, right-half-plane zero, small-signal analysis, tri-state boost converter.

A novel method using a reduced multivariate polynomial model has been developed for biometric decision fusion where simplicity and ease of use could be a concern. However, much to our surprise, the reduced model was found to have good classification accuracy for several commonly used data sets from the Web. In this paper, we extend the single output model to a multiple outputs model to handle multiple class problems. The method is particularly suitable for problems with small number of features and large number of examples. Basic component of this polynomial model boils down to construction of new pattern features which are sums of the original features and combination of these new and original features using power and product terms. A linear regularized least-squares predictor is then built using these constructed features. The number of constructed feature terms varies linearly with the order of the polynomial, instead of having a power law in the case of full multivariate polynomials. The method is simple as it amounts to only a few lines of Matlab code. We perform extensive experiments on this reduced model using 42 data sets. Our results compared remarkably well with best reported results of several commonly used algorithms from the literature. Both the classification accuracy and efficiency aspects are reported for this reduced model. Index Terms--Pattern classification, parameter estimation, pattern recognition, multivariate polynomials, and machine learning.

A challenging problem in the design of boost converters operating in continuous-conduction mode is posed by the dynamically shifting right-half-plane (RHP) zero in the converter's small-signal control-to-output transfer function. The paper proposes a novel tri-state boost converter without such a zero in the transfer function. The additional degree of freedom introduced in the converter in the form of a freewheeling interval has been exploited through an easy control technique to achieve this elimination. The absence of the RHP zero allows the control scheme to achieve larger bandwidth under closed-loop conditions, resulting in fast response. Analytical, simulation and experimental results of the tri-state boost converter have been presented and compared with those of the classical boost converter both under open-loop and under closed-loop operating conditions. The results clearly demonstrate the superior dynamic performance of the proposed converter. The proposed converter can be used in applications wherever fast-response boost action is needed. Index Terms--Boost converter, dc-dc converter, right-half-plane zero, small-signal analysis, tri-state boost converter.

The on-line implementation and results from a hybrid short-term electrical load forecaster that Is being evaluated by a power utility are documented In this paper. This forecaster employs a new approach involving a parallel neural-fuzzy expert system, whereby Kohonen's self-organizing feature map with unsupervised learning, is used to classify daily load patterns. Post-processing of the neural network outputs is performed with a fuzzy expert system which successfully corrects the load deviations caused by the effects of weather and holiday activity. Being highly automated, little human interference is required during the process of load forecasting. A comparison made between this model and a regression-based model currently being used in the Control Centre has shown a marked improvement in load forecasting results. Keywords : Artificial neural network, fuzzy expert system, hybrid fuzzy-neural model, short-term load forecast

FRESIM is a microscopic time-stepping simulation model for freeway corridor traffic operations. To enable FRESIM to realistically simulate expressway traffic flow in Singapore, parameters that govern the movement of vehicles needed to be recalibrated for local traffic conditions. This paper presents the application of a genetic algorithm as an optimization method for finding a suitable combination of FRESIM parameter values. The calibration is based on field data collected on weekdays over a 5.8 km segment of the Ayer Rajar Expressway. Independent calibrations have been made for evening peak and midday off-peak traffic. The results show that the genetic algorithm is able to search for two sets of parameter values that enable FRESIM to produce 30-s loop-detector volume and speed (averaged across all lanes) closely matching the field data under two different traffic conditions. The two sets of parameter values are found to produce a consistently good match for data collected in different days.

An integrated framework for modeling and evaluating the economic impacts of environmental dispatching and fuel switching is presented in this paper. It explores the potential for operational changes in utility commitment and dispatching to achieve least cost operation while complying to rigorous environmental standards. The work reported here employs a heuristics-guided evolutionary algorithm to solve this multiobjective constrained optimization problem, and provides the decision maker a whole range of alternatives along the Pareto-optimal frontier. Heuristics are used to ensure the feasibility of each solution, and to reduce the computation time. The capabilities of this approach are illustrated via tests on a 19-unit system. Various emission compliance strategies are considered to reveal the economic trade-offs that come into play.

The optimization of electromagnetic devices to meet an object with constraints is well understood and requires a starting point for the iterative process. Further, where there are several objects, a solution that is a compromise between the competing objects has to be accepted. Traditional exact methodology - such as gradients-based minimization of a suitable object function - for such problems is costly and therefore depends all the more on a good starting design to minimize cost. This paper makes use of fuzzy concepts in an expert system framework and proposes an efficient method for finding a reasonably approximate optimum solution for an electromagnetic system with competing objects. This may then be used as the starting solution for a more exact, but expensive process.

This paper describes the implementation and forecasting results of a hybrid fuzzy neural technique, which combines neural network modeling, and techniques from fuzzy logic and fuzzy set theory for electric load forecasting. The strengths of this powerful technique lie in its ability to forecast accurately on weekdays, as well as, on weekends, public holidays, and days before and after public holidays. Furthermore, use of fuzzy logic effectively handles the load variations due to special events. The Fuzzy-Neural Network (FNN) has been extensively tested on actual data obtained from a power system for 24-hour ahead prediction based on forecast weather information. Very impressive results, with an average error of 0.62% on weekdays, 0.83% on Saturdays and 1.17% on Sundays and public holidays have been obtained. This approach avoids complex mathematical calculations and training on many years of data, and is simple to implement on a personal computer.

A methodology for evaluation of transient stability of medium size interconnected longitudinal power systems has been developed using a hybrid neural network/pattern recognition approach. Assessment of transient stability is done using a fast pattern recognition algorithm at each load level, accurately predicted by a neural network on a half-hourly basis. As opposed to the conventional approaches, this hybrid strategy can make fast decisions with less computations.

This paper proposes a new technique for freeway incident detection using a constructive probabilistic neural network. Incident detection is one of the important components in Intelligent Transportation Systems. It identifies traffic abnormality based on input signals obtained from traffic flow sensors. To date, the development of Intelligent Transportation Systems has urged the researchers in incident detection area to explore new techniques with high adaptability to changing site traffic characteristics. Recent works show that the basic probabilistic neural network is one of the best choices for this purpose. However, it suffers from high memory requirement and the lack of practical model adaptation and network pruning methods. Recent work in probabilistic neural network (PNN) research has led to the development of constructive probabilistic neural network (CPNN), which incorporates a clustering technique with an automated training process. CPNN has been presented in this paper to solve two problems in traffic network incident detection. The work reported in this paper was conducted on Ayer Rajah Expressway (AYE) in Singapore for incident detection model development, and subsequently on I-880 freeway in California, for model adaptation. The developed model achieved incident detection performance of 92.00% detection rate and 0.81% false alarm rate on AYE, and 91.30% detection rate and 0.27% false alarm rate on I-880 freeway using the proposed adaptation method. In addition to its superior performance, the network pruning method employed here facilitated remarkable model size reduction by a factor of 11 compared to a conventional probabilistic neural network. A more impressive size reduction by a factor of 50 was achieved after the model was adapted for the new site. The results from this paper suggest that CPNN is a better adaptive classifier for incident detection problem with a changing site traffic environment. Index Terms--Incident detection, model adaptation, network pruning, probabilistic neural network (PNN).