L.L. Lai

GA optimized PID controllers for automatic generation control of two area reheat thermal systems under deregulated environment

This paper proposes for a genetic algorithm (GA) tuned PID controllers for automatic generation control of two - area reheat thermal system under deregulated environment. A PID controller with its inherent superb capabilities of containing peak deviation, steady state errors and oscillatory behaviour of a dynamic system is an obvious choice for automatic generation control of interconnected power systems. The main issue towards the use of a PID controller is how to optimize the three gains of the controller. The trial and error method of finding the gains by indirect optimization using ISE technique with an appropriate performance index appears to be not wise enough because of its space complexity. GA appears to be the right choice in finding the optimum gains for the controllers. Hence, in this work, the gains of the PID controllers are optimized using floating point GA. The results of the GA optimized PID controllers on a two area reheat thermal system are compared with those with optimized through trial and error method. The GA optimized controllers are found to be superior in terms of peak transient deviation, settling times, and dynamic oscillations.

Wavelet-GA-ANN Based Hybrid Model for Accurate Prediction of Short-Term Load Forecast

This paper proposes a hybrid model developed through wiser integration of wavelet transforms, floating point GA and artificial neural networks for prediction of short-term load. The use of wavelet transforms has added the capability of capturing of both global trend and hidden templates in loads, which is otherwise very difficult to incorporate into the prediction model of ANN. Auto-configuring RBF networks are used for predicting the wavelet coefficients of the future loads. Floating point GA (FPGA) is used for optimizing the RBF networks. The use of GA optimized RBF network has added to the model the online prediction capability of short-term loads accurately. The performance of the proposed model is validated using Queensland electricity demand data from the Australian National Electricity Market. Results demonstrate that the proposed model is more accurate as compared to RBF only model.

Creating Efficient Visual Codebook Ensembles for Object Categorization

An image comprises information, such as color, texture, shape, and intensity, which humans use in parallel for perception. Based on this knowledge, three methods of constructing visual codebook ensembles are proposed in this paper. The first technique introduced diverse individual visual codebooks by randomly choosing interesting points. The second technique was based on a random subtraining image data set with random interesting points. The third method directly utilized different patch information for constructing an ensemble with high diversity. The codebook ensembles were learned to capture and convey image properties from different aspects. Based on these codebook ensembles, different types of image presentations could be obtained. A classification ensemble could be learned based on the different expression data sets from the same training image set. The use of a classification ensemble to categorize new images can lead to improved performance. The detailed experimental analyses on several data sets revealed that the present ensemble approaches were resistant to variations in view, lighting, occlusion, and intraclass variations. In addition, they resulted in state-of-the-art performance in categorization.

Wavelet transform and neural networks for fault location of a teed-network

A new technique using wavelet transforms and neural networks for fault location in a tee-circuit is proposed in this paper. Fault simulation is carried out in EMTP96 using a frequency dependent transmission line model. Voltage and current signals are obtained for a single phase (phase-A) to ground fault at every 500 m distance on one of the branches, which is 64.09 km long. Simulation is carried out for 3 cycles (60 ms) with step size /spl Delta/t, of 2.5 /spl mu/s to abstract the high frequency component of the signal and every 100 points have been selected as output. Two cycles of waveform, covering pre-fault and post-fault information are abstracted for further analysis. These waveforms are then used in wavelet analysis to generate the training pattern. Four different mother wavelets have been used to decompose the signal, from which the statistical information is abstracted as the training pattern. RBF network was trained and cross-validated with unseen data.

Continuous space optimized artificial ant colony for real- time typhoon eye tracking

For real-time typhoon eye tracking, artificial ant colony (AAC) methodology has been proved valuable for the efficient & effective identification of snake contour model boundary, which was built to simulate the real unclear typhoon eye whirly shape. While satellite digital photograph technology make it possible to capture real-time meteorological information; by means of constructing solution space and heuristic information, the contour of non-clear typhoon eye can be tracked intelligently. However, the practical conditions and meteorological phenomena are very complicated, only using discrete energy parameters as the heuristic information to lead the intelligent directing of ants are not reliable enough. In this paper, continuous space multi-kernel functions are introduced to optimize the heuristic information. In order to supply more practical factors for the energy converging procedure, corresponding Gaussian parameters calculation method will be given. In comparison, the iteration numbers can be decreased concerning same complexity of the problem to be solved, which proves that proposed optimization could provide the improvement on the practicability and effectiveness of original solutions.

Particle swarm optimization for economic dispatch of units with non-smooth input-output characteristic functions

This paper proposes an application of particle swarm optimization (PSO) to solve economic dispatch (ED) of units with non-smooth input-output characteristic functions. The IEEE 30-bus system with 6 generating units has been used as the simulation system to show the effectiveness of the algorithm. Results are compared to those detail by evolutionary programming (EP). It shows the PSO can produce slightly better results than those from EP

Fault location of a teed-network with wavelet transform and neural networks

A new technique using wavelet transforms and neural networks for fault location in a tee-circuit is proposed in this paper. Fault simulation is carried out in EMTP96 using a frequency dependent transmission line model. Voltage and current signals are obtained for a single phase (phase-A) to ground fault at every 500 m distance on one of the branches, which is 64.09 km long. Simulation is carried out for 3 cycles (60 ms) with step size /spl Delta/t, of 2.5 /spl mu/s to abstract the high frequency component of the signal and every 100 points have been selected as output. Two cycles of waveform, covering pre-fault and post-fault information are abstracted for further analysis. These waveforms are then used in wavelet analysis to generate the training pattern. Two different mother wavelets have been used to decompose the signal, from which the statistical information is abstracted as the training pattern. RBF network was trained and cross-validated with unseen data.

GA based algorithm for optimum allocation of reactive power under deregulated environment

An algorithm using floating point genetic algorithm (FPGA) was developed to solve the problem of optimum allocation of reactive power in power systems under open market environment. The performance of the proposed model is validated on IEEE-14 bus system with modifications to incorporate the varying working conditions of power systems like, change of tap settings of transformers, variable reactive power compensations, etc. Results of the FPGA demonstrate that the algorithm is well competent in achieving the near optimal allocation of reactive power under practical constraints and price based conditions.

GA tuned differential evolution for economic load dispatch with non-convex cost function

This paper proposes a genetic algorithm (GA) tuned differential evolution (DE) method for solving economic dispatch (ED) problem with non-smooth cost curves. The tuning of the weights in differential evolution is the key issue in designing an efficient differential evolution algorithm. Their values are dependent on nature and characteristic of objective function. As there is no explicit rule or guideline in determining these parameters, they are generally determined after a number of experimentations. In this paper floating point GA is used in tuning these parameters. The developed algorithm is experimented on a medium size of 40 units. The performance of the proposed algorithm is compared with standard Improved Fast Evolutionary Programming (IFEP) techniques. The simulation results demonstrate that GA tuned DE method is very efficient in finding higher quality solutions in high order non-convex ED problems.

Adaptive RBFN model for 2D spatial interpolation

As known, radial basis function (RBF) network is considered as an effective methodology to make prediction in spatial space, with spatial information fusion at different layers of RBF; the hidden layers fusion is able to give better result. The novelty of this paper is to propose an adaptive RBF network construction method, which combines the traditional incremental algorithm and real-time responsivity analysis. In the process of training, classification output error rate will be calculated to evaluate the responsivity. Experiments were carried out with practical weather sites data sets based on three other algorithms, i.e. Voronoi diagram, IDW (inverse distance weighted), topogrid algorithm. Compared results have shown that the proposed method has advantages in terms of both performance and precision. In addition, the adaptive attributes make it convenient to implement interpolation between variational source data sets.