An Optimal Design Method for Multilayer Feedforward Networks`Michael (original) (raw)
Related papers
Neural Network-based Power Flow Model
2022 IEEE Green Technologies Conference (GreenTech)
Power flow analysis is used to evaluate the flow of electricity in the power system network. Power flow calculation is used to determine the steady-state variables of the system, such as the voltage magnitude /phase angle of each bus and the active/reactive power flow on each branch. The DC power flow model is a popular linear power flow model that is widely used in the power industry. Although it is fast and robust, it may lead to inaccurate line flow results for some critical transmission lines. This drawback can be partially addressed by data-driven methods that take advantage of historical grid profiles. In this paper, a neural network (NN) model is trained to predict power flow results using historical power system data. Although the training process may take time, once trained, it is very fast to estimate line flows. A comprehensive performance analysis between the proposed NNbased power flow model and the traditional DC power flow model is conducted. It can be concluded that the proposed NN-based power flow model can find solutions quickly and more accurately than DC power flow model.
International Journal of Engineering Research and Technology (IJERT), 2014
https://www.ijert.org/assessment-and-comparative-study-of-different-enhanced-artificial-neural-networks-based-power-flow-solutions https://www.ijert.org/research/assessment-and-comparative-study-of-different-enhanced-artificial-neural-networks-based-power-flow-solutions-IJERTV3IS030670.pdf This paper presents the development of Neural Networks based fast load flow solution method, which can be used for such real time applications. A feed-forward model of the neural network based on back propagation algorithm (BP) and a radial basis function neural network (RBFNN) are proposed to solve the load flow problem under different loading/contingency condition for computing voltage magnitudes and angles of a power system. A comparative study is presented to assess the performance of different models of ANNs. The RBFNN has many advantageous features such as optimized system complexity, minimized learning, less computation time for training and simulation and recall times as compared to single layer and multi-layer perceptron models. The effectiveness of the proposed ANNs models for on-line application is demonstrated by computation of buses voltage magnitudes and voltages angles for different loading/contingency conditions in three typical test systems also, the Iraqi National Grid load flow problem is solved by two efficient ANN models. The proposed models (RBFNN) have been found to provide sufficiently accurate results and a robustness fast load flow solution which can be efficiently applied to on-line (real-time) implementation.
A Backpropagation Based Alternative to the Classical Power Flow Algorithm
Today, the power flow or load flow algorithm may be defined as an iterative procedure for approximating the steady state condition of a given power system. This methodology has sufficed for a long time and has been effective enough to meet power system planning demands, in most cases. However, a dilemma exists. The present basic power flow model, the Newton-Raphson, is not appreciably improvable in terms of speed of convergence and accuracy of solution. In fact, these two defining parameters oppose each other. The most successful attempts to improve upon the basic model either trade accuracy for speed (as in the Fast Decoupled Load Flow, which reduces the class of solvable problems) or vice versa (as in the Second Order Load Flow, which significantly complicates the algorithm. This paper proposes an alternative to the Newton-Raphson algorithm, namely an artificial neural network (ANN) method based on the backpropagation (BPN) learning algorithm. BPN, which has a feed forward network architecture and eliminates the divergence problems associated with iteration as well as significantly reducing the time of convergence. The accuracy of solution is determined by the range and amount of training provided; therefore, speed of convergence and accuracy of solution are independent of each other. A network architecture was developed and implemented and the results are detailed in this paper.
International Journal of Computer Applications, 2014
Determination of optimum feed forward artificial neural network (ANN) design and training parameters is an extremely important mission. It is a challenging and daunting task to find an ANN design, which is effective and accurate. This paper presents a new methodology for the optimization of ANN parameters as it introduces a process of training ANN which is effective and less human-dependent. The derived ANN achieves satisfactory performance and solves the timeconsuming task of training process. A Genetic Algorithm (GA) has been used to optimize training algorithms, network architecture (i.e. number of hidden layer and neurons per layer), activation functions, initial weight, learning rate, momentum rate, and number of iterations. The preliminary result of the proposed approach has indicated that the new methodology can optimize designing and training parameters precisely, and resulting in ANN where satisfactory performance.
Heuristic principles for the design of artificial neural networks
Information and Software Technology, 1999
Artificial neural networks have been used to support applications across a variety of business and scientific disciplines during the past years. Artificial neural network applications are frequently viewed as black boxes which mystically determine complex patterns in data. Contrary to this popular view, neural network designers typically perform extensive knowledge engineering and incorporate a significant amount of domain knowledge into artificial neural networks. This paper details heuristics that utilize domain knowledge to produce an artificial neural network with optimal output performance. The effect of using the heuristics on neural network performance is illustrated by examining several applied artificial neural network systems. Identification of an optimal performance artificial neural network requires that a full factorial design with respect to the quantity of input nodes, hidden nodes, hidden layers, and learning algorithm be performed. The heuristic methods discussed in this paper produce optimal or near-optimal performance artificial neural networks using only a fraction of the time needed for a full factorial design.
The author's reference source, 2020
Over the past two decades, the feedforward neural network (FNN) optimization has been a key interest among the researchers and practitioners of multiple disciplines. The FNN optimization is often viewed from the various perspectives: the optimization of weights, network architecture, activation nodes, learning parameters, learning environment, etc. Researchers adopted such different viewpoints mainly to improve the FNN's generalization ability. The gradient-descent algorithm such as backpropagation has been widely applied to optimize the FNNs. Its success is evident from the FNN's application to numerous real-world problems. However, due to the limitations of the gradient-based optimization methods, the metaheuristic algorithms including the evolutionary algorithms, swarm intelligence, etc., are still being widely explored by the researchers aiming to obtain generalized FNN for a given problem. This article attempts to summarize a broad spectrum of FNN optimization methodologies including conventional and metaheuristic approaches. This article also tries to connect various research directions emerged out of the FNN optimization practices, such as evolving neural network (NN), cooperative coevolution NN, complex-valued NN, deep learning, extreme learning machine, quantum NN, etc. Additionally, it provides interesting research challenges for future research to cope-up with the present information processing era.
Design of Multi-Layer Neural Networks for Butterworth Filter Optimization
2009
In this paper a proposed design of five multi-layer feed-forward Artificial Neural Networks (ANNs) is presented for optimized Butterworth filter. The first and second network perform Butterworth ideal Low Pass Filter (LPF) and typical LPF. The third ANN performs Band Pass Filter (BPF). The fourth network perform multi-BPF which consists of two layers, the first layer consists of six tansig neurons and the second layer consists of one purline neuron, and the fifth feed-forward network is designed to perform the High Pass Filter (HPF) which consists of three layers, the first layer consists of three tansig neurons, the second layer consists of three tansig neurons and the third layer consists of one purline neuron. Back-propagation training algorithm is used to train the proposed networks with Mean Square Error (MSE) equals 10-10. Simulation and test programs are implemented by using MATLAB.
The Optimal Multi-layer Structure of Backpropagation Networks
2006
A novel algorithm obtained by using Bayesian information criterion (BIC) is presented to systematically choose the optimal multilayer network structure, via the number of hidden layers and hidden nodes of each layer, of backpropagation (BP) networks. Simulation results with daily data on stock prices in the Thai market show that the algorithm performs satisfactorily. Moreover, the proposed algorithm is also compared to Daqi-Shouyi method.
MATHEMATICALLY DESIGNED ARTIFICIAL NEURAL NETWORKS WITH GAUSS NEWTON ALGORITHM
A simplified representation of some discrete problem is used to predict working of that problem. Mathematical model makes certain resembling on that problem. A multilayer Feedforward neural network is one of the computational mathematical model that can be performed as a linear or nonlinear mapping within the real world problems. Mathematically designed Artificial Neural Networks is applied on Multilayer Feedforward neural network with Backpropagation under Gauss Newton algorithm to overcome the complexity of conventional model. Also proposed model makes a prediction about real world problem which requires a model on classification for clinical findings and obtained results are compared with existing results.