Program Topic: Modeling & Computing Methods

W4-2Wednesday 15:20-17:00

15:20 Comparing Bisection Numerical Algorithm with Fractional Short Circuit Current and Open Circuit Voltage Methods for MPPT Photovoltaic Systems

Muamer Shebani (Memorial University of Newfoundlan, Canada); Tariq Iqbal ( & Memorial University of Newfoundland, Canada); John Quaicoe (Memorial University of Newfoundland, Canada)

Muamer M. Shebani, Student Member, IEEE, Tariq Iqbal, and John E. Quaicoe, Senior Member, IEEE

Abstract—The maximum power produced by a photovoltaic (PV) system varies according to the variation in the solar irradiance and temperature. Maximum power point tracking (MPPT) algorithms are implemented to extract maximum power from PV system. This paper presents a bisection numerical algorithm (BNA) based MPPT, and it compares the algorithm’s tracking accuracy and performance to Fractional Short-Circuit Current (FSCC) and Fractional Open Circuit Voltage (FOCV) methods. This comparison uses the same DC-DC boost converter, PI controller, and load to examine the tracking accuracy for each method. The mathematical model for the PV system is developed using a single diode model, and it is implemented in Matlab/Simulink environment to examine each method. Simulation results for different solar irradiations are presented. The results show that the BNA has the best maximum power tracking accuracy in comparison with the FSCC and FOCV methods.

15:40 A Study on Power-Flow and Short-Circuit Algorithms Capable of Analyzing the Effect of Load Current on Fault Current Using the Bus Impedance Matrix

Insu Kim (Alabama A&M University, USA); Ronald Harley (Georgia Institute of Technology, USA)

In short-circuit studies, load makes difficult to determine fault current that flows in positive-, negative-, and zero-sequence networks. Therefore, it is often ignored because the magnitude of load current is much less than that of fault current that flows from generators when a fault occurs. As distributed generation resources such as photovoltaic systems, wind farms, and non-linear generators based on power electronics have being deployed into power system networks, load current that flows from them may affect a magnitude of fault current. Thus, the objective of this study is to develop a short-circuit algorithm that analyzes the effect of load current on fault current. For this purpose, this study initially develops a power-flow analysis algorithm that iterates to calculate current to be injected and determines voltage using the bus impedance matrix. Then, the proposed short-circuit algorithm uses as input data the results of the proposed power-flow algorithm. To verify the algorithms developed in MATLAB, a distribution system with a distributed generator is presented in a case study. Then, this study (a) calculates the power flow of a distribution system, (b) generates a single line-to-ground fault on the case study, and (c) changes the capacity of load, the capacity of a distributed generator, and the location of a fault. Finally, it examines the effect of load and distributed generation on a magnitude of fault current.

16:00 MATLAB/Simulink Modelling and Experimental Results of a PEM Electrolyzer Powered by a Solar Panel

Mohamed Albarghot (Memorial University of Newfoundland, Canada); Luc Rolland (Memorial University of Newfoundland, Canada)

Abstract – in this paper, the solar panels are used to power an electrolyzer to separate the water into hydrogen and oxygen gas. The electrical equivalent circuit for the proton exchange membrane electrolyzer was developed and implemented in MATLAB/Simulink along with the atmospheric hydrogen storage tank. The voltage (2 volt) and current (1 ampere) were supplied in a similar manner in order to compare the simulated and experimental results. The hydrogen amount is calculated to be 7.345 (ml/min A) from the model as well as the experimental set up. The experimental and simulation results were matched.

16:20 Analysis of Transients in a Micro-grid Using Wavelet Transformation

Yunqi Wang (University of New South Wales, Australia); Jayashri Ravishankar (University of New South Wales, Australia); Phu Le (University of New South Wales, Australia); Toan Phung (University of New South Wales, Australia)

The increasing penetration of distributed generation (DG) into the electrical network presents new challenges to reliable operation of system due to the special characteristics of DG units and their low inertia. Transient in the micro-grid causes more serious disturbances in the reliable operation of the network. The transients only occur for a few cycles, which are difficult to detect and cannot be identified by using digital measuring and recording instruments. This paper utilises wavelet transformation method to detect and analyze the transient signals in the micro-grid. The results show that disturbances are successfully observed by applying wavelet transform. The variation of coefficients is found to be much smaller in grid-connected mode.

16:40 An AC Power Flow Linearization for Power System Optimization Using Linear Programming

Torsten Sowa (RWTH Aachen University & Institute for High Voltage Technology, Germany); Alexander Stroband (RWTH Aachen University, Germany); Wilhelm Cramer (RWTH Aachen University, Germany); Simon Koopmann (RWTH Aachen University, Germany)

This paper presents a method for the linearization of the non-linear power flow equations, which can be used in mixed integer linear optimizations. The power flow equations are linearized around an operating point using the Taylor approximation. The linearization implies an approximation error, which can be iteratively reduced by modifying the operating point. In addition to existing approaches, controllable assets like voltage regulated transformers or phase shifters are integrated into the linearized grid constraints. The model is exemplarily applied to an operation planning model of distributed energy resources considering grid restrictions. The results show that this approach reduces the approximation error significantly and that it is robust for different real distribution grids as well as for different generation and load scenarios.

F1-1Friday 08:20-09:40

08:20 Design of Synchrophasor Estimation Systems with Convex Semi-Infinite Programming

Francisco Messina (Universidad de Buenos Aires, Argentina); Pablo Marchi (CSC-CONICET, Argentina); Leonardo Rey Vega (University of Buenos Aires, Facultad de Ingeniería & CONICET, Argentina); Cecilia Galarza (University of Buenos Aires, Argentina)

In this paper, we present a design methodology for synchrophasor estimation systems based on a convex semi-infinite optimization approach. Concretely, we present a suitable objective function for the problem and show how different performance constraints can be formulated. The advantage of this over most methods is that it allows us to control precisely the behavior of the system on both frequency and time domains. Thus, it gives an optimal and extremely flexible tool for a designer who wishes to obtain a particular desired performance. In particular, we show the advantages of this methodology by comparing it with standard designs on a study case on the basis of the IEEE Std. C37.118.1-2011.

08:40 Recursive Estimation of π-Line Parameters for Electric Power Distribution Grids

Alexander Prostejovsky (Technical University of Denmark, Denmark); Oliver Gehrke (DTU Electrical Engineering, Denmark); Anna Kosek (Technical University of Denmark, Denmark); Thomas Strasser (AIT Austrian Institute of Technology, Austria)

Electrical models of power distribution grids are used in applications such as state estimation and Optimal Power Flow (OPF), the reliability of which depends on the accuracy of the model. This work presents an approach for estimating distribution line parameters from Remote Terminal Unit (RTU) measurements which are subject to measurement device tolerances and random noise. Building upon an earlier work which introduced a measurement tolerance compensation model, we aim to improve a) the robustness towards noisy data and \b) the estimate of the parallel susceptance. For this purpose, we employ an Extended Kalman Filter (EKF) whose measurement noise covariance matrix is modified in order to account for all noisy variables in the overdetermined system. Simulations confirm the advantages of the EKF over the previously used Least-Squares (LSQ) estimator. In the low random noise cases considered in this paper, the EKF yields a four-fold improvement over the LSQ for the parallel susceptance across all quantization ranges. For the highest levels of random and quantization noise, the EKF performs about 1.5 to 3 times better than the LSQ for all line parameters. Furthermore, the EKF shows more consistent behavior when applied to data obtained from a laboratory distribution grid, which exhibits uncertainties that are not accounted for in the models.

09:00 Simulation Platform Developed to Study and Identify Critical Cases in a Future Smart Grid

Lucian Mihet-Popa (Oestfold University College & Politehnica University of Timisoara, Norway); Yi Zong (Technical University of Denmark, Denmark); You Shi (Technical University of Denmark, Denmark); Voicu Groza (University of Ottawa, Canada)

This paper proposes a simulation platform developed to study and identify critical cases in a Smart Grid. A distribution network with different Distributed Energy Resources (DER) Components, connected along the feeders, is analyzed, having the objective to identify limitations of existing simulation and planning tools, with a particular focus on the challenges imposed by the introduction of Smart Grid technologies. Another important issue of the paper is to identify critical load cases, as well as the voltage variations with the highest potential able to implement the grid model and the worst case scenarios developed

09:20 Dynamic Modeling of Diesel Generator Based on Electrical and Mechanical Aspects

Sihem Benhamed (Université du Québec à Rimouski, Canada); Hussein Ibrahim (Wind Energy TechnoCentre, Canada); Karim Belmokhtar (Wind Energy TechnoCentre, Canada); Hatem Hosni (Université du Québec à Rimouski, Canada); Adrian Ilinca (Université du Quebec à Rimouski, Canada); Daniel Rousse (Ecole de Technologie Superieure, University of Quebec, Canada); Ambrish Chandra (Ecola de Technologie, Superieure, Canada); DrishtySingh Ramdenee (ITMI, Canada)

Nowadays the studies of diesel generators are limited to present mechanical dynamic of the process or the electrical one, this is due to the complexity and the high no linearity of the DG. This paper gives a revue of different model which can describe the total dynamic process of the diesel generator. Then a developed model is proposed to study the interaction between the mechanical and electrical aspects in a DG. The validity of the proposed model is verified by an application in a study case. The developed model has been implemented in Matlab/Simulink, and the simulation result confirms the dynamic performances of the system compared with the operational data of the DG of TechnoCentre éolien (TCE).

F2-1Friday 11:00-12:20

11:00 SVD-based Reduced-Order Rational Approximation for EMT Analysis

Abner Ramirez (Center for Research and Advanced Studies of Mexico, Mexico); Edgar Medina (CINVESTAV-Guadalajara, Mexico)

This paper introduces a simple and effective model order reduction (MOR) approach based on singular value decomposition (SVD) and aimed to electromagnetic transient (EMT) analysis. The proposed approach initially adopts a rational approximation obtained by the vector fitting (VF) software tool for an arbitrary frequency range. Then, it applies SVD-based truncation separately to low-frequency range and to the error given by the original function and the low-frequency approximation in the high-frequency range. Finally, the resultant reduced approximations are assembled for EMT solution allowing the use of two different time-steps. The obtained reduced-order model achieves computational savings compared to the original full-size system given by VF. A case study is presented to validate the proposed method.

11:20 Non-Intrusive Load Monitoring Using Wavelet Design and Co-Testing of Machine Learning Classifiers

Jefferson Chung (University of Ontario Institute of Technology, Canada); Jessie Gillis (University of Ontario Institute of Technology, Canada); Walid Morsi (University of Ontario Institute of Technology, Canada)

This work aims to investigate the effect of implementing co-testing in Non-Intrusive Load Monitoring. Wavelet Design is used to extract features from the switching transients of loads. The features extracted are evaluated on a real test bed using One-Against-the-Rest and co-testing approaches. The results show the effectiveness of both approaches and it is seen that classification using co-testing provides a higher overall classifier accuracy. This is due to a significant decrease in misclassification occurrences.

11:40 A Single-Phase Dynamic Phasor Estimation System: Analysis of Performance and Computational Cost

Pablo Marchi (CSC-CONICET, Argentina); Francisco Messina (Universidad de Buenos Aires, Argentina); Leonardo Rey Vega (University of Buenos Aires, Facultad de Ingeniería & CONICET, Argentina); Cecilia Galarza (University of Buenos Aires, Argentina)

This paper presents a new single-phase technique for dynamic phasor estimation using a sample by sample digital processing approach. The estimation process is performed in three stages. In the first stage, an efficient FIR filtering of
the input signal is made to reject out-of-band and harmonic interference. The second stage uses a parameterized EPLL to track amplitude, phase, frequency and ROCOF. The final stage compensates the delay and any possible phase and amplitude
distortion. The proposed design is compared against a well known window DFT method known as IpD²FT, which requires a large buffer and iterative process for each report. The goal is to develop a PMU system compliant with the IEEE Std. C37.118.1a and compare both algorithms in terms of performance and computational cost.

12:00 Power System Tracking State Estimation Based on Stochastic Fractal Search Technique Under Bad Measurements Conditions

Hossam Mosbah (Dalhousie University, Canada); Mohamed El-Hawary (Dalhousie University, Canada)

Tracking State Estimation (TSE) predicts potential security contingencies. This paper discusses the application of Fractal Search which is a stochastic meta-heuristic algorithm under various scenarios including normal operation where load fluctuates linearly and noisy and bad data measurements introduced at different time instants during the study period. The proposed algorithm is tested on IEEE 5, 14, 30, and 57 bus systems and the results are compared with those obtained Genetic Algorithm (GA) and Particle Swarm Optimization (PSO) to demonstrate the validity of the proposed approach.

F3-1Friday 13:20-15:00

13:20 Evaluation of Wind Turbine Characteristics Built-in Model in Matlab Simulink

Abobkr Abobkr (College of Engineering Technology Houn, Libya); Mohamed El-Hawary (Dalhousie University, Canada)

Due to growing concerns over climate change, more and more countries are looking to renewable energy sources to generate electricity. Therefore, wind turbines are increasing in popularity. The amount of mechanical power that can be extracted from the wind by a turbine rotor depends on the characteristics of wind turbine. Furthermore, the aim of the characteristics of any wind turbine is to provide an expectation of how to design the generator’s control system in order to track these characteristics and extract maximum power from the wind. The characteristics of wind turbines is the focus of this paper. A new wind turbine characteristics methodology is proposed and summarized in a new flowchart. A Matlab simulator was written to evaluate the performance coefficient of wind turbine. Unlike the conventional simulations which use the wind turbine block in the MATLAB /SIMULINK library, a block was modeled by supplementing the drive train subsystem with aerodynamic model

13:40 Neural Network Backpropagation Algorithm Control for PEM Fuel Cell in Residential Applications

Mohamed Chemsi (Universite du Quebec à Trois-Rivieres, Canada); Kodjo Agbossou (Universite du Quebec à Trois-Rivieres, Canada); Alben Cardenas (Universite du Quebec a Trois-Rivieres, Canada)

Integration of renewable energy in residential applications is one of the most promising options to reduce the impact of human activities on the environment. Residential loads are generally AC loads with high variations generating low-frequency ripples that can harm the sources especially the fuel cell. This paper investigates the use of a neural network algorithm to control a power converter for fuel cell (FC) system supplying a thermostatic residential load. The proposed algorithm allows a limitation of the FC current ripple under 2% without compromising the transient response of the system and permitting the correctly supply the AC load. The model of a two stages conditioning system, including a DC-DC boost converter and a full bridge IGBT, single-phase inverter with their control, has been implemented using MATLAB/Simulink and validated by simulation

14:00 An Approach to Precise Modeling of Photovoltaic Modules Under Changing Environmental Conditions

Seyedkazem Hosseini (Université du Québec, Canada); Shamsodin Taheri (Université du Québec en Outaouais, Canada); Masoud Farzaneh (UQAC/CIGELE, Canada); Hamed Taheri (École de Technologie Stupérieure, Canada)

Equivalent circuit models of photovoltaic (PV) modules are used to investigate the performance of PV systems for different operating conditions. The electrical characteristics of PV modules change as environmental conditions vary. Hence, for an accurate study of the PV modules when insolation and temperature are variable, it is necessary to verify the dependence of the parameters of the PV module model to insolation and temperature. This paper proposes a single-diode model of PV modules based on a novel approach for extracting the climate dependent parameters. The parameters of the model are updated for given environmental conditions so that the model is able to precisely yield the corresponding real characteristics of the PV module. Comparison between the simulated model and experimentally measured data demonstrates the effectiveness of the developed approach in determining the PV characteristics under variable weather conditions. The proposed model can be helpful to properly design and select PV systems.

14:20 On Optimization of SVMs Kernels and Parameters for Electricity Price Forecasting

Abdussalam Mohamed (Dalhousie University, Canada); Mohamed El-Hawary (Dalhousie University, Canada)

Forecasting electricity price is a popular task in a modern electricity market. Forecasting of future prices allows market trading participators to gain insight and make decisions. Selecting the best training parameters is often involved in forecasting in order to obtain optimal prediction. Support Vector Regression (SVR) provides a well-established and powerful regression method to fit data and find minimal-risk slack variables around the fit line. Finding the best fit strongly depend on input feature set and the tuning of hyper-parameters. SVR quadratic programming (QP) and parameters optimization are applied to improve the forecasting accuracy. In this paper, quadratic programming optimization solving techniques are applied to improve the performance of SVMs, and a research conducted to study the influence of optimized hyperparameters using several known SVM kernels for electricity price forecasting.