Iroshani Jayawardene

Reservoir based learning network for control of two-area power system with variable renewable generation

The penetration of renewable energy sources into the electric power system is rapidly increasing. Integrating variable renewable energy sources into the transmission grid introduces challenges in real time power system operation. This causes power and frequency fluctuations and raises stability concerns. In this paper, a 200MW photovoltaic (PV) plant is integrated into a two-area four-machine power system. In order to maintain the system frequency, a dynamic tie-line power flow control is implemented using predicted PV power as an input to the automatic generation controller in Area 1, which transfers power to Area 2 with PV generation. The prediction performances of two learning reservoir based networks, an echo state network (ESN) and an extreme learning machine (ELM), are investigated for day and night time operations. The experimental study is performed using actual weather data from Clemson, SC and a real time simulation of a utility-scale PV plant integrated power system. Phasor measurement units (PMUs) are used to provide input signals to automatic generation controllers in the two area power system. Typical tie-line power flow control results based on ESN and ELM models are presented to show the impact of predicting PV power in improved automatic generation control with variable generation. ESN and ELM models provide minimal tie-line power flow deviations from reference power flows during day and night time operations, respectively.

Comparison of echo state network and extreme learning machine for PV power prediction

The increasing use of solar power as a source of electricity has introduced various challenges to the grid operator due to the high PV power variability. The energy management systems in electric utility control centers make several decisions at different time scales. In this paper, power output predictions of a large photovoltaic (PV) plant at eight different time instances, ranging from few seconds to a minute plus, is presented. The predictions are provided by two learning networks: an echo state network (ESN) and an extreme learning machine (ELM). The predictions are based on current solar irradiance, temperature and PV plant power output. A real-time study is performed using a real-time and actual weather profiles and a real-time simulation of a large PV plant. Typical ESN and ELM prediction results are compared under varying weather conditions.

Comparison of Adaptive Neuro-Fuzzy Inference Systems and Echo State Networks for PV Power Prediction☆

Abstract Use of photo-voltaic (PV) power as a source of electricity has been rapidly growing. Integration of large PV power into the grid operation introduces several challenges. Uncertainty of PV power generation causes frequency fluctuations and power system instabilities. Due to this, short term PV power prediction has become an important area of study. Short term PV power prediction supports proper decision-making in control centers. Power generation output of a PV plant is highly dependent on different weather conditions such as solar irradiance, temperature and cloud covers. Weather data analysis and prediction can be considered as big data due to its complexity and dynamically changing characteristics. In this paper, an adaptive neuro-fuzzy inference system (ANFIS) is developed and compared with an echo state network (ESN) for short term PV power prediction. The ANFIS approach consists of three ANFIS modules for predicting solar irradiance and temperature, and then estimating the PV power, respectively. The ESN on the other hand predicts the PV power based on current weather parameters. A weather database containing data sampled every second is used in developing the ANFIS and ESN based PV power prediction systems. Results are compared under different Clemson, SC weather conditions with the two approaches.

Denial of Service Attack on Tie-Line Bias Control in a Power System With PV Plant

The use of synchrophasor networks consisting of phasor measurement units (PMUs) makes it possible to monitor, analyzes, and control the electric power grid in real-time. PMU measurements of frequencies, currents, voltages, and phase angles are transmitted to system control centers through synchrophasor networks. Delayed or missing measurements from PMUs in closed-loop applications could lead to power system instability. Although the use of virtual private network (VPN) tunnels eliminates many security vulnerabilities, VPNs are still vulnerable to denial of service (DoS) attack that exploits a side-channel vulnerability. In this paper, the authors detail their analysis of DoS attack on a two-area four machine power system with a utility-scale photovoltaic (PV) plant. Automatic generation control (AGC) is used to implement a tie-line bias control in one of the areas. The impact of PMU packet dropping on AGC operation and countermeasures are presented in this paper. Cellular computational network (CCN) prediction of PMU data is used to implement a virtual synchrophasor network (VSN). The data from VSN is used by the AGC when the PMU packets are disrupted by DoS attacks. Real-time experimental results show the CCN based VSN effectively inferred the missing data and mitigated the negative impacts of DoS attacks.

Cellular computational extreme learning machine network based frequency predictions in a power system

Frequency is one of the most important characteristics in power system monitoring, control and protection. Frequency variations can be observed with significant changes in operating conditions. High penetration levels of renewable energy pose variability and uncertainty challenges for grid operation. It is essential to have innovative methodologies to take necessary actions to overcome these challenges. Power system frequency prediction provides an insight to better system control and protection. In this paper, a cellular computational extreme learning machine network (CCELMN) based frequency prediction approach is presented. Results are compared with those obtained with independent ELM models and persistence model and shown to outperform.

Frequency Prediction of Synchronous Generators in a Multi-Machine Power System with a Photovoltaic Plant Using a Cellular Computational Network

Power systems with synchronous generators and solar photovoltaic (PV) experience frequency and power fluctuations due to high variability of PV power. Automatic generation control is implemented to control power outputs of the generators and stabilize the system frequency. It is desirable with increasing levels of PV penetration to have foresight of frequency fluctuations to empower advanced control systems. A new methodology is presented in this paper for predicting frequency of synchronous generators in a power system with solar PV. A cellular computational network (CCN) is used to perform the frequency prediction over multi-time scale. CCNs are decentralized and distributed computing paradigms. Thus, CCNs are suitable for fast prediction of frequency of synchronous generators distributed spatially across a power system. The inputs to cells of the CCN are derived from phasor measurement unit (PMU) measurements of frequency and voltage phasor at the respective generator buses. Past, current and predicted measurements enable multi-timescale predictions of synchronous generator frequencies in a power system. Typical multi-time scale frequency predictions using the CCN are illustrated on a two-area four machine power system with solar PV integrated.

Tie-line bias control and oscillations with variable generation in a two-area power system

Integrating variable generation sources such as utility-scale photovoltaic (PV) plants into the transmission grid is desirable with the increasing quest for cleaner sources of electric power generation and reducing cost of utility-scale PV. In a multi-area power system, inter-area flows have to be controlled according to some power purchase agreements and transaction contracts. Automatic generation control (AGC) with knowledge of PV plant generation is introduced for variable tie-line bias control and maximum utilization of PV power in the power system. Typical results to illustrate the proposed real-time tie-line bias control are presented on a two-area four machine power system with a utility-scale PV plant. Furthermore, an analysis of the tie-line power oscillations due to cloud cover and disturbances for different levels of PV power penetration are carried out and the presence of low frequency modes are pointed out.

Situational intelligence for online coherency analysis of synchronous generators in power system

This paper presents a situational intelligence (SI) based approach to carry out coherency analysis of synchronous generator in a power system in an online manner. A cellular computational network (CCN) is used as the SI algorithm. CCN is a framework for distributed multi-timescale frequency prediction by utilizing the local and neighboring phasor measurement units (PMUs). The predicted frequency values are utilized for coherency analysis. The advantages of the CCN are scalability and distributedness which caters for on-line predicted coherency analysis for large power systems. The multi-time scale frequency predictions mitigates or minimizes delays in power system measurements and provides an insight to the power system coherent behavior apriori. The simulation studies on the New York-New England IEEE benchmark power system are presented to demonstrate that CCN based SI can be utilized in online coherency analysis. Predicted measurements can enhance resiliency to bad data. Furthermore, it is possible to utilize this approach for adaptive control of wide area power systems.