Communicating with Living Neuronal Networks on Microelectrode Arrays for Computation and Control

Robert L. Ortman, Christopher J. Rozell, Ganesh Kumar Venayagamoorthy and Steve M. Potter

Effective input patterns must evoke computationally useful responses in cultures of rat cortical neurons on MEAs. Candidate electrical stimulation patterns are evaluated for evoked-response separability and reliability via a support vector machine (SVM)-based method. Genetic algorithm is used to construct subsets of highly separable patterns.

Computation Using Latency Dynamics in Living and Artificial Neural Networks

Riley T. Zeller-Townson, Ganesh Kumar Venaygamoorthy and Steve M. Potter

Spiking Neural Networks use the precise timing of action potentials to convey meaning. The conduction delays between neurons are one set of parameters that can be tuned to improve network performance on computational tasks, however no biologically inspired delay learning rules have been adopted by the artificial neural network community. This work shows the computational properties of delay update rules that are based on how delay change in living neural networks, as well as how the actual biological data can be used to improve performance for a prediction task.

Computation with Action Potential Delay Dynamics

Riley T. Zeller-Townson, Jonathan P. Newman, Ganesh K. Venayagamoorthy and Steve M. Potter

Computation with Action Potential Delay Dynamics

Biologically Inspired Artificial Neural Networks, such as Spiking Neural Networks (SNNs), promise to provide significant advances over classic Artificial Neural Networks (ANNs) by performing computations in ways similar to the living brain. SNNs use discrete action potentials, which require a finite amount of time to travel between neurons. Most SNNs assume this axonal conduction delay to be constant, in spite of growing biological evidence that this conduction delay shows both long term and short term plasticity. We are working to explore the computational implications of these dynamics.

Reservoir-Computing-Based, Biologically-Inspired Artificial Neural Network (BIANN) for Online Modeling of a Single Machine Infinite Bus (SMIB) System

Jing Dai, Ronald G Harley, Ganesh Kumar Venayagamoorthy and Steve M. Potter

BIANN uses biologically plausible spiking neuron models. It bridges the gap between oversimplified ANNs and living neural networks. Effective encoding, decoding and training mechanisms for BIANN still need to be developed. A reservoir computing based training approach is proposed for the BIANN to serve as a novel modeling and control tool for practical applications. The BIANN is able to provide accurate one and five steps-ahead predictions of the rotor speed and terminal voltage of a generator in a SMIB, for online monitoring purposes.

Dynamic Performance Model for Wind Turbine Generators

Prajwal Gautam and  Ganesh Kumar Venayagamoorthy

Wind turbine power curves are based on the industry standard IEC 61400-12-1. Power curves are used for planning purposes and estimating total wind power production. Wind velocity are collected and averaged over 10-minute periods.  Traditional methods do not explain varying characteristics in wind dynamics where multiple power productions are observed for same wind speed. When the input parameters such as wind speed and wind directions are known and the output parameter wind power are known for an installed wind turbine generation plant, a dynamic computational network such as neural network is used to develop operation model and estimate the wind power generation.

Scalable Monitoring and DSOPF Control for Smart Grids

Karthikeyan Balasubramaniam and Ganesh Kumar Venayagamoorthy

An adaptive, optimal, real-time controller based on adaptive critics design called dynamic stochastic optimal power flow (DSOPF) controller is proposed. Stochastic nature in power system can arise as a result of load and generation stochastic behaviors and due to random noise in PMU data which arises due to communication noise and measurement error. DSOPF controller can perform real-time control action but system wide information cannot be made available to DSOPF controller in real-time because of power system communication delays which can range from a few milliseconds to several seconds depending on distance and communication media.

If state variables can be predicted ahead of time, then communication delay can be compensated for. Hence, a scalable wide area monitoring system that can predict state variables ahead of time is developed. Scalability is achieved by using cellular architecture called cellular computational network (CCN). This module can effectively compensate for communication delays and hence can enable DSOPF controller to perform real-time control with system wide information.

Scalable Integrated Situational Awareness System for Smart Grid

Bipul Luitel and Ganesh Kumar Venayagamoorthy

In a smart grid, monitoring of system variables such as voltages and speed deviations of generators is important for assessing its stability, and making proper control decisions. Development of wide area monitoring system is, hence, important for situational awareness; especially in a smart grid where integration of renewable resources, distributed generation and bidirectional power flows can lead to instabilities if proper control action is not taken at the right time, place and context.

Emergent Intelligence in Cellular Neural Networks

Bipul Luitel and Ganesh K. Venayagamoorthy

Emergent Intelligence in Cellular Neural NetworksIntelligence in CNN emerges from a group of cells that are learning subsystems, all of which either utilize the same or  a different learning method, that either adapt themselves in synchrony with the other subsystems or independent of the others in their own pace. Intelligence in CNN emerges over time through progressive learning and adaptation of these distributed interacting subsystems represented as cells.

Damping Electromechanical Oscillations in Large-Scale Power Systems Using Intelligent Aggregated Control

Diogenes Molina, Ganesh Kumar Venayagamoorthy and Ronald G Harley

Poorly damped oscillations can constraint the safe operating region of power systems, prevent more economical operation, and increase the probability of wide-spread blackouts. Controllers capable of monitoring and injecting signals at multiple generating stations across the system can help mitigate these oscillations and improve overall performance. Methodologies for designing such controllers using approximate dynamic programming system aggregation techniques are proposed.