Search

When supporting commercial or defense systems such as aircraft avionics, guidance and control, electronic warfare, or propulsion systems, a perennial challenge is providing effective test and diagnosis strategies to minimize downtime, thereby maximizing system availability. One can argue that one of the most effective ways to maximize downtime is to be able to detect and isolate as many faults that either exist or are emerging in a system at one time as possible. This is referred to as the "multiplefault diagnosis" problem, and it is a problem that is known to be computationally intractable (i.e., NP-complete) [1]. While several tools have been developed over the years to assist in addressing the multiple-fault diagnosis problem, considerable work remains to provide the best diagnosis possible given the information collected through observations, gripes, test results, and historical data. Recently, a new model for evolutionary computation has been developed called the "Factored Evolutionary Algorithm" (FEA) [2]. In FEA, a target optimization problem is broken down into subproblems that exhibit some kind of overlap. Then the optimization algorithm of choice (e.g., simulated annealing, genetic algorithm, particle swarm optimization) is applied to each of the subproblems, and the subproblems periodically share information with neighboring subproblems along the points of overlap. [ABSTRACT FROM AUTHOR]

Due to uncertainties in generation and load, optimal decision making in electrical energy markets is a complicated and challenging task. Participating agents in the market have to estimate optimal bidding strategies based on incomplete public information and private assessment of the future state of the market, to maximize their expected profit at different time scales. In this paper, we present an agent-based model to address the problem of short-term strategic bidding of conventional generation companies (GenCos) in a power pool. Based on the proposed model, each GenCo agent develops a private probabilistic model of the market (using dynamic Bayesian networks), employs an online learning algorithm to train the model (sparse Bayesian learning), and infers the future state of the market to estimate the optimal bidding function. We show that by using this multiagent framework, the agents will be able to predict and adapt to approximate Nash equilibrium of the market through time using local reasoning and incomplete publicly available data. The model is implemented in MATLAB and is tested on four test case systems: two generic systems with 5 and 15 GenCo agents, and two IEEE benchmarks (9-bus and 30-bus systems). Both the day-ahead (DA) and hour-ahead (HA) bidding schemes are implemented. The results show a drop in market power in the 15-agent system compared to 5-agent system, along with a Pareto superior equilibrium in the HA scheme compared to the DA scheme, which corroborates the validity of the proposed decision making model. Also, the simulations show that introduction of an HA decision making stage as an uncertainty containment tool, leads to a more stable and less volatile price signal in the market, which consequently results in flatter and improved profit curves for the GenCos. [ABSTRACT FROM PUBLISHER]

There exists a wide field of strong theoretical models, which provide a basis for diagnostic and probabilistic reasoning. Among these are Bayesian methods, decision trees, D-matrices, logic based frameworks, and more. The IEEE 1232?2010 Standard for Artificial Intelligence Exchange and Service Tie to All Test Environments (AI-ESTATE) [1] strives to provide a common interface for interacting with these reasoning models to establish a common platform for prognosis and diagnosis. In this work, we propose a platformagnostic net-centric framework and interface, conformant to the AI-ESTATE standard. The IEEE AI-ESTATE provides platform-agnostic models and a service interface facilitating the exchange of diagnostic knowledge and reasoning capabilities. The Numerical Intelligent Systems Laboratory at Montana State University has been working continuously on promoting and enhancing this standard. Within AI-ESTATE is defined a set of Reasoner Manipulation Services (i.e., a service API). These services provide a standardized way to interact with a reasoning server for the purposes of diagnostic and prognostics. Previously, the AIESTATE service API had been implemented locally within our tool, called the Standards-based Analysis Platform for Predictive Health and Intelligent Reasoning Environment (SAPPHIRE). [ABSTRACT FROM PUBLISHER]