Distributed Intelligent Agents for Decision Making at Local Der Levels

Distributed Energy Resources (DER), when integrated properly with the electric grid can provide both economic and societal benefits. Economic benefits in the form of reduced energy costs for end users that employ DER are well documented. Societal benefits in the form of achieving a more robust electric grid, less vulnerable to terrorist attack, are still more difficult to quantify. But it is safe to say that maintaining the historically high reliability of the U.S. power system is a fundamental requirement for the U.S. economy. Seamless penetration of DER assets is only possible with increased communications and information flow at all levels. Intelligent software agents can provide the needed communications, information and control enhancements. The project team, Alternative Energy Systems Consulting, Inc. (AESC) and the Center for Networked Distributed Energy (CNDE) at Colorado State University developed a conceptual design for a multilevel agent/agency hierarchy for use in coordinating grid power transactions from the distribution up to the transmission level. The project team further refined the agent based concept at the distribution level into the “Power Neighborhood” agent-based control concept. In this case, a power neighborhood consists of multiple sites collaborating via a Dutch auction-based process to allocate DER resources in response to dynamic pricing or grid disturbances. In order to evaluate the feasibility of the power neighborhood concept, AESC developed the Distributed Intelligent Agent for Decision Making (DIADM) simulation software. The DIADM simulation software was subsequently installed on computer(s) at Colorado State University and used to test basic functionality and the feasibility of the agent-based approach. The test scenarios examined by CNDE personnel were selected to demonstrate the feasibility of a decentralized architecture for power system control using agent-based modeling. Testing revealed that the site agents were able to cover their individual site critical loads, and their collective neighborhood loads by establishing bilateral transactions with other site agents (both within their own and adjacent neighborhoods) and the distribution substation agent via the auction website. The agents operating within the power neighborhoods were able to allocate DER assets amongst the various sites in both an hour-ahead auction environment as well as in response to a signal indicating the loss of grid supplied power. Testing showed that the agents, operating via Dutch auctions and without a central authority, quickly allocated DER assets and converged on a solution. Testing also showed how neighborhoods with disparate DER penetration levels are able to both reallocate resources within their respective neighborhoods as well as contribute to an adjacent neighborhood’s ability to respond. In doing so, neighborhoods with excess DER assets were able to reduce their overall costs through the sale of excess capacity. The test results are qualitative in nature since timing and performance constraints imposed by both DER asset performance and the latency of inter-agent communications were not simulated. However, test results clearly show that a network of agents, acting within “power neighborhoods” can quickly allocate resources both in response to dynamic pricing and if signaled that a partial or total loss of grid supplied power is eminent. Furthermore, that use of a Dutch auction based process eliminates the need for a centralized control thus providing an inherently more open and extensible solution. DISTRIBUTED INTELLIGENT AGENTS FOR DECISION MAKING AT LOCAL DER LEVELS Project Final Report for the Period July 2003 – December 2004 Date Issued/Published December 2004 Gerald L. Gibson, PE Alternative Energy Systems Consulting, Inc. Wade Troxell, Ph.D. Colorado State University PREPARED FOR THE UNITED STATES DEPARTMENT OF ENERGY OFFICE OF FOSSIL ENERGY Work Performed Under Contract No. DG02-03ER83604