A Perspective on the Research and Development of Applications of Artificial Intelligence in the Electric Power Industry

Viewed as manufacturing, the electric power industry is more similar to process industries such as petroleum, chemicals and pharmaceuticals than to the manufacture of automobiles, electronic devices and clothing by discrete, assembly-fine operations. However, it has most of the same challenges as these other industries, often carried to their extremes. For instanze, while electricity has essentially zero shelf llfe and it’s transportation is effectively instantaneous, a local fault in that transportation/transmission system can instantly halt deliveries everywhere. With the advent of deregulation, competition ~d unbundling, the business aspects of electric power are becoming virtually indistinguishable from most other manufacturing industries. Issues of maintenance scheduling and product mix are becoming far more important than ever before. In the future, various unbundled areas of the industry will have strong similarities, as well as some significant differe~es, to other industries as varied as telecommunications and retail sales. The purpose of this "Perspective" is to outline some ofthese similarities and differences, and to describe some of ways in which artificial intelligence i s being used to address these challenges in the context of the electric power industry. The Electric Power Research Institute The Electric Power Research Institute (EPRI) was formed in 1973 to apply advanced science and technology for the benefit of its member utilities and their customers. Funded through voluntary contributions by over 600 member utilities, EPRI’s work covers a wide range of technologies related to the generation, delivery, and use of electricity, with special attention paid to cost-effectiveness and environmental concerns. By far the majority of EPRI research is devoted to the development of products and procedures which are application-specific and me destined forthe immediate use ofits members. However, a portion of its effort, managed within the Strategic Research and Development Office, is directed toward exploring new technological ideas, pursuing advanced concepts, and fostering areas of science with potential for breakthroughs. Many of the EPRI’s product developments and research projects involve artificial intelligence. Background On The Electric Power Industry The North American power network may realistically be considered to be the largest machine in the world since its transmission lines connect all the electric generation and distribution on the continent. Computer simulations support all the planning and most of the operational control that goes into assuring the success of its primary function: to deliver bulk electric power from generation sources to load areas reliably and economically. Because the assurance of reliability has been the overwhelming oal in performing this function, artificial intelligence has had relatively little employment. However, power sector deregulation and privatization are now taking place worldwide, and with the advent of free competition in the electric power industry, new ways me being sought to improve efficiency without seriously diminishing reliability. Deregulation and the introduction of competition is being achieved through the unbundling of electrical services: converting the historic, vertical integration of generation, transmission and distribution into separate companies, or at least separate services, each optimizing its performance based on different criteria and all operating at arms length. Common wisdom, based on the experience of other industries and other nations, expects that in five years there will be only about fifty companies. The generation companies will be completely deregulated, except for some lingering environmental constraints. The distribution companies will still be regulated, along the lines of today’s local telephone companies, but major industrial/commercial customers, and cooperatives of individual residential customers, may generate their own power or buy it from the lowest bidder. The transmission Wildberger 219 From: Proceedings of the AI and Manufacturing Research Planning Workshop. Copyright © 1996, AAAI (www.aaai.org). All rights reserved. companics will bc partly regulated in an attempt to ensure open access and non-discriminatory pricing for "’wheeling" power between any generator and any user or distributor, while maintaining some level of system security despite their lack of control of either generation or load. On a limited basis, several utilities are beginning to employ remote control, distributed sensing, and communications methods to improve network performance. High-speed electronic power controllers such as FACTS (flexible AC transmission system) devices represent especially promising technology. By replacing the slow mechanical switches now used to manage system operations, these controllers offer for the first time the potential to dynamically fine-tune transmission so that power delivery can respond instantly to changcs in demand without the burden of maintaining large amounts of "spinning reserve" generation. However, these devices represent a two-edged sword. On the one hand, they me capable of controlling the system’s inherent modal behavior and directing the flow of active power to where it is wanted by accepting changes in reactive power elsewhere. But, on the other hand, increased use of highspeed electronic controllers like FACTS presents more opportunities for large disturbances to occur. In addition, the use of economically beneficial, high voltage direct current (HVDC) interconnectious also encourages dependence on geographically remote power sources, making the new controls associated with them essential for secure & stable operation. It is ironic that, just as system-wide control and information exchange is becoming possible through technology, political considerations are preventing it from taking place. However, this provides both an opportunity and a challenge for the use of artificial intelligence to aid each separately controlled part of the system in the optimization of its own profitability by adapting rapidly to changes in the other parts. Forecasting, producing estimates based on limited information, and calculating the cost/benefit of additional information will be essential to success in a market that will include complex "derivatives" based on future options to power production, transmission capacity and segmented delivery services. The electric power industry is similar in many ways to the telecommunications and transportation industries, but it has important differenceswhich cannot bc ignored in its practical operations, whether regulated or market driven, integrated or unbuudled. While electricity also "flows" (from high voltage to low voltage locations), its transmission is inherently different from that of gas or water: ̄ Power flows through the grid in inverse relation to the impedance on each line. ̄ Electric power systems use phase shiflers rather than valves. ̄ Providing the required flow on one line often results in "loop flows" on several other lines ̄ Despite batteries and capacitors, and in contrast to "line packing" of gas or the use of reservoirs for water, there is no practical way to store large amounts of electricity forany significant length of time. ̄ Reliable electric service is critically dependent on the ability to respond to changed conditions instantaneously. ̄ Global stability is cssential for local efficiency, but every local change has some effecton global stability. For an electric power network, there are three basic operating requirements: ̄ all components within their thermal ratings ̄ all voltages within upper and lower limits ̄ all generators synchronized Electricity has the shortest shelf-life of any product that we manufacture. Its perishability is partly compensated by the ability to transport it at almost the speed of light. But the infrastructure required forthat transportation is made up of local parts that have limited capacities, and the viability of the whole system depends in an extremely complex way on the performance of each of those parts. Knowledge-Based Operation & Maintenance