Electric power network decision effects.

INTRODUCTION

Electric power network system design and operation calls for a large number of decisions by many decision-makers over an extended period of time. These decisions are typically made in isolation and do not take into account the fact that in such an interconnected network, it is difficult to take an action that affects only the decision-maker (Watts, 2003). In increasingly interconnected engineering systems, others are not only affected by our actions (decisions), but we are in turn influenced by the actions and reactions of others. The butterfly effect, which proposes that "the flap of a butterfly's wings in Brazil [can] set off a tornado in Texas," acknowledges the connected nature of the world that produces unanticipated impacts (Hilborn, 2004). Decision-makers need to acknowledge their networked reality by considering prescriptive, descriptive, and normative decision approaches (Howard, 1992; Raiffa, 2002). This article first reviews relevant factors in decision-making and then proposes that in a network, some decisions can be considered individually, whereas others should be considered plurally. An illustrative case study for electric power networks shows when the manager should consider the interconnected network in order to make the best decision and when the manager can safely eliminate this step. The final section summarizes and concludes.

BACKGROUND

Decision-making can be considered to fall within two broad categories: individual decision-making and plural decision-making (Raiffa, 2002). Below this dichotomy are three major perspectives. The first is descriptive decision-making, which seeks to determine how people make decisions and perhaps to build models that mimic and predict human behavior. This behavioral-based study is grounded in psychology and seeks to explain how and why people think, learn, and behave when making decisions. The second perspective is normative decision-making, which seeks to determine how decisions should be made and to develop models to help humans make rational and internally consistent choices (von Neumann and Morgenstern, 1947). Economic decision-making and game theory are normative; each relies on all decisions being made rationally within the context of available information. The third perspective is prescriptive decision-making, which is guided by the question of how decisions could be made better (Raiffa, 2002). In a sense, this is a combination of the descriptive and normative perspective; it recognizes the heuristics and cognitive biases present in the descriptive view and attempts to improve upon unaided decision-making.

Plural decision-making, or decision-making involving two or more individuals, can be divided into two orientations: separate and joint. The first involves decisions made by individuals separately but whose ultimate results are influenced by the actions of others in an influencing network (Monge and Contractor, 2003; Wasserman and Faust, 1994). The term influencing network refers to nodes (decision-makers) that are connected via linkages that represent an interaction that determines the results of the actions taken at each node. The decision-makers at each node of the network interact with each other indirectly via the influencing network. Game theory can be used in this realm to help decision-makers determine which actions to take.

[FIGURE 1 OMITTED]

The second plural decision-making orientation is when individual decisions are made in direct consultation with other individuals in the influencing network (Raiffa, 2002). Often this occurs when two or more individuals will be affected by the actions of any single party, so they voluntarily confer to establish common ground to guide their individual actions. Negotiation theory is used in this realm to allow individual decision-makers to collaborate in taking actions that will be agreeable to all parties. Figure 1 shows an organizational structure of the hierarchical relationships described above.

Networked decision-makers must address the added complexity that not only do their decisions impact others in the network, but the reactions of those other decision-makers can in turn impact the outcome to the original decision-maker. If we return to the butterfly effect metaphor that illustrates the dependence of the state of chaotic behavior, we may conclude that in any system of independent nodes (i.e., managers) with some mechanism that can link these nodes (e.g., common product, common supplier, or common customer), small changes (i.e., decisions) in one part of the system will affect all of the nodes of the system. If this premise is true, it would imply that there is no such thing as an individual decision as indicated in Figure 1. Rather, all decisions are plural since any choice made by an individual that exists in the connected world may affect one or more other individuals, whose reaction would in turn affect the original decisionmaker.

Ideally, all decisions should then be made with full consideration of all possible impacts throughout the network and the resulting reactions and reflected impact back to the original decision-maker. This is not realistically practical, so a systems analytic methodology is needed to determine what level and extent of decision network analysis is justified. In some cases, the variability or noise in the network overshadows the impact of a "small" decision made at a particular node. This can be treated with individual decision analysis. In other cases, network noise does not dampen the impacts of a relatively "large" decision, whose impacts and reactions can be transmitted across the network. Such decisions should be treated with plural decision analysis. The question becomes: at what point does the analysis need to change from individual to plural decision-making? We believe that some notion of decision size will determine when it is best to shift from individual to plural analysis. The notion of size of a decision has not been well defined. We will next introduce an illustrative case study to help define one dimension of decision size.

ELECTRIC POWER NETWORK ANALYSIS

This section presents an electric power network case study. Oldham et al. (1986) considered the issue of determining the appropriate electric power generating capacity. Both monetary and non-monetary impacts are considered. Survey results were employed to assess health effects as well as land, water, and property impacts. Shaalan et al. (1994) employed interval analysis of electric power system modification. Engineering performance and economic performance are considered simultaneously, and the interval model facilitates consideration of parameter uncertainty and sensitivity analysis. The analysis presented here expands on this work by adding consideration of environmental impacts, dealing with a deregulated but networked industry, and addressing lifecycle design changes at the plant level directly. The system analyzed here possesses several desirable features for considering individual vs. plural decision-making. First, the nature of the product is such that its differentiability can be carefully controlled in the number and intensity of attributes. This allows a market rich enough to reflect variability between producers but also limits the number of attribute dimensions in order to make solutions tractable. Second, the market is defined in such a way that there is a network of relationships between each producer along several dimensions. Among these dimensions are common supply requirements, geographic proximity, and infrastructure interconnections. If only one network dimension were used, valuable insight could have been lost by not fully understanding what grouping of producers' characteristics is important in controlling the output measure of market share. Third, a single decision-maker determines a performance measure that reflects customers' desires.

The product used here is electric power. This product has few customer-discernable distinguishing characteristics and thus provides fewer alternate causal explanations for the obtained results. Electricity's physical attributes--electrical potential (voltage) and alternating current frequency (hertz)--both are regulated by the interconnections with other electrical generating equipment in the network and thus are not a customer-discernable and -selectable attribute. However, reliability is an important physical attribute of the product.

The electric power industry began functioning under government regulation in the early 1900s (Casazza and Delea, 2003). Due to many factors, the industry began the path to deregulation in the late 1970s, and the current structure is a marketplace that is best represented as an auction (Rau, 2003). The electrical generators offer to supply quantities of power for a given price, and the customers bid to buy power to supply their required loads. Between these two is the dispatcher who can buy power from the generators and then resell it to the customers. The sound business rule of "buy low and sell high" has been the mode of operation for the dispatchers (also referred to as independent system operators or ISOs). These dispatchers control the transmission system and thus decide who can provide power to the transmission system, which will eventually satisfy customer loads. The dispatcher will then determine how much power will be accepted by the system from each of the interconnected electrical power generators and thus determines the price that the customer will pay.

We consider three attributes that the dispatcher can consider in making generator loading selection decisions: how much the customer will pay for electrical power, the reliability of that power, and the environmental impact of electric power generation. The three attributes of economic cost, reliability, and environmental impact of generating the electrical power are often of a conflicting nature. For example, increasing reliability by adding redundant components or using oversized components with a higher failure threshold will often increase cost. Similarly, decreasing environmental impacts by adding air pollution control systems and/or substituting cleaner burning fuels can also increase costs. Thus, actions taken to improve one attribute can worsen another.

Print Get the Mag Weekly Updates