Agent based simulation of artificial electricity markets

We have developed new approaches to the modelling of the customer reaction. The basic to problem here is to model how customers respond to different price patterns, also those that have not been used in the experiments, and then to find the price pattern that produces the desired changes in the utility's load pattern. Our idea is to model the consumers and producers as optimizing goal seeking agents and then simulate the electricity distribution system as an artificial market. This approach has connections to current research distributed artificial intelligence. We have developed a prototype software called Power Agents P[HÄM95d], P[HÄM96c].

(Full picture 42 KB)

Thus far, the consumer agents represent residential customers. The customer response is assumed to be based on the maximization of the subjective valuation of services provided by the electricity used minus the price paid for it. Currently we are working with an agent model that describes the behavior of customers with electric space heating T[PAR94], T[HÄM96]. P[HÄM00b]. Different ways to use multicriteria optimization procedures are evaluated and the whole coordination problem is analyzed as a Stackelberg game P[HÄM97b].

We are extending our research on artificial markets the analysis of electricity production markets. There are new electricity exchange market places in the Nordic countries.

(Full picture 50 KB)

Short-term load forecasting forms an important part of an energy management system. Accurate forecasts are a prerequisite for efficient production and purchase planning as well as for security assessment of the power system. We have developed a system for short-term load forecasting in communal utilities R[RUU88], T[KAR90]. The computer implementation of the system, called FLOATER, is now owned by IVO International, a large Finnish energy consulting company, and it is in operational use in three Finnish utilities.

Related to the work on dynamic pricing of electricity, we have developed approaches for modeling the load of different types of customers. Because of large measurement errors, robustness of the model is emphasized. Moreover, because of the very large number of data, computational considerations are extremely important. We have chosen the type of model where the load is modelled as the sum of the dominating rhythm component (daily or weekly rhythm) and the weather sensitive part of the load P[RÄS95a]. An example of the decomposition of the load into the rhythm component and the weather sensitive component is shown in the figure below. A computer implementation of the system, called LoadLab, has also been developed P[RÄS96]. The program has been used in the analysis of dynamic pricing experiments and in load analysis projects.

(Full picture 9 KB)

(Full picture 50 KB)

The electric utility industry has organized power pools to coordinate the generation and transmission of electric power. When there is an opportunity for cost savings, the members are confronted with the problem of allocating the savings in an equitable manner. We have developed contracts for multi--period electricy exchange considering situations where side payments between the companies are ruled out (P[EHT87, EHT88a, EHT89c, RUU89b, RUU91a, RUU92, RUU94a]). The exchange is accomplished through barter where the commodities are deliveries of electricity in the different periods of the cycle.

The axiomatic approach to bargaining can be used to define the equity properties of the cost allocation mechanism. Because the conditions for the compromise solution are, in general, nonseparable with respect to time the problem becomes computationally complex. We have developed a hierarchical computational method to solve these problems. The approach has turned out to be efficient in applications to real power systems.

Natural resource problems

The use of control and game theory methods in international resource management problems is of growing importance. We have applied dynamic game theory to modelling and analysis of problems where two or more agents utilize the same or interdependent resources. Thus, the agents face dynamic decision making problems related to the choice between cooperative and noncooperative or myopic and foresighted management policies. We have analysed finite and infinite time horizon fishery games where agreements among the agents are not binding. Noncooperative solutions of games (P[KAI85a, KAI85b, DOC89, HÄM89a, KAI89a]) have been analysed. The properties of cooperative and noncooperative games have been combined in order to model negotiations where the purpose is to reach safe equilibrium agreement (P[HÄM84a, HÄM85a, KAI86, KAI87a, KAI88b, EHT88b, EHT93a, EHT95a]). The United Nations' Conference on Straddling Stocks and Highly Migratory Fish Stocks has been analysed using game theoretic approaches P[KAI95c], P[KAI95e], P[KAI98a], P[BJO00], R[LIN98].

The theory of Stackelberg (leader-follower, principal-agent) games has been aplied to model dynamic incentive contract. Models of output share incentives have been analyzed in connection with harvest share contracts in fishing under cartels and social coordination (P[HÄM86a, HÄM90c]). The Norwegian spring-spawning herring fishery has been studied using risk analysis P[TOU00], R[TOU98].

Pollution and environmental problems

Air pollution is mainly due to sulphur and nitrogen emisions which are partially deposited on the ground of the emitter countries. Due to the remote transportation of pollutants by winds, large amounts of the emissions fall on the ground in the neighbouring countries. Sulphur and nitrogen compounds affect the properties of soil, surface water, and ground water, for which reason this particular pollution process is referred to as acidification.

We have developed and analysed two- and three-country transboundary air pollution models in a game theory setting. The project analyses noncooperative and cooperative solutions of the acid rain game and evaluates the bilateral and trilateral emission reduction agreements among Finland, Russia, and Estonia P[KAI91e, KAI92c, KAI92d, TAH93, P[KAI95a]. P[KAI98b],

The climate on the earth is determined by the balance between energy received from the sun and energy reradiated back into space. The radioactively important gases, greenhouse gases (e.g. CO2, nitrous oxide, methane), are transparent to incoming visible radiation but they absorb invisible thermal radiation. The thermal balance on the earth is changed resulting in overall climate change. This phenomenon is referred to as the greenhouse effect.

Game theory analyses show that the design of cooperative programs for international environmental agreements may be a challenging and complicated task when we are dealing with stock pollutants and when the players are asymmetric with respect to emission volumes, emission abatement costs, or damages caused by the environmental change. The project analyses the dynamic properties of multilateral environmental agreements. In particular, we study the problem of constructing self-enforcing environmental equilibria in the context of environmental policy problems arising at an international level, in particular in the context of global climate change P[KAI95d].

Methane is an important greenhouse gas and contributes to the anticipated climate change. During the last two centuries, the atmospheric methane concentrations have been increasing significantly. In the biochemical cycle of methane, boreal peatlands act as methane sources. We have initiated a research project for analysing the methane emission dynamics in boreal peatlands P[KET96b].

(Full picture 80 KB)