A brief introduction to our agent-based model of the el-certificate market – The NEMM

The following introduction provides an overview to NEMM, our “New Elcertificate Market Model”.  The main goal of the NEMM project is to develop a model of the certificate market that takes account of the behavior, activities, and information available to market participants. The project has two key components: interviews and model development. A number of pilot interviews with market participants have been conducted throughout spring and summer 2014 in order to better understand actors’ behavior, decision making processes, drivers and approaches. The results of these interviews were then used as the basis for the development of an agent-based simulation model of the Swedish-Norwegian elcertificate market. NEMM models the individual decision makers in the market, their behaviour and interactions, instead of the taking the traditional “top down” equilibrium approach of traditional economics. NEMM is quite unique – we are not aware of any other agent-based approaches to quota markets like the el-cert market, and it provides a unique insight into the future development of the elcertificate market.

The project is partly financed by the Research Council of Norway and has been undertaken by Optimeering together with Thema Consulting Group. In addition, several project partners from the power sector in Norway and Sweden sponsor the project.

What follows is a brief overview of NEMM – how it is put together, the types of agents modelled, and how they interact. If you are interested in finding out more, please contact us – we’d love to show you some results. We are also presenting at the upcoming Montel el-certificate conference in January – based, of course, on the model and results.

Model overview

The building blocks

The two key elements of NEMM are the Environment and the modelled Agents.

Lysbilde02

The Environment

The Environment is the modelled representation of the real world context of the joint Swedish-Norwegian electricity certificate market. All the information and elements in the model that the simulations itself does not change exists in the Environment. For instance all potential projects, either under construction, awaiting investments decisions or in other parts of the early stages, are represented here.

The Environment also includes information such as all current, historic and future power demand, certificate quota (i.e. certificate demand), power prices (spot and future), all power plant and project data, market regulations, tax schemes and the actual weather conditions given at specified times in the simulation.

Input data

The data needed to populate the Environment is provided via the model’s database. This database includes, among other things, detailed and up to date information regarding all relevant power plants and projects, including capacity, load factor, capital expenditure, operations expenditures, rates of technology improvement, estimated construction time, etc. Certificate demand is provided in terms of the quota curve, expected annual power demand and realistic within-year demand profiles.

Agents

The Agents represent the heterogeneous actors in the market. Agents are of different “types” (e.g. producers and obligated suppliers or purchasers), reflecting the roles they play in the certificate market. In addition, Agents of the same type can be different in terms of utility functions (how they determine what are successful actions, and what are not), learning, horizon, decision-making, size, region, cost of capital, etc. The number of Agents in a specific simulation can vary simulation to simulation, and can change (new agents can be added or existing agents can go bankrupt) during the simulation. The different types of Agents are described below.

Developer agent

Developer agents undertake all project development activities and construction of new power plants. Any investment decision or decision to apply for concession is made by this agent, but with respect to the company policy and required rate of returns given by the parent Company. The investment decisions are evaluated as per the developer agent’s strategy, and may take account of market analysis (provided by an analysis agent), the current and future market price (certs and power), the (estimated) relative quality of the project compared to competing projects, or a combination of these.

Producer agent

Once a power plant begins operation, it is transferred to a producer agent, who is then responsible for the sale of the resulting certificates in the certificate market. The Producer agent offers its certificates into the market based on its particular market strategy, its current position, how well its strategy has performed so far, and market insight provided by an analysis agent.

Purchaser agent

Represents a retailer obligated to buy certificates certificate eligible power demand. In the same manner as the Developer agent and the Producer agent, it depends on an analysis agent for its understanding of the certificate market. Purchaser agents who do not acquire sufficient certificates face a penalty.

Trader agent

Trader agents have no underlying demand or production of certificates and partake in the market to exploit arbitrage opportunities.

Analysis agent

Conduct different kinds of market analysis of the certificate and power market, and shares this with the agents within the same Company (see below). The Analysis agent provides both short-term and long-term market analysis to their associated agents. The Analysis agent also estimates the future demand and supply of certificates within the Company.

Company agent

This agent is an entity providing the linkage between different types of agents belonging to the same organisation. For instance, an Analysis agent provides its analysis only with the Developer agent or the Producer agent in the same Company agent. The Company agent does not partake in any processes in the simulations directly, but provides its “child” agents (one or more of a developer, producer, purchaser, and trader) with information such as company policy, cost of capital, and other guidelines. All agents in the model will have a Company agent as a “parent” agent, and a Company agent will have at least one “child” agent associated with it.

Model dynamics

The NEMM model can be considered to consist of three main and distinct processes, and the detailed interaction between these processes. The figure below illustrates the three processes: Market, Analysis, and Investments.

NEMM Components

The market process

This represents the day-to-day trading in the model. It is in this process that Purchaser agents, Trader agents and Producer agents sell and buy certificates and the current certificate price is set. In short, the agents participating in the market post bids of offers according to their particular strategy, utility (how they determine what is a good strategy and what is not), their position and previous market outrun and liquidity. The agents develop (change) their strategy according to how successful or otherwise said strategy has been in the market.

 Analysis

The analysis component is where the analysis agents build up a picture of the future direction of the modeled certificate market. There are three types of analysis conducted at present: prognosis of short term market price developments, prognosis of medium to longer term market price directions (called the “certificate value signal”, or CV signal) and a project cost-curve analysis. The short term market analysis is a prognosis of the price in the next trading period, and is based on a statistical model of the market. The CV signal estimates value based on the size of the overall market position (i.e. how short or long the market is). The cost-curve analysis captures a “fundamental” approach to the market and essentially calculates the cost of the marginal plant required to meet the certificate demand.

The certificate value signal is used by all actors in evaluating bid/offer alternatives, and reflects the fact that those actors who are able (willingly or unwillingly) to wait before they sell (or buy) certificates will make an assessment of the likely future movements in certificate price. Agents in NEMM make this assessment using a minimum amount of information – the current price, and how short or long the market is. The project cost curve based analysis is used by some (but not all) developer agents in determining investments, and some agents in determining floor and ceiling price levels.

Short term market price expectations

The price expectation for the next trading period (short-term price expectations) provided by the Analysis agent is based on a statistical model of previous market prices. That is, the market participants have an expected certificate price for the next trading period given as a function of the past few prices (though they will not necessarily bid that price). The prognosis “learns” by updating itself to changing market price levels.

The Certificate value signal

The CV signal is the expected direction of future certificate price movement – up, sideways, or down. The concept behind the signal is simple, and is based on the number of total certificate shortfall at the given time point (that is, the demand from “now” to 2035 less the production to 2035 from all existing and under construction power plants). The signal makes the assumption that the current market price reflects the markets evaluation of the chance the market will be long (and thus certificates will ultimately have no value) and the chance the market will be short (and the certificates will ultimately have a very high value). By comparing the total certificate shortfall now with the total certificate shortfall at the end of the holding horizon (assuming in both cases only production from existing and under construction plant), other agents associated with the Analysis agent can assess whether the market is expected to become tighter or looser, and thus if prices should go up or down.

Project cost curve based analysis

This essentially takes a “fundamentals” point of view on the future price of certificates. It evaluates all future demand and all future certificate production from the plants already in operation or under construction. It then assumes that all certificate demand will be met in the most cost-effective manner based on the long run marginal cost of all potential projects and the time of build out. It hence indirectly assumes that the certificates market will ensure an optimal build-out to meet demand and that the price of certificates will be the certificate price needed by the marginal project build in order to break even.

As all Analysis agents are uncertain in their estimates of future demand, production and economics of potential projects, both the Certificate value signal and the cost-curve based analysis will differ among the Analysis agents. Note also that different agents will use their analysis differently, and some will not use it at all – reflecting how actors in the market actually operate.

The Investment process

The investment process is the collective term describing all the decisions that each developer agent must take for all projects within every investment period. The decisions to be made by each developer are:

  • Should we spend resources on identifying new projects?
  • Should we apply for concession for a particular project?
  • Should we start the preconstruction for an investment ready project?

These decision steps are a simplification of the decision steps taken in actual investment cases and described to us in the pilot interviews. However, they provide a reasonable approximation to the approach taken by all those interviewed that undertake development activities.

Investment strategies

NEMM currently implements three investment strategies (with each developer agent using one of these). The strategies are based on the results of the pilot interviews, and capture the investment decision-making approaches described to us by the interviewed actors. The three strategies are:

The cost-curve strategy. A developer agent using this strategy will move a project from one step to the next based on its LRMC compared to the LRMC of the marginal plant assessed by the developer company’s analysis agent (see the section on project cost curve based analysis above).

The market price-based strategy. A developer using this strategy will invest in a project only if current (i.e. modeled at time of investment) certificate and power prices are sufficient. Such developers will in addition not invest in a project if it is sufficiently more expensive than an assessed marginal plant (even if prices are high enough).

Finally, the third strategy is the restricted price-based strategy. Similar to the price based strategy, a developer will invest in a project only if the value of the “hedge-able” (or “PPA-able”) power and certificate income is sufficient to cover investment costs. This captures those investments that require bank financing under current market conditions.

Model results

The results of NEMM are currently only available for our project partners. If you are interested in finding out more, please do not hesitate to contact us.