In the InvestAgent project, the agent-based electricity market model AMIRIS was extended to include the investment perspective. To this end, a new module was implemented to represent investment decisions in the electricity sector. The analyses made possible by this allow consideration of an almost unlimited number of investors, which represent companies investing in the electricity sector. Investments by private individuals, such as in rooftop PV systems or home storage, are not yet explicitly modelled. Instead, the installed capacity for these technologies is specified externally in the simulations. Investors in the model can make investments in wind, solar or battery storage. Decommissioning decisions are also included. The project involved the development of test cases to verify that the model mechanics function as desired, as well as a case study of a long-term transformation pathway for the German electricity system. The involvement of stakeholders in the project and a brief outlook after project completion are also discussed below. The slides from the final workshop and a link to the project website of project participant ZIRIUS at the University of Stuttgart can be found under More information.
Method
Forecast
In the model, investors decide based on a long-term price and utilisation forecast for the respective investment option. The forecast can vary between investors or several investors can work with a shared forecast. The forecast covers a period of 20 years, which is almost the entire lifetime of the technologies. A long-term dispatch simulation is used to create the forecast. The development of the power plant fleet and fuel prices, which investors base their decisions on, is largely based on existing scenario studies, such as the Ariadne scenarios. In addition, “miniature plants” for the possible investment options, namely wind, solar and battery storage, are introduced into the simulation. These are small units of the respective technology with a capacity of 1 MW, which therefore have no impact on the price result. This assumes that investors are price takers and their investments have no impact on prices.
Investment Decision
The expected costs and revenues of the investment options are extracted from the previously generated price and utilisation forecast. These are then used to calculate the expected net present value (NPV) for an investor. NPV is a common measure of profitability for companies considered in the project and is used by most investors, sometimes together with other profitability indicators, as determined by the survey conducted in the project. Investment only takes place if the NPV is greater than zero, i.e. if the investment achieves the desired rate of return over its lifetime. In addition, a stricter minimum requirement for the internal rate of return can also be specified. Investors then initially select the investment option with the maximum NPV. They invest as much as possible until either a previously defined maximum share for the technology in an investor’s portfolio is reached or the investor’s investment budget for the respective investment round is exhausted. Once the maximum share of a technology is reached, investment is made in the technology with the second highest NPV, provided it is positive. This continues until the investor’s investment budget for the round is completely exhausted.
The profitability of investments can be influenced by policy instruments. In the model, renewable energies can either receive no funding or be funded through a capacity premium, i.e. a payment in Euros per installed MW, or a market premium in Euros per MWh produced. If support instruments are specified, this is incorporated into the profitability forecast of investors and leads to higher NPVs. Investors in the model decide on their investments simultaneously. They are only linked by the “realisation perspective”, i.e. the simulated electricity market, as described in the following section.
Simulation Process
Investors can make their investment decisions in annual investment rounds. The newly installed capacities from all investors are combined and passed to a dispatch simulation. This dispatch simulation corresponds to the electricity market assumed to be “realised”. The electricity market is simulated for one year with an hourly resolution. Afterwards, a reassessment of past decisions is possible and further capacities can be invested in a new investment round. This is repeated until the last year of the simulation is reached.
After each simulated year, the revenues and costs achieved are returned to the investors. The realised revenues and costs may differ from the forecast revenues and costs if the investor based their investment decision on a forecast that deviates from the realisation. Thus, incorrect decisions can be explicitly represented. Investors can use the realised costs and revenues to decide whether to continue operating or decommission their plants. This is described in the following section.
Decommissioning Decisions
A plant can be decommissioned if its revenues are below the incurred costs. To this end, each investor defines a look-back window and analyses, for example, the revenues of the previous five years of the simulation. If the plant has been at a loss for a minimum number of years, e.g. four years, within this period, the investor decides to decommission the plant. Both the look-back window and the maximum number of years, for which a deficit is accepted, are freely selectable.
Test Cases and Case Study
Test Cases
Several test cases were developed to verify the desired functions of the individual model mechanics. It was checked, for example, whether investments only take place if the forecast NPV is greater than zero. It was also checked whether changed forecasts lead to a change in profitability assumptions and, if applicable, to different investment decisions. The logic for decommissioning decisions and the interactions between capacity expansion and electricity price levels were also subjected to testing. A preliminary parameterisation was used for the test cases. In order to isolate the tested model logics as much as possible, strongly simplified considerations were sometimes made. It was not aimed for making statements about the future development of the electricity system already. Some of the test cases were presented at a conference. Further test cases are described in a publication currently in progress at the end of the project. Figure 1 shows an exemplary comparison of the expansion developments for the base case and a case in which the investor assumes a higher expansion of PV.
A case study simulated a long-term transformation pathway for the German electricity system. The “Elek” scenario from the Ariadne Kopernikus project serves as the basis for the forecast of the investors. The capacity mix in 2026 serves as the starting point for the simulation of investments. The investors in AMIRIS can make their own investment decisions for the years of the transformation pathway between 2026 and 2045, which may differ from those in the Ariadne scenario, for example if the NPV of individual technologies is not positive. The investment volumes generated in the Ariadne scenario were used as the basis for the parameterisation of the investment budgets. Assumptions about investment expenditure as well as variable and fixed costs of the technologies were taken from the dataset of the Danish Energy Agency. It was analysed which capacity expansion occurs. Figure 2 shows the overall capacity development in the scenario considered. Uncertain parameters were also varied and their sensitivity to forecast NPVs and the resulting investments was examined. Figure 3 shows an exemplary illustration of the results of a sensitivity analysis for NPVs when varying the gas price.
Figure 2: Development of the total capacity in the scenario considered.
The developments for all technologies except gas turbines, wind, PV and batteries are taken from the Ariadne scenario “Elek” and are specified exogenously. The capacity development for gas turbines, wind, PV and batteries was determined endogenously. A significant increase in total capacity is noticeable. No expansion is visible for gas turbines throughout, as they have a negative NPV.
Figure 3: Development of the net present values with a variation of the gas price by +/- 25 %.
The effects of the changed gas price on the net present value of gas turbines and PV are low. For gas turbines, this is because they are the price-setting technology in the model. The reason for PV systems is that they hardly produce in hours when gas is price-setting. Higher NPVs for wind and batteries can be determined at a higher gas price. Wind produces in some hours where gas is price-setting. Batteries benefit from increased spreads due to higher gas prices. The NPVs of wind and PV represent a hypothetical situation without funding.
One goal of the project was to align the logics of investment decisions as closely as possible with the real investment behaviour of companies. To this end, a survey on investment decisions in the electricity sector was conducted in the project in cooperation with the project participant ZIRIUS at the University of Stuttgart. The key findings of the survey can be found here. In addition, in-depth interviews with investors and banks were conducted under the leadership of ZIRIUS. Essential elements of the model development by DLR were aligned with the findings of the survey and the interviews.
Furthermore, the project was accompanied by an advisory board consisting of members of associations and companies in the energy industry across almost all stages of the value chain. This practical advisory board provided valuable impetus for the research work at practice advisory board meetings. A future workshop "Actors in the German Electricity System" was also conducted in the project. A summary can be found on the ZIRIUS project website. The project results were presented shortly after project completion in a public web workshop. The slides from the final workshop can be found here.
Outlook after Project Completion
Further exploitation and refinement of the work is planned after project completion. A scientific publication on the method of investment modelling with test cases for the model mechanics is already well advanced. A further publication on the case study for the long-term transformation pathway and the influence of funding instruments on investments will be further processed after the project ends. The model is published and available open source on GitLab. The scenario data will also be made available after the project ends. A publication of the data on the Open Energy Platform is planned for this purpose.
Furthermore, it is planned to refine individual elements of the model after the project ends. For example, an update of the forecast in the simulation process should be implemented and its effects on the development of investment decisions checked. This also allows a reaction to capacity expansions that have occurred in the meantime. Further research is also needed to develop approaches to further objectify and simplify the model parameterisation. Central challenges here consist in determining suitable data and estimators for the investment budgets of the investors and their assumptions for long-term forecasts.
