Plans for 2014

Plans for the 2014 Power TAC competition

The goal for this page is to develop and document a consensus among server and broker developers on features and improvements for the 2014 competition. Please use the "Discussion" page attached to this page to add your comments; this page will be periodically updated as we approach a consensus. In order to participate, you will likely need to create an account on this wiki if you don't already have one.

The timeline for the 2014 schedule starts in August 2013. By the end of September 2013 we should have agreement on features for the game and for any changes in the simulation server. The full Specification should be ready by mid-December, and the simulation server for 2014 should be feature-complete before mid-January 2014.

Tournament schedule

When the venue and date for the 2014 competition are available, they will be published here.

In addition to the annual tournament, we plan to run approximately four trial tournaments, to test new features and provide broker developers with opportunities to refine and stabilize their systems. We may decide to use "themes" for at least two of these trials; for example, we could run one with 30% penetration of electric vehicles or solar PV in the customer population, and we could run one or two using weather data from fairly extreme locations (winter in Minneapolis, summer in Houston).

Game design updates

A number of suggestions have been made to enhance the Power TAC scenario in various ways for the 2014 competition. We need to decide which ones to include, keeping in mind the need for a careful balance between realism and ease of analysis.

• New genco models representing wind parks and solar PV parks. These are well along in development as of July 2013.
• Some variations in genco and customer models between tournaments, or possibly even between games within a tournament. For example, population sizes could vary, along with penetration of distributed wind and solar resources. Note that varying such parameters between games in a tournament could seriously complicate data analysis.
• An Electric Vehicle model as a new customer type.
• Enhancement to the Tariff model to allow payment to customers for balancing capacity rather than simply production or consumption used for balancing. The current scheme only allows for discounts on price, and does not allow for quantifying balancing capacity.
• Simulation of the effects of grid congestion through location-marginal prices in the wholesale market. The lack of a transmission model is a major criticism from power-systems people. The question is whether the impact of grid congestion can be simulated with appropriate distributions, or whether some sort of actual grid model must be added.
• Long-term contracts for delivery of energy through the wholesale market, in addition to the existing day-ahead market.
• Specific market rules about the treatment of distributed renewables. Examples include the feed-in tariffs used in Germany and Spain, and the net-metering rules in many states in the U.S. Keep in mind that implementing such rules will require tying a consumption tariff to a production tariff; currently they are separate.

These are just headlines, of course. Before any of these features can be added, we will need a detailed specification of how they are to work.

Simulation server

Beyond implementing the new models required by game design changes, the primary work in the simulator is cleanup and documentation, and some improvements in the Visualizer.

Broker framework

The existing broker framework needs to be better documented in some areas, and it needs to do a better job of detecting timing anomalies, such as when a context-switch on the broker's machine delays a time-update message.

A new broker framework is needed (or perhaps more than one) to make it possible to write a broker using tools that are more familiar to economists, business scholars, and engineers. Suggestions include Matlab, R, Python, Repast Simphony, and Clojure. There are two possible approaches to this problem:

1. Do some "market research" to decide which platforms would attract the biggest and highest-quality audience, and implement that. 2. Find a volunteer to do a port and help them do a good job, then see if anyone picks it up.

Given the availability of development resources, the second approach is more likely to succeed.

Tools

Power TAC is a research tool, and it's ultimately only as good as the quality of research it produces. The simulation is complex and produces large log files. We will all be more successful if we collaborate on a set of analysis tools that make our research more productive.

Experiment manager

Some work has started on a version of the Tournament Scheduler that will serve as a platform for designing and running multi-game experiments. This is an essential element for empirical research using multiple brokers and multiple configurations.

Log analysis tools

The logfile analyzer works for extracting data from individual game logs, but it's also important to be able to extract data from large numbers of games and analyze them together in various ways. Some teams have ad-hoc sql-based tools for this. It would be a great service to the community to share ideas and code and come up with a flexible and easy-to-use toolset for this. Even better if it could be integrated into the experiment manager.