WEBVTT 00:00.000 --> 00:11.600 One, if you don't hear me just give me some indication that I shall speak louder, right? 00:11.600 --> 00:18.000 So my name is Christoph, I'm from the German Aerospace Center, and despite that name we 00:18.000 --> 00:23.440 are actually doing energy research, and I want to present you our electricity market model 00:23.440 --> 00:30.320 Amiris, and Amiris stands for the open agent-based market model for the investigation 00:30.320 --> 00:37.200 of renewable and integrated energy systems, so quite long title, but otherwise we wouldn't 00:37.200 --> 00:41.040 fill in all of the letters that we had before. 00:41.040 --> 00:49.280 So, right, and it's an agent-based market model for the power market, and it models 00:49.280 --> 00:56.160 business oriented decisions for dispatched considering the regulatory framework, and we focus 00:56.160 --> 01:02.240 on renewable energy sources and flexibility options, and as I'm here, this model is completely 01:02.240 --> 01:07.040 open source, you can find it on our GitLab repository. 01:07.040 --> 01:12.160 The key functionality of this model is that we simulate the trading and operation of power 01:12.160 --> 01:15.200 generation plans and flexibility options. 01:15.200 --> 01:21.040 We model from a business oriented perspective, so we model the behavior under uncertainty 01:21.040 --> 01:28.000 of the individual agents with an hourly resolution, or even higher, if you want to, and spatial 01:28.000 --> 01:34.000 resolution of market zones, so for example, your national energy market, but you can 01:34.000 --> 01:38.160 define that at any range that you want to. 01:38.160 --> 01:44.240 And inputs to that model are the power plant capacities, efficiencies, and availability 01:44.240 --> 01:50.400 of these power plants, feeding potentials for renewables, demand, time series, fuel prices, 01:50.400 --> 01:58.480 and CO2 prices, you feed that into amiris, and what you get is electricity prices, CO2 emissions, 01:58.480 --> 02:04.960 system costs for support instruments, for example, the dispatch of the power plans, and 02:04.960 --> 02:09.760 you can derive market values for all of the units in the simulation. 02:10.640 --> 02:16.960 Now, example application of this model is the question to model market dynamics, 02:16.960 --> 02:21.920 so how will electricity prices evolve in different futures scenarios that you can think of? 02:23.280 --> 02:30.240 How support instruments impact the behavior of the agents, if there is enough support for 02:30.240 --> 02:38.880 renewable energy sources, maybe, or even too much, or how to household electricity tariffs, 02:38.880 --> 02:44.560 impact the flexibility level of these households, especially from heat pumps, electric vehicles, 02:44.560 --> 02:50.400 or storage units in that household, for example. So these are example applications that we 02:50.400 --> 02:56.800 applied a year before, we have several other applications as well. Now, as an agent-based model, 02:56.800 --> 03:02.320 the central thing of our model are the agents, right? And we have different types of agent in our 03:02.320 --> 03:09.360 model. Of course, we consider markets, at markets we determine prices for goods. For example, 03:09.360 --> 03:14.400 at the day ahead market, we can determine the electricity price. But we have also other markets 03:14.400 --> 03:20.640 like the fuel market or the CO2 market, which are things are kind of stopped, which will only 03:20.640 --> 03:27.920 feed in the prices that you pre-decided on for the simulation. We have traders that operate on 03:28.000 --> 03:33.760 these markets, they fulfill marketing strategies, so they bid for goods, or they offer goods, 03:33.760 --> 03:42.240 and so we have like supply traders or demand traders. And those will create bits and place them 03:42.240 --> 03:50.480 at the electricity market. Then, once prices are awarded and the bits are awarded, they will feed 03:50.480 --> 03:55.600 back to the power plant operators and they will control the power plants and operate them. 03:56.160 --> 04:04.160 Flexibility providers allow two model electrolysis units, for example, or storage units, 04:04.160 --> 04:10.880 or heat pump units, for example, and they try to minimize the cost, maximize the profit from the 04:10.880 --> 04:17.600 operation, and they need forecasts that we have in our model, and these forecasts are provided 04:17.600 --> 04:26.480 by dedicated forecasting agents. Also, we have explicit consideration of support policies, and for 04:26.480 --> 04:33.920 that we have a support policy agent. Now, you can see how all of these agents interact with each 04:33.920 --> 04:38.480 other. We have different kind of connections between those agents. We have flows of money, 04:38.480 --> 04:44.560 flows of energy, and also flows of information between these different agents, and these then 04:44.560 --> 04:51.360 form the whole simulation. And as a result of these simulations, you can see here, 04:52.320 --> 04:59.360 back testing results, all of the data is open source, open access, so you can download these data 04:59.360 --> 05:04.640 from our example repository, and run the simulation. As you can see here, is electricity time 05:05.520 --> 05:13.200 series over the last over a few weeks in the year 2019 for the German country, and you can see 05:13.200 --> 05:19.360 we achieve quite a good fit for the historical data in gray where our model results are blue. 05:21.920 --> 05:28.720 Now, how do we actually use it? So, I'm here with this kind of simulation written in Java using 05:28.720 --> 05:33.920 the open framework for distributed agent-based modeling of energy systems, also long name, 05:33.920 --> 05:40.400 and it's also developed by us. That's the cost probably, and that's famed for short, in that 05:40.400 --> 05:46.720 case, and here we define a mirror, which then communicates with the input output model written in 05:46.720 --> 05:55.120 Python, which collects your configuration files as yammer standards, puts all your CSV data in there, 05:56.400 --> 06:02.480 makes a binary input file, runs the Java simulation in on one or multiple cores, usually 06:02.480 --> 06:10.160 only use one my laptop, and then you create binary output files, which then are translated into 06:10.480 --> 06:18.480 readable CSV files again using famio. This also completely opensource, so feel free to build your 06:18.480 --> 06:26.320 own models with that framework as well. And for an easy access, we have a Python wrapper for this 06:26.320 --> 06:33.520 ameraspy that can make life way easier. So, if you want to use ameras, it's basically creating 06:33.600 --> 06:40.400 a Python environment running pip install, ameraspy, ameras install will download the whole 06:40.400 --> 06:47.680 model and all of the example files, and then run the model with an amerasrun command and the download 06:47.680 --> 06:52.880 model and the files, and after it maximum five minutes if you have Java and Python already 06:52.880 --> 06:59.200 installed in a system, you will have your first result ready for one of the examples in the 06:59.200 --> 07:07.840 simulation here. If you want to connect your own model, maybe, to our to our mirrors, it's 07:07.840 --> 07:14.400 quite easy. You can we provide a URL model service class within our code that allows easy connection 07:14.400 --> 07:21.920 with your model, basically anything. So, it's an API web API approach, basically, so you can 07:21.920 --> 07:26.880 then web request to your API and interchange data, and if you don't have an API, it's super easy to 07:26.880 --> 07:36.640 set one up. So, you can couple with virtually any model at runtime of the ameras model to your own one, 07:36.640 --> 07:41.280 and we use this a lot if we don't want to code our special code in Java, for example, or we 07:41.280 --> 07:47.760 have a different model in Python or something, then we use this kind of remote call to the 07:47.760 --> 07:54.320 specialized model, and this works quite well. For example, for load shifting optimization or heat 07:54.320 --> 07:59.200 pump operation, we have used this, and we are currently using this also for machine learning 07:59.200 --> 08:07.120 based predictions of the electricity price forecasts. We develop our model as following the 08:07.120 --> 08:12.320 fair for research software principles, so we want to have our model findable, accessible, 08:12.320 --> 08:20.160 interoperable, and reusable. All of these links are under downloadable, can follow all of these links 08:20.240 --> 08:26.800 on the downloads from this presentation, and we also have some external users already, which we are 08:26.800 --> 08:33.520 quite proud of, and if you want to join our community and want to work with us or work with the 08:33.520 --> 08:38.400 model, feel free to do so downloaded use it, and if you have any questions, so any, 08:39.920 --> 08:44.320 not happy with our model and complaints, maybe something like that, feel free to post your 08:44.320 --> 08:51.760 questions or complaints or suggestions at the OpenMot forum, send us an email, or discuss with us 08:51.760 --> 08:57.760 at our open forum that we have every Friday at 10 o'clock, a century European time, and of course, 08:57.760 --> 09:01.520 feel free to raise issues or contribute in our repositories. Thank you. 09:01.760 --> 09:15.040 Thank you. Any questions? Yes. 09:24.320 --> 09:29.040 So in our group, we are doing a lot of research on the integration of renewables, and what you 09:29.040 --> 09:34.720 require support instruments, so we have several research projects where we try to understand 09:34.720 --> 09:39.680 which kind of support instruments that there are, so like feed-in tariffs or market premium 09:39.680 --> 09:47.200 stuff like that, work best for the future, and exactly that, yeah. So, yes? 09:59.040 --> 10:05.760 And then that input is a view of money, which then creates a cut-in electricity, which creates a 10:05.760 --> 10:12.720 mic of use, but in a kind of a cascaded event, because since you have known so close, inflation money, 10:13.280 --> 10:18.400 or some have stopped paying, because of any more money, so then they have electricity, and that 10:18.400 --> 10:26.880 stops the flow of power, so are you interested? So, I try to repeat the question, so you ask if we 10:26.960 --> 10:34.720 kind of model complex spill over effects between the different flows that are presented? 10:38.400 --> 10:48.320 I'd say no. Yeah, so kind of the system that we presented are is kind of static as well, so 10:48.320 --> 10:55.840 the flows in there, you can't change the flows numbers, but not really kind of create new connections 10:55.920 --> 11:00.960 or make a kind of evolving system during this simulation, you would need to do that manually, 11:00.960 --> 11:08.640 probably, right? Thank you.