1 00:00:07,600 --> 00:00:12,100 The purpose of this talk is to discuss an agent-based model of the power sector, called 2 00:00:12,100 --> 00:00:13,300 EMLab-generation. 3 00:00:13,300 --> 00:00:20,000 The underlying question really is: `how are we going to change this huge system to make 4 00:00:20,000 --> 00:00:26,980 the energy transition happen?' You can get a glimpse of how huge it really is... by looking around. 5 00:00:28,660 --> 00:00:32,660 We are now at the Port of Rotterdam, and this area is called the 'Maasvlakte'. 6 00:00:32,660 --> 00:00:39,660 As a Dutchman I'm proud to show you this land; reclaimed from the see in the 1960s, 7 00:00:39,660 --> 00:00:42,920 and this is now a flourishing industrial area. 8 00:00:42,920 --> 00:00:47,800 Also as a researcher it is fascinating: it's a huge source of CO2 emissions, has a few 9 00:00:47,800 --> 00:00:51,140 of the newest coal power plants, making it a large electricity production center, and 10 00:00:51,140 --> 00:00:54,670 it is one of the largest energy consumption areas. 11 00:00:54,670 --> 00:00:56,800 So how can we get the energy transition underway? 12 00:00:56,800 --> 00:01:03,399 In a previous lecture by Igor Nikolic, you already got to know Agent-Based Modeling. 13 00:01:03,399 --> 00:01:08,570 Agent-based models are a computerized laboratory, which help to explore what goes on here at 14 00:01:08,570 --> 00:01:15,470 the Port of Rotterdam area, specifically in the power sector, because they do not ignore, 15 00:01:15,470 --> 00:01:21,070 but rather embrace the complexity of the system! How can we start to understand and shape the 16 00:01:21,070 --> 00:01:24,100 system in the direction we want it to go? 17 00:01:24,100 --> 00:01:25,530 How can we get this system to change? 18 00:01:25,530 --> 00:01:31,840 We are so keen on reducing CO2 emissions - well, the things that need to change are right before 19 00:01:31,840 --> 00:01:38,840 our eyes!! As an agent-based modeler, I immediately start to think: well if you want to lower 20 00:01:39,280 --> 00:01:41,610 our emissions, we need to change someone's behavior. 21 00:01:41,610 --> 00:01:42,299 Whose? 22 00:01:42,299 --> 00:01:43,880 In what way? 23 00:01:43,880 --> 00:01:45,310 To what extent? 24 00:01:45,310 --> 00:01:46,729 How can we achieve that? 25 00:01:46,729 --> 00:01:50,560 So maybe we need some policies. 26 00:01:50,560 --> 00:01:54,610 So... what are the long-term effects of climate policies? 27 00:01:54,610 --> 00:01:56,780 What model can help you to tackle the problem? 28 00:01:56,780 --> 00:02:02,000 At least you need the existing power plants in there and the owners operating them. 29 00:02:02,000 --> 00:02:06,380 You represent the power plants, model the owners as agents that do the operation: 30 00:02:06,380 --> 00:02:13,040 they sell electricity to the market -- representing the consumption -- and choose when to run 31 00:02:13,040 --> 00:02:13,970 which power plants. 32 00:02:13,970 --> 00:02:18,040 You make sure that they are smart enough to run the cheapest plants: at lower demand the 33 00:02:18,040 --> 00:02:22,560 cheaper plants run and at peak demand also the expensive plants run. 34 00:02:22,560 --> 00:02:27,310 Accordingly electricity prices vary. 35 00:02:27,310 --> 00:02:32,959 As of 2005, in Europe we have an emissions trading scheme, which is a market for CO2 credits. 36 00:02:32,959 --> 00:02:38,120 By limiting the available credits, CO2 emissions become costly, and they are reduced. 37 00:02:38,120 --> 00:02:42,160 In order to see the effects of a CO2 price, we added it to the model. 38 00:02:42,160 --> 00:02:44,260 We find that not much changes in outcomes. 39 00:02:44,260 --> 00:02:51,160 The cause: we really need different kinds of power plants here! How do those get there? 40 00:02:51,160 --> 00:02:54,250 Well, companies need to invest in new generators. 41 00:02:54,250 --> 00:02:57,790 So you add the companies' ability to make investments. 42 00:02:57,790 --> 00:03:03,019 An investment decision is very much bound by uncertainty: it is based on expected power 43 00:03:03,019 --> 00:03:10,019 prices, expected CO2 prices, expected fuel prices, technology developments, company profile. 44 00:03:10,510 --> 00:03:15,280 In turn, these are - at least partly - affected by past investments. 45 00:03:15,280 --> 00:03:22,260 So all decisions together determine the system-wide developments and performance. 46 00:03:22,260 --> 00:03:28,390 Now we end up with a model with existing power plants, the owners that operate them, 47 00:03:28,390 --> 00:03:34,800 the market to sell the electricity to, a CO2 market with limited credits to make sure there is 48 00:03:34,800 --> 00:03:41,400 a CO2 price that reduces the emissions, and the owners invest in new capacity as they see fit. 49 00:03:41,500 --> 00:03:46,629 With the model, we now study the influence of policy on these investments in the electricity 50 00:03:46,629 --> 00:03:51,980 market, the CO2 emissions that result from this, and therefore also the demand for CO2 51 00:03:51,980 --> 00:03:57,049 credits, the CO2 price that results from that, and so on. 52 00:03:57,049 --> 00:04:03,860 With agent-based modeling, this model can explore heterogeneity of actors, consequences 53 00:04:03,860 --> 00:04:08,659 of imperfect expectations and investment behavior outside of ideal conditions. 54 00:04:08,659 --> 00:04:15,129 It is also very intuitive: you can actually see the actors act & interact during the simulation 55 00:04:15,129 --> 00:04:18,109 through markets. 56 00:04:18,109 --> 00:04:21,609 The real analysis is done with data of many runs. 57 00:04:21,609 --> 00:04:22,879 So what do we find? 58 00:04:22,879 --> 00:04:27,719 Well, it is not at all straightforward to structurally change this system. 59 00:04:27,719 --> 00:04:34,719 The CO2 market works fine on the very long term, but it is not unlikely that it leads 60 00:04:35,400 --> 00:04:39,469 to a rather expensive decarbonization path. 61 00:04:39,469 --> 00:04:44,059 Making the energy transition cheaper requires an incentive that is stable enough for businesses 62 00:04:44,059 --> 00:04:48,270 to justify significant renewable investments over longer decades. 63 00:04:48,270 --> 00:04:54,379 Our modeling suggests that without additional policy interventions such as a CO2 price floor 64 00:04:54,379 --> 00:04:56,599 this is not to be expected. 65 00:04:56,599 --> 00:05:02,129 As Margot Weijnen already mentioned: worldwide developments in shale gas have a significant 66 00:05:02,129 --> 00:05:06,589 effect on our effort to make our electricity sector renewable. 67 00:05:06,589 --> 00:05:11,809 Can you then say: we should not have built a new coal plant here at the Maasvlakte? 68 00:05:11,809 --> 00:05:17,080 Should we blame the owner for 6 Mton CO2 emissions per year? 69 00:05:17,080 --> 00:05:21,379 Not really, because our modeling also indicates that this plant had a reasonable business 70 00:05:21,379 --> 00:05:24,569 case when the decision was made. 71 00:05:24,569 --> 00:05:29,089 Capturing the most relevant factors underlying investment decisions in the model is key to 72 00:05:29,089 --> 00:05:31,319 embrace the complexity of the power system. 73 00:05:31,319 --> 00:05:35,869 By doing so, we aim to understand better how the system as a whole functions, and have 74 00:05:35,869 --> 00:05:39,839 an impact on where it may be headed. 75 00:05:39,839 --> 00:05:43,809 It does not provide you with perfect prediction, but it should help a lot in the discussion 76 00:05:43,809 --> 00:05:49,389 on how to make our energy transition going! This modeling is part of TU Delft's energy 77 00:05:49,389 --> 00:05:53,800 modeling lab and is open source. Do you want to explore it yourself? 78 00:05:53,800 --> 00:05:56,849 For more information, see http://emlab.tudelft.nl.