#319 Analysis of Market Scenarios using Agent Based Modelling of Complex Electrical Power System
Authors: Yogesh Bichpuriya, Narayanan Rajagopal and Krishna V Prasad
Electrical power system is a complex system having multiple agents interacting with each other to maintain the dynamic balance of demand and supply. Though these interactions were limited in the traditional system. Now, the electrical system is going through a major transformation from its traditional avatar to a future grid. The transformation brings challenges in different aspects of the system e. g., planning, operation and control etc. Increasing proportion of Intermittent distributed renewable energy, evolving market structure and interactions, and active participation of customers/prosumers are some of the major factors which should be considered for decision making.
In such a complex system with multiple agents and enhanced and extended interactions, existing modelling and simulation approaches to understand the behaviour of the system and to make optimal decisions would not be sufficient. In a complex system, understanding of agents and their interactions is also important for policy makers and regulators.
In this work, we present modelling of different agents and their interactions in the electricity market scenario. We use Agent Based Modelling approach to define these agents and to model their interactions. We will also discuss some use-cases to illustrate the impact of an agent’s behaviour on the other market agents and their strategy. The model will be useful for doing impact analysis of any new regulation or market mechanism. It will also help in doing what-if analysis by different agents in order to make a decision.
#322 Hierarchical Structure of World Trade Flow Network
Authors: Le Ahn Quang, Ashadun Nobi, Nam Jung, Tae Ho Lee, Seung Eun Maeng and Jae Woo Lee
We analyze the structural change of the trade flow network for different commodities from 1995 to 2013. The MST (minimum spanning tree) is constructed from data of the trade flow and investigates the hierarchical organization of the MST by the method of hierarchical path. We find that the tree for man-made products such as machinery and transport equipment and manufactured goods is more hierarchical than the natural products such as mineral fuels. We also observe that the hierarchy F is changed over time notably due to globalization and crises. The interesting finding is that when we remove China from trade flow network, the trend of the F is changed significantly which implies the role of China on world trade network in recent years. Finally, the network stability is identified by Jaccard index. We observe that the MSTs for the man-made products are more stable than those of the natural products.
#323 Chloroplasts and mitochondria exhibit wonderful synchrony in evolution with host genomes
Authors: Michael Sadovsky, Julia Putinzeva, Anna Chernyshova and Vaselina Fedotova
Very strong correlation between genome-wide determined structure and phylogeny of the genome bearer is found. Triplet frequency dictionary is the structure, and organelles genomes are the entities used to figure out the structural proximity. Clusterizing in the triplet frequencies space has been carried out, both by K-means, and elastic maps technique. The set of genomes was clusterized into 2, 3, ..., 8 classes, consequently, and the composition of the classes was studied. The composition of the classes was found to be very far from a random one: usually, two classes developed for K-means clusterization with K classes merged into a single class, as the clusterization has been developed for (K-1) classes. Besides, the observed phylogenetic composition of the classes strongly follows the phylogeny. Indeed, a clusterization for two classes over the mitochondrion genomes always yields a perfect separation of chordata species from invertibrate ones. Similarly, such clusterization carried out over the chloroplast genomes yields a perfect separation of higher Embryophytes from non-vascular plant. A growth of the number of classes identified through K-means always followed the classic phylogeny of organisms, in general. It should be stressed that the proximity in structure space (that is triplet frequency distribution determined genome-wide, with neither elimination or prevalence of some specific regions of a genome) was determined over the organellae genomes, while the phylogenetic proximity has been determined in traditional way, through morphology (that is determined by nuclear genome, in turn). These two genetic systems are physically separated and independent, and evolve separately. The high level of correspondence between the custom phylogeny and proximity in triplet frequency space proves the synchrony in evolution of these two genetic systems. The method shows a high sensitivity, if applied to cluster the set of genomes of lower taxonomy rank.
#329 Lost strings in bacterial genomes: random or targeted process?
Authors: Michael Sadovsky and Natalia Rudenko
Total number of k-tipples that could be found in DNA sequence grows exponentially. Obviously, for any genome there exists the specific length d of the strings so that some d-tipples are absent, in the sequence, just because of the Dirichlet principle. Thus, the key question here is whether the lost d-tipples are randomly eliminated from a sequence, or they exhibit some preference and order in the elimination pattern. Suppose, the shortest words that are eliminated due to natural selection, are not randomly wipped out from a sequence. Then, one should expect that phylogenetically close species (races, etc.) must exhibit a close proximity in the lost words sets. To test the idea, we selected a proper set of bacterial genomes, and check, whether the list of avoided strings exhibits a reasonable level of concordance. Some combinatorial issues have also been studied, to differ the natural selection processes from a noise. Reciprocally, we tested a list of mostly independent bacterial (and yeast) genomes for the concordance of the lists of the eliminated words. Finally, we tested the observed lists of the avoided strings (d-tipples) against a randomly selected entities. It has been found that the lists of the avoided strings is not random, for the bacterial and yeast genomes.
#336 Reinforcement learning accounts for conditional cooperation behavior in social dilemma games
Authors: Takahiro Ezaki, Yutaka Horita, Masanori Takezawa and Naoki Masuda
Behavioral experiments using repeated multiplayer social dilemma games in groups and networks have revealed that players condition their decisions on the fraction of cooperative partners in the previous round of the repeated game. This behavior is referred to as the (moody) conditional cooperation. The origin of conditional cooperation and its moody variant remains unclear. Here we provide a proximate explanation by numerical simulations. We show that players adopting a variant of the so-called Bush-Mosteller reinforcement learning rule show the targeted behavior. In the model, players are assumed to have no access to information about what other players are doing such that they cannot explicitly use conditional cooperation rules. We found that the reinforcement learners that showed (moody) conditional cooperation obeyed a behavioral pattern similar to the GRIM strategy, a well-known strategy in the repeated prisoner's dilemma game. A reinforcement-learning variant of the GRIM strategy seems to better explain the extant experimental results than the Pavlov strategy, which is an established strong competitor in the repeated prisoner's dilemma game.
Authors: Yogesh Bichpuriya, Narayanan Rajagopal and Krishna V Prasad
Electrical power system is a complex system having multiple agents interacting with each other to maintain the dynamic balance of demand and supply. Though these interactions were limited in the traditional system. Now, the electrical system is going through a major transformation from its traditional avatar to a future grid. The transformation brings challenges in different aspects of the system e. g., planning, operation and control etc. Increasing proportion of Intermittent distributed renewable energy, evolving market structure and interactions, and active participation of customers/prosumers are some of the major factors which should be considered for decision making.
In such a complex system with multiple agents and enhanced and extended interactions, existing modelling and simulation approaches to understand the behaviour of the system and to make optimal decisions would not be sufficient. In a complex system, understanding of agents and their interactions is also important for policy makers and regulators.
In this work, we present modelling of different agents and their interactions in the electricity market scenario. We use Agent Based Modelling approach to define these agents and to model their interactions. We will also discuss some use-cases to illustrate the impact of an agent’s behaviour on the other market agents and their strategy. The model will be useful for doing impact analysis of any new regulation or market mechanism. It will also help in doing what-if analysis by different agents in order to make a decision.
#322 Hierarchical Structure of World Trade Flow Network
Authors: Le Ahn Quang, Ashadun Nobi, Nam Jung, Tae Ho Lee, Seung Eun Maeng and Jae Woo Lee
We analyze the structural change of the trade flow network for different commodities from 1995 to 2013. The MST (minimum spanning tree) is constructed from data of the trade flow and investigates the hierarchical organization of the MST by the method of hierarchical path. We find that the tree for man-made products such as machinery and transport equipment and manufactured goods is more hierarchical than the natural products such as mineral fuels. We also observe that the hierarchy F is changed over time notably due to globalization and crises. The interesting finding is that when we remove China from trade flow network, the trend of the F is changed significantly which implies the role of China on world trade network in recent years. Finally, the network stability is identified by Jaccard index. We observe that the MSTs for the man-made products are more stable than those of the natural products.
#323 Chloroplasts and mitochondria exhibit wonderful synchrony in evolution with host genomes
Authors: Michael Sadovsky, Julia Putinzeva, Anna Chernyshova and Vaselina Fedotova
Very strong correlation between genome-wide determined structure and phylogeny of the genome bearer is found. Triplet frequency dictionary is the structure, and organelles genomes are the entities used to figure out the structural proximity. Clusterizing in the triplet frequencies space has been carried out, both by K-means, and elastic maps technique. The set of genomes was clusterized into 2, 3, ..., 8 classes, consequently, and the composition of the classes was studied. The composition of the classes was found to be very far from a random one: usually, two classes developed for K-means clusterization with K classes merged into a single class, as the clusterization has been developed for (K-1) classes. Besides, the observed phylogenetic composition of the classes strongly follows the phylogeny. Indeed, a clusterization for two classes over the mitochondrion genomes always yields a perfect separation of chordata species from invertibrate ones. Similarly, such clusterization carried out over the chloroplast genomes yields a perfect separation of higher Embryophytes from non-vascular plant. A growth of the number of classes identified through K-means always followed the classic phylogeny of organisms, in general. It should be stressed that the proximity in structure space (that is triplet frequency distribution determined genome-wide, with neither elimination or prevalence of some specific regions of a genome) was determined over the organellae genomes, while the phylogenetic proximity has been determined in traditional way, through morphology (that is determined by nuclear genome, in turn). These two genetic systems are physically separated and independent, and evolve separately. The high level of correspondence between the custom phylogeny and proximity in triplet frequency space proves the synchrony in evolution of these two genetic systems. The method shows a high sensitivity, if applied to cluster the set of genomes of lower taxonomy rank.
#329 Lost strings in bacterial genomes: random or targeted process?
Authors: Michael Sadovsky and Natalia Rudenko
Total number of k-tipples that could be found in DNA sequence grows exponentially. Obviously, for any genome there exists the specific length d of the strings so that some d-tipples are absent, in the sequence, just because of the Dirichlet principle. Thus, the key question here is whether the lost d-tipples are randomly eliminated from a sequence, or they exhibit some preference and order in the elimination pattern. Suppose, the shortest words that are eliminated due to natural selection, are not randomly wipped out from a sequence. Then, one should expect that phylogenetically close species (races, etc.) must exhibit a close proximity in the lost words sets. To test the idea, we selected a proper set of bacterial genomes, and check, whether the list of avoided strings exhibits a reasonable level of concordance. Some combinatorial issues have also been studied, to differ the natural selection processes from a noise. Reciprocally, we tested a list of mostly independent bacterial (and yeast) genomes for the concordance of the lists of the eliminated words. Finally, we tested the observed lists of the avoided strings (d-tipples) against a randomly selected entities. It has been found that the lists of the avoided strings is not random, for the bacterial and yeast genomes.
#336 Reinforcement learning accounts for conditional cooperation behavior in social dilemma games
Authors: Takahiro Ezaki, Yutaka Horita, Masanori Takezawa and Naoki Masuda
Behavioral experiments using repeated multiplayer social dilemma games in groups and networks have revealed that players condition their decisions on the fraction of cooperative partners in the previous round of the repeated game. This behavior is referred to as the (moody) conditional cooperation. The origin of conditional cooperation and its moody variant remains unclear. Here we provide a proximate explanation by numerical simulations. We show that players adopting a variant of the so-called Bush-Mosteller reinforcement learning rule show the targeted behavior. In the model, players are assumed to have no access to information about what other players are doing such that they cannot explicitly use conditional cooperation rules. We found that the reinforcement learners that showed (moody) conditional cooperation obeyed a behavioral pattern similar to the GRIM strategy, a well-known strategy in the repeated prisoner's dilemma game. A reinforcement-learning variant of the GRIM strategy seems to better explain the extant experimental results than the Pavlov strategy, which is an established strong competitor in the repeated prisoner's dilemma game.
Home |
Calls |
Programme |
Registration |
Contact |
Copyright CCS © 2016