The global grid

in Renewable Energy : An International Journal (2013), 57

This paper puts forward the vision that a natural future stage of the electricity network could be a grid spanning the whole planet and connecting most of the large power plants in the world: this is the ... [more ▼]

This paper puts forward the vision that a natural future stage of the electricity network could be a grid spanning the whole planet and connecting most of the large power plants in the world: this is the “Global Grid”. The main driving force behind the Global Grid will be the harvesting of remote renewable sources, and its key infrastructure element will be the high capacity long transmission lines. Wind farms and solar power plants will supply load centers with green power over long distances. This paper focuses on the introduction of the concept, showing that a globally interconnected network can be technologically feasible and economically competitive. We further highlight the multiple opportunities emerging from a global electricity network such as smoothing the renewable energy supply and electricity demand, reducing the need for bulk storage, and reducing the volatility of the energy prices. We also discuss possible investment mechanisms and operating schemes. Among others, we envision in such a system a global power market and the establishment of two new coordinating bodies, the “Global Regulator” and the “Global System Operator”. [less ▲]

Stratégies d'échantillonnage pour l'apprentissage par renforcement batch

in Revue d'Intelligence Artificielle [=RIA] (2013)

Meta-learning of Exploration/Exploitation Strategies: The Multi-Armed Bandit Case

in Filipe, Joaquim; Fred, Ana (Eds.) Agents and Artificial Intelligence: 4th International Conference, ICAART 2012, Vilamoura, Portugal, February 6-8, 2012. Revised Selected Papers (2013)

The exploration/exploitation (E/E) dilemma arises naturally in many subfields of Science. Multi-armed bandit problems formalize this dilemma in its canonical form. Most current research in this field ... [more ▼]

The exploration/exploitation (E/E) dilemma arises naturally in many subfields of Science. Multi-armed bandit problems formalize this dilemma in its canonical form. Most current research in this field focuses on generic solutions that can be applied to a wide range of problems. However, in practice, it is often the case that a form of prior information is available about the specific class of target problems. Prior knowledge is rarely used in current solutions due to the lack of a systematic approach to incorporate it into the E/E strategy. To address a specific class of E/E problems, we propose to proceed in three steps: (i) model prior knowledge in the form of a probability distribution over the target class of E/E problems; (ii) choose a large hypothesis space of candidate E/E strategies; and (iii), solve an optimization problem to find a candidate E/E strategy of maximal average performance over a sample of problems drawn from the prior distribution. We illustrate this meta-learning approach with two different hypothesis spaces: one where E/E strategies are numerically parameterized and another where E/E strategies are represented as small symbolic formulas. We propose appropriate optimization algorithms for both cases. Our experiments, with two-armed “Bernoulli” bandit problems and various playing budgets, show that the metalearnt E/E strategies outperform generic strategies of the literature (UCB1, UCB1-T UNED, UCB-V, KL-UCB and epsilon-GREEDY); they also evaluate the robustness of the learnt E/E strategies, by tests carried out on arms whose rewards follow a truncated Gaussian distribution. [less ▲]

Optimized Look-Ahead Tree Policies: A Bridge Between Look-Ahead Tree Policies and Direct Policy Search

in International Journal of Adaptive Control and Signal Processing (2013)

Direct policy search (DPS) and look-ahead tree (LT) policies are two popular techniques for solving difficult sequential decision-making problems. They both are simple to implement, widely applicable ... [more ▼]

Direct policy search (DPS) and look-ahead tree (LT) policies are two popular techniques for solving difficult sequential decision-making problems. They both are simple to implement, widely applicable without making strong assumptions on the structure of the problem, and capable of producing high performance control policies. However, computationally both of them are, each in their own way, very expensive. DPS can require huge offline resources (effort required to obtain the policy) to first select an appropriate space of parameterized policies that works well for the targeted problem, and then to determine the best values of the parameters via global optimization. LT policies do not require any offline resources; however, they typically require huge online resources (effort required to calculate the best decision at each step) in order to grow trees of sufficient depth. In this paper, we propose optimized look-ahead trees (OLT), a model-based policy learning scheme that lies at the intersection of DPS and LT. In OLT, the control policy is represented indirectly through an algorithm that at each decision step develops, as in LT using a model of the dynamics, a small look-ahead tree until a prespecified online budget is exhausted. Unlike LT, the development of the tree is not driven by a generic heuristic; rather, the heuristic is optimized for the target problem and implemented as a parameterized node scoring function learned offline via DPS. We experimentally compare OLT with pure DPS and pure LT variants on optimal control benchmark domains. The results show that the LT-based representation is a versatile way of compactly representing policies in a DPS scheme (which results in OLT being easier to tune and having lower offline complexity than pure DPS); while at the same time, DPS helps to significantly reduce the size of the look-ahead trees that are required to take high-quality decisions (which results in OLT having lower online complexity than pure LT). Moreover, OLT produces overall better performing policies than pure DPS and pure LT and also results in policies that are robust with respect to perturbations of the initial conditions. [less ▲]

Biorthogonalization Techniques for Least Squares Temporal Difference Learning

Poster (2012, December 07)

We consider Markov reward processes and study OLS-LSTD, a framework for selecting basis functions from a set of candidates to obtain a sparse representation of the value function in the context of least ... [more ▼]

We consider Markov reward processes and study OLS-LSTD, a framework for selecting basis functions from a set of candidates to obtain a sparse representation of the value function in the context of least squares temporal difference learning. To support efficient both updating and downdating operations, OLS-LSTD uses a biorthogonal representation for the selected basis vectors. Empirical comparisons with the recently proposed MP and LARS frameworks for LSTD are made. [less ▲]

Cooperative frequency control with a multi-terminal high-voltage DC network

in Automatica (2012), 48(12), 31283134

We consider frequency control in power systems made of several non-synchronous AC areas connected by a multi-terminal high-voltage direct current (HVDC) grid. We propose two HVDC control schemes to make ... [more ▼]

We consider frequency control in power systems made of several non-synchronous AC areas connected by a multi-terminal high-voltage direct current (HVDC) grid. We propose two HVDC control schemes to make the areas collectively react to power imbalances, so that individual areas can schedule smaller power reserves. The first scheme modifies the power injected by each area into the DC grid as a function of frequency deviations of neighboring AC areas. The second scheme changes the DC voltage of each converter as a function of its own area's frequency only, relying on the physical network to obtain a collective reaction. For both schemes, we prove convergence of the closed-loop system with heterogeneous AC areas. [less ▲]

Policy search in a space of simple closed-form formulas: towards interpretability of reinforcement learning

in Lecture Notes in Computer Science (2012), 7569

In this paper, we address the problem of computing interpretable solutions to reinforcement learning (RL) problems. To this end, we propose a search algorithm over a space of simple losed-form formulas ... [more ▼]

In this paper, we address the problem of computing interpretable solutions to reinforcement learning (RL) problems. To this end, we propose a search algorithm over a space of simple losed-form formulas that are used to rank actions. We formalize the search for a high-performance policy as a multi-armed bandit problem where each arm corresponds to a candidate policy canonically represented by its shortest formula-based representation. Experiments, conducted on standard benchmarks, show that this approach manages to determine both efficient and interpretable solutions. [less ▲]

Coordinated primary frequency control among nonsynchronous systems connected by a multi-terminal HVDC grid

in IET Generation, Transmission & Distribution (2012), 6(2), 99-108

The authors consider a power system composed of several non-synchronous AC areas connected by a multiterminal high-voltage direct current (HVDC) grid. In this context, the authors propose a distributed ... [more ▼]

The authors consider a power system composed of several non-synchronous AC areas connected by a multiterminal high-voltage direct current (HVDC) grid. In this context, the authors propose a distributed control scheme that modifies the power injections from the different AC areas into the DC grid so as to make the system collectively react to load imbalances. This collective reaction allows each individual AC area to downscale its primary reserves. The scheme is inspired by algorithms for the consensus problem extensively studied by the control theory community. It modifies the power injections based on frequency deviations of the AC areas so as to make them stay close to each other. A stability analysis of the closed-loop system is reported as well as simulation results on a benchmark power system with five AC areas. These results show that with proper tuning, the control scheme makes the frequency deviations converge rapidly to a common value following a load imbalance in an area. [less ▲]

Learning exploration/exploitation strategies for single trajectory reinforcement learning

in Proceedings of the 10th European Workshop on Reinforcement Learning (EWRL 2012) (2012)

We consider the problem of learning high-performance Exploration/Exploitation (E/E) strategies for finite Markov Decision Processes (MDPs) when the MDP to be controlled is supposed to be drawn from a ... [more ▼]

We consider the problem of learning high-performance Exploration/Exploitation (E/E) strategies for finite Markov Decision Processes (MDPs) when the MDP to be controlled is supposed to be drawn from a known probability distribution pM( ). The performance criterion is the sum of discounted rewards collected by the E/E strategy over an in finite length trajectory. We propose an approach for solving this problem that works by considering a rich set of candidate E/E strategies and by looking for the one that gives the best average performances on MDPs drawn according to pM( ). As candidate E/E strategies, we consider index-based strategies parametrized by small formulas combining variables that include the estimated reward function, the number of times each transition has occurred and the optimal value functions V and Q of the estimated MDP (obtained through value iteration). The search for the best formula is formalized as a multi-armed bandit problem, each arm being associated with a formula. We experimentally compare the performances of the approach with R-max as well as with e-Greedy strategies and the results are promising. [less ▲]

Comparison of Different Selection Strategies in Monte-Carlo Tree Search for the Game of Tron

in IEEE Conference on Computational and Intelligence in Games 2012 (2012)

Monte-Carlo Tree Search (MCTS) techniques are essentially known for their performance on turn-based games, such as Go, for which players have considerable time for choosing their moves. In this paper, we ... [more ▼]

Monte-Carlo Tree Search (MCTS) techniques are essentially known for their performance on turn-based games, such as Go, for which players have considerable time for choosing their moves. In this paper, we apply MCTS to the game of Tron, a simultaneous real-time two-player game. The fact that players have to react fast and that moves occur simultaneously creates an unusual setting for MCTS, in which classical selection policies such as UCB1 may be suboptimal. In this paper, we perform an empirical comparison of a wide range of selection policies for MCTS applied to Tron, with both deterministic policies (UCB1, UCB1-Tuned, UCB-V, UCBMinimal, OMC-Deterministic, MOSS) and stochastic policies (Epsilon-greedy, EXP3, Thompson Sampling, OMC-Stochastic, PBBM). From the experiments, we observe that UCB1-Tuned has the best behavior shortly followed by UCB1. Even if UCB-Minimal is ranked fourth, this is a remarkable result for this recently introduced selection policy found through automatic discovery of good policies on generic multi-armed bandit problems. We also show that deterministic policies perform better than stochastic ones for this problem. [less ▲]