Search for: All records

Prosumers with generation and storage capabilities can supply en- ergy back to the grid, or trade their surplus with other prosumers for their mutual benefit. A prosumer aggregation that facilitates such trades will price the energy being traded to achieve an objective such as profit maximization, social welfare, or market equilibrium. We propose the use of reinforcement learning to design a trans- active controller to price energy in a prosumer aggregation. This has an advantage over other decentralized pricing mechanisms as it does not rely on iterative price settlement or load estimation by prosumers, and estimates the price in a day ahead manner. We present numerical case studies to evaluate our controller, and dis- cuss extensions to implement this in real prosumer aggregations. 

This paper considers off-street parking for the cruising vehicles of transportation network companies (TNCs) to reduce the traffic congestion. We propose a novel business that integrates the shared parking service into the TNC platform. In the proposed model, the platform (a) provides interfaces that connect passengers, drivers and garage operators (commercial or private garages); (b) determines the ride fare, driver payment, and parking rates; (c) matches passengers to TNC vehicles for ride-hailing services; and (d) matches vacant TNC vehicles to unoccupied parking garages to reduce the cruising cost. A queuing-theoretic model is proposed to capture the matching process of passengers, drivers, and parking garages. A market-equilibrium model is developed to capture the incentives of the passengers, drivers, and garage operators. An optimization-based model is formulated to capture the optimal pricing of the TNC platform. Through a realistic case study, we show that the proposed business model will offer a Pareto improvement that benefits all stakeholders, which leads to higher passenger surplus, higher drivers surplus, higher garage operator surplus, higher platform profit, and reduced traffic congestion. 

We propose an approach to design a Model Predictive Controller (MPC) for constrained Linear Time Invariant systems performing an iterative task. The system is subject to an additive disturbance, and the goal is to learn to satisfy state and input constraints robustly. Using disturbance measurements after each iteration, we construct Confidence Support sets, which contain the true support of the disturbance distribution with a given probability. As more data is collected, the Confidence Supports converge to the true support of the disturbance. This enables design of an MPC controller that avoids conservative estimate of the disturbance support, while simultaneously bounding the probability of constraint violation. The efficacy of the proposed approach is then demonstrated with a detailed numerical example. 

The value created by aggregating behind-the-meter distributed energy storage devices for grid services depends on how much storage is in the system and the power network operation conditions. To understand whether market-driven distributed storage investment will result in a socially desirable outcome, we formulate and analyze a network storage investment game. By explicitly characterizing the set of Nash equilibria (NE) for two examples, we establish that the uniqueness and efficiency of NE depend critically on the power network conditions. Furthermore, we show it is guaranteed that NE support social welfare for general power networks, provided we include two modifications in our model. These modifications suggest potential directions for regulatory interventions. 

Frequency regulation is crucial for balancing the supply and demand of modern electricity grids. To provide regulation services, it is important to understand the capability of flexible resources to track regulation signals. This paper studies the problem of submitting capacity bids to a forward regulation market based on historical regulation signals. We consider an aggregator who manages a group of flexible resources with linear dynamic constraints. He seeks to find the optimal capacity bid, so that real-time regulation signals can be followed with an arbitrary guaranteed probability. We formulate this problem as a chance-constrained program with unknown regulation signal distributions. A sampling and discarding algorithm is proposed. It provably provides near-optimal solutions at a guaranteed probability of success without knowing the distribution of the regulation signals. This result holds for resources with arbitrary linear dynamics and allows arbitrary intra-hour data correlations. We validate the proposed algorithm with real data via numerical simulations. Two cases are studied: (1) CAISO market, where providers separately submit capacity estimates for regulation up and regulation down signals, (2) PJM market, where regulation up and down capacities are the same. Simulation results show that the proposed algorithm provides near-optimal capacity estimates for both cases. 

Early detection of incipient faults is of vital im- portance to reducing maintenance costs, saving energy, and enhancing occupant comfort in buildings. Popular supervised learning models such as deep neural networks are considered promising due to their ability to directly learn from labeled fault data; however, it is known that the performance of supervised learning approaches highly relies on the availability and quality of labeled training data. In Fault Detection and Diagnosis (FDD) applications, the lack of labeled incipient fault data has posed a major challenge to applying these supervised learning techniques to commercial buildings. To overcome this challenge, this paper proposes using Monte Carlo dropout (MC-dropout) to enhance the supervised learning pipeline, so that the resulting neural network is able to detect and diagnose unseen incipient fault examples. We also examine the proposed MC-dropout method on the RP-1043 dataset to demonstrate its effectiveness in indicating the most likely incipient fault types. 

In the classical risk limiting dispatch (RLD) formulation, the system operator dispatches generators relying on information about the distribution of demand. In practice, such information is not readily available and therefore is estimated using historical demand and auxiliary information (or features) such as weather forecasts. In this paper, instead of using a separated estimation and optimization procedure, we propose learning methods that directly compute the RLD decision rule based on historical data. Using tools from statistical learning theory, we then develop generalization bounds and sample complexity results of the proposed methods. These algorithms and performance guarantees, developed for the single-bus network, are then extended to a general network setting for the uniform reserve case.