Smart meter devices enable a better understanding of the demand at potential risk private information leakage. One promising solution to mitigating such is inject noises into data achieve certain level differential privacy. In this article, we cast one-shot non-intrusive load monitoring (NILM) in compressive sensing framework, and bridge gap between NILM inference accuracy privacy's parameters. We then derive valid theoretical bounds offer insights on how privacy parameters affect...

Conventional wisdom to improve economic dispatch effectiveness is design the load forecasting method as accurately possible. However, this approach can be problematic due temporal and spatial correlations between system cost prediction errors. This observation motivates us jointly treat two forms of by adopting notion end-to-end machine learning. Thus, we first propose a task-specific learning criterion conduct for maximal benefits. To reduce approach's computational burden over-fitting...

Flexible resources are increasingly significant for the reliable operation of power grids due to high penetration renewable energy. Thermostatically controlled loads (TCLs) one common flexible resources, whose control has been extensively studied. Yet, much can be improved. We investigate scheduling TCLs facing uncertain temperatures and dynamic prices. Classical approaches often employ chance-constrained program or robust optimization handle such uncertainties. However, these either require...

The increasing number of electric vehicles (EVs) on the road brings both opportunities and challenges to power system. For EV charging stations (EVCSs), it is often difficult conduct effective operations due incomplete information in EVs' departure times opacity their preference information. To tackle this challenge, we seek design optimal deadline differentiated dynamic price menu that offers multiple choice-pairs deadlines prices. We prove such menus can incentivize EVs truthfully reveal...

The Internet of Things (IoT) enables reliable and fast data collection transmission, providing key infrastructure for power generation, distribution, control in the smart grid. This IoT-enabled grid tackles challenges brought by renewable penetration new ways: Accurate real-time information allows application artificial-intelligence-powered computation. We employ deep learning framework consider problem storage facing uncertainties generation. propose both model-based model-free frameworks...

The high penetration of renewable energy brings significant uncertainty to the power grids. Taking economic dispatch (ED) as an example, inaccurate prediction generations dramatically increases cost and risks grid's reliable operation. accurate distribution knowledge enables modeling ED stochastic programming with joint chance constraints, which various classical methods can tackle. However, in practice, such is inaccessible, we only observe samples from some unknown distribution. This makes...

The integration of distributed energy resources, particularly wind energy, presents both opportunities and challenges for the modern electrical grid. On supply side, farms frequently encounter penalties due to power's intermittency variability. incorporation storage systems can mitigate these through real-time power adjustments. However, uncertainties in future renewable generation significantly impede optimal control, existing algorithms either lack theoretical guarantees, or fail...

Renewable energy penetration increases the power grid's operational uncertainty, threatening economic effectiveness and reliability of grid. In this paper, we examine how uncertainty affects unit commitment (UC), a classical electricity market procedure. Stochastic programming has helped handle for UC performed well with distribution knowledge, but lack such information in practice deteriorates effectiveness. Such dilemma becomes more pronounced when dealing joint chance constraints solely...

The smart grid benefits and suffers from meter data. Proper use of massive data can improve energy services but may raise privacy concerns. For example, user consumption profiling, a classic method, identify patterns based on the collected load profiles users. Thus, these individual needs to be protected. However, most existing works focus transmission calculation privacy, often require additional computation, communication, or platform construction costs. In contrast, noise-injection-based...

Conventional wisdom to improve the effectiveness of economic dispatch is design load forecasting method as accurately possible. However, this approach can be problematic due temporal and spatial correlations between system cost prediction errors. This motivates us adopt notion end-to-end machine learning propose a task-specific criteria conduct dispatch. Specifically, maximize data utilization, we an efficient optimization kernel for process. We provide both theoretical analysis empirical...

Global urbanization has enabled worldwide economic growth over the past decade. Such legend yields a dramatically increasing population for major metropolises, which heavily burdens transportation sector. The huge personal demands warrant an efficient platform to support dynamic mobile services and their operation. In this work, we focus on complete dispatch ride-hailing platform. We first model uncertainty in both supply side demand side. Then propose network flow accelerated algorithm...

The transportation sector is one of the main consumers global energy. So, its electrification crucial for a sustainable future. However, slow developments in public infrastructure can be major bottleneck such electrification. An increasing number electric vehicle (EV) charging stations are being built across world to improve this infrastructure. Competition amongst EV improves market efficiency. In paper, effect competition on setting service surcharge investigated. design characterization...

Wind power, a prominent renewable energy resource, contributes to substantial yet volatile supply the power grid, posing significant risks both and demand. On side, wind farms often incur penalty costs due discrepancy between actual committed generation specified in electricity contracts. To address this issue, storage systems are widely adopted for real-time regulation of delivery, effectively mitigating fluctuations. However, effective control is hindered by uncertainties surrounding...

Probabilistic forecasting can characterize the uncertainties and dynamic trends of future residential load, while massive data are required for popular methods. In this study, we consider probabilistic load users who only willing to provide limited samples due privacy concerns. To address challenge, analyze characteristics employ clustering-based few-shot learning methods augment data. Meanwhile, combine different models, known as model ensemble, further improve performance. Compared with...

EV charging scheduling is a fundamental problem for infrastructure management as EVs impose both challenges and opportunities on the power grid. The primary difficulty comes from uncertainty associated with arriving EVs, including their arrival time, departure volume. This makes in single station already challenging. Clearly, joint across different stations even more To this end, paper, we first develop risk-limiting approach based chance-constrained optimization offline setting. enable...

Reinforcement Learning (RL) algorithms are known to suffer from the curse of dimensionality, which refers fact that large-scale problems often lead exponentially high sample complexity. A common solution is use deep neural networks for function approximation; however, such approaches typically lack theoretical guarantees. To provably address we observe many real-world exhibit task-specific model structures that, when properly leveraged, can improve efficiency RL. Building on this insight,...