Smart Distribution Boards (Smart DB), Non-Intrusive Load Monitoring (NILM) for Load Device Appliance Signature Identification and Smart Sockets for Grid Demand Management

Abstract

Traditionally, the choices to balance the grid and meet its peaking power needs are by installing more spinning reserves or perform load shedding when it becomes too much. This problem becomes worse as more intermittent renewable energy resources are installed, forming a substantial amount of total capacity. Advancements in Energy Storage System (ESS) provides the utility new ways to balance the grid and to meet its peak demand by storing un-used off peak energy for peak usage. Large sized ESS—mega watt (MW) level—are installed by different utilities at their substations to provide the high speed grid stabilization to balance the grid to avoid installing more capacity or triggering any current load shedding schemes. However, such large sized ESS systems and their required inverters are costly to install, require much space and their efficacy could also be limited due to network fault current limits and impedances. In this paper, we propose a novel approach and trial for 3000+ homes in Singapore of achieving a large capacity of demand management by developing a smart distribution board (DB) in each home with the high speed metering sensors (>6 kHz sampling rate) and non-intrusive load monitoring (NILM) algorithm, that can assist home users to perform the load/appliance profile identification with daily usage patterns and allow targeted load interruption using the smart sockets/plugs provided. By allowing load shedding at device or appliance level, while knowing their usage profile and preferences, this can allow such an approach to become part of a new voluntary interruptible load management system (ILMS) that requires little user intervention, while minimizing disruption to them, allowing ease of mass participation and thus achieving the intended MW demand management capacities for the grid. This allows for a more cost effective way to better balance the grid without the need for generation capacity growth, large ESS investment while improving the way to perform load shedding without disruptions to entire districts. Simply, home users can now know and participate with the grid in interruptible load (IL) schemes to target specific home appliance, such as water heaters or air conditioning, allowing interruptions during certain times of the day, instead of the entire house, albeit with the right incentives. This allows utilities to achieve MW capacity load shedding with millions of appliances with their preferences, and most importantly, with minimal disruptions to their consumers quality of life. In our paper, we will also consider coupling a small sized Home Energy Storage System (HESS) to amplify the demand management capacity. The proposed approach does not require any infrastructure or wiring changes and is highly scalable. Simulation results demonstrate the effectiveness of the NILM algorithm and achieving high capacity grid demand management. This approach of taking user preferences for appliance level load shedding was developed from the results of a survey of 500 households that indicates >95% participation if they were able to control their choices, possibly allowing this design to be the most successful demand management program than any large ESS solution for the utility. The proposed system has the ability to operate in centralized as part of a larger Energy Management System (EMS) Supervisory Control And Data Acquisition (SCADA) that decide what to dispatch as well as in autonomous modes making it simpler to manage than any MW level large ESS setup. With the availability of high-speed sampling at the DB level, it can rely on EMS SCADA dispatch or when disconnected, rely on the decaying of the grid frequency measured at the metering point in the Smart DB. Our simulation results demonstrate the effectiveness of our proposed approach for fast grid balancing.

Applications of Multi-Agent Deep Reinforcement Learning: Models and Algorithms

“…P. 3 Higher quality of service reflects better system performance, such as higher throughput [35], lower end-to-end delay (or shorter packet delivery time [6]), and lower blocking probability (or the probability of an agent being denied of services due to poor system performance) [31]. P. 4 Higher energy efficiency reduces energy consumption of devices, such as sensors and smart sockets embedded in wireless modules [64]. A balanced adjustment of on-time and off-time during the peak and non-peak times, respectively, increases energy efficiency [56].…”

“…The controllers maximize the global reward to ensure a comfortable level in a collaborative (X.2.2) manner. The proposed approach addresses the challenge of ultradensification (C.1) caused by a large number of sensors, including smart sockets [64], motion sensors, temperature sensors, and door sensors.…”

Toward Robust Non-Intrusive Load Monitoring via Probability Model Framed Ensemble Method

“…Usually, the smart socket is deployed to the plug of target appliance, and utilized for the monitoring of only certain appliance. In addition to the signal acquisition and processing module, the communication module is also required in smart socket [4]. Therefore, multiple appliances monitoring requires multiple sockets, and it is really expensive to deploy the smart socket-based sensing networks in a house.…”

An Enhanced Ensemble Approach for Non-Intrusive Energy Use Monitoring Based on Multidimensional Heterogeneity

“…The exploration enthusiasm for such a project had lasted for a period of time, but it decreased due to the high financial costs associated with too many sockets [5]. Besides, this method of electricity monitoring is strictly confined to socketed appliances, and it is not friendly to residents due to intrusive installations [6]. Hence, the socket-based sensor network, considered as the intrusive way, was replaced once non-intrusive approaches achieved reliable performance.…”