Smart Non-Intrusive Appliance Load-Monitoring System Based on Phase Diagram Analysis

Smart Cities, Journal Year: 2024, Volume and Issue: 7(4), P. 1936 - 1949

Published: July 23, 2024

Much of today’s power grid was designed and built using technologies organizational principles developed decades ago. The lack energy resources classic networks are the main causes development smart to efficiently use resources, with stable safe operation. In such a network, one fundamental priorities is provided by non-intrusive appliance load monitoring (NIALM) in order analyze, recognize determine electricity consumption each consumer. this paper, we propose new system approach for characterization signature based on data-driven method, namely phase diagram. Our aim appliances different types consumers that can exist within building.

Language: Английский

---

FOCCA: Fog-cloud continuum architecture for data imputation and load balancing in Smart Grids

Computer Networks, Journal Year: 2025, Volume and Issue: unknown, P. 111031 - 111031

Published: Jan. 1, 2025

Language: Английский

---

General NILM Methodology for Algorithm Parametrization, Optimization and Performance Evaluation

Buildings, Journal Year: 2025, Volume and Issue: 15(5), P. 705 - 705

Published: Feb. 23, 2025

The research area of NILM exhibits a high heterogeneity regarding approaches and characteristics, especially in terms the applied algorithms, measurement data, quantities, features used, as well congruent appliance event state definitions. Therefore, performance evaluation establishment comparability is not straightforward. aim presented work was to address these challenges through development an application-oriented, general methodology for parametrization, optimization, existing algorithms. based on framework applicable wide range data. Temporary, individual measurements are utilized build extended database providing reliable ground truth common metrics. definition also formulated. application focused event-based algorithms data commercial building one significant appliance, relation total energy demand building. proved be suitable intended purpose. Two different event-detection could optimized their input parameters, able identify operation behavior optimally.

Language: Английский

---

Non-Intrusive Load Identification Based on Multivariate Features and Information Entropy-Weighted Ensemble

Energies, Journal Year: 2025, Volume and Issue: 18(9), P. 2369 - 2369

Published: May 6, 2025

In non-intrusive load monitoring (NILM), single-dimensional features exhibit limited representational capacity, while feature fusion at the layer often leads to information loss due dimensional transformation, as well risk of explosion caused by newly added features. To address these challenges, this paper proposes a identification method based on multivariate and entropy-weighted ensemble. Specifically, one-dimensional numerical related power current are input into traditional machine learning models, two-dimensional image binary V-I trajectory processed deep neural network model Swin Transformer. Information entropy is employed adaptively determine weight each classification model, weighted voting strategy utilized combine decisions multiple models obtain final result. This approach achieves decision layer, effectively avoiding transformations fully leveraging complementary advantages from different dimensions. Experimental results show that proposed accuracies 99.48% 99.54% public datasets PLAID WHITED, respectively.

Language: Английский

---

EOLD: A reinforcement learning-based energy-optimised load disaggregation framework for demand-side energy management

Renewable Energy, Journal Year: 2025, Volume and Issue: unknown, P. 123536 - 123536

Published: May 1, 2025

Language: Английский

---

A Comprehensive Review of Behind-the-Meter Distributed Energy Resources Load Forecasting: Models, Challenges, and Emerging Technologies

Energies, Journal Year: 2024, Volume and Issue: 17(11), P. 2534 - 2534

Published: May 24, 2024

Behind the meter (BTM) distributed energy resources (DERs), such as photovoltaic (PV) systems, battery storage systems (BESSs), and electric vehicle (EV) charging infrastructures, have experienced significant growth in residential locations. Accurate load forecasting is crucial for efficient operation management of these resources. This paper presents a comprehensive survey state-of-the-art technologies models employed process BTM DERs recent years. The review covers wide range models, from traditional approaches to machine learning (ML) algorithms, discussing their applicability. A rigorous validation essential ensure model’s precision reliability. Cross-validation techniques can be utilized reduce overfitting risks, while using multiple evaluation metrics offers assessment predictive capabilities. Comparing predictions with real-world data helps identify areas improvement further refinement. Additionally, U.S. Energy Information Administration (EIA) has recently announced its plan collect electricity consumption identified U.S.-based crypto mining companies, which exhibit abnormal patterns due rapid fluctuations. Hence, some case studies been presented that focus on irregular buildings equipped DERs. These activities underscore importance implementing robust anomaly detection address deviations typical usage profiles. Thus, our proposed framework, DERs, considering smart meters (SMs). Finally, thorough exploration potential challenges emerging based artificial intelligence (AI) large language (LLMs) suggested promising approach.

Language: Английский

---

Comparing four machine learning algorithms for household non-intrusive load monitoring

Energy and AI, Journal Year: 2024, Volume and Issue: 17, P. 100384 - 100384

Published: Sept. 1, 2024

The combination of Machine Learning (ML), smart energy meters, and availability household appliance profile data has opened new opportunities for Non-Intrusive Load Monitoring (NILM). However, the number options makes it challenging in selecting optimal combinations different applications, which requires studies to examine their trade-offs. This paper contributes one such study that investigated four established ML approaches – K Nearest Neighbour (KNN), Support Vector (SVM), eXtreme Gradient Boosting (XGBoost) Convolutional Neural Network (CNN) performance classifying events from Alternating Current (AC) Root Mean Square (RMS) where sampling frequency training dataset set size was varied (10 Hz–1 kHz 50–2000 examples per class, respectively). computational expense during training, testing storage also assessed evaluated with reference real-world applications. CNN classifier trained on AC at 500 Hz 11,000 gave best F1-score 0.989 followed by KNN 0.940. required models 3̃MB, is very close fitting cost-effective embedded system microcontrollers. would prevent high-rate needing be sent cloud as analysis could performed edge computing Internet-of-Things (IoT) devices.

Language: Английский

---

Disaggregating Household Water End-Uses: A Comparative Study Between XGBoost and TabNet

Published: Aug. 16, 2024

Language: Английский

---