International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 11 Issue: 04 | Apr 2024
p-ISSN: 2395-0072
www.irjet.net
AN APPROACH FOR DEMAND SIDE MANAGEMENT USING K- MEANS CLUSTERING Shraddha Sharma1, Prof. Seema Pal2 1 Dept. of Electrical Engineering, Jabalpur Engineering College, Jabalpur, Madhya Pradesh, India
2Assistant professor, Dept. of Electrical Engineering, Jabalpur Engineering College, Jabalpur, Madhya Pradesh,
India ---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Smart meter is an advanced metering
enormous amount of high frequency data, which exhibits the characteristics of Big data i.e. velocity, volume, variability, variety and value thus, require a robust communication infrastructure for data processing and storage at utility end. This data being highly dimensional in nature, greatly impacts the analysis and deduction process making it less efficient due to curse of dimensionality. This challenge necessitates the use of dimensionality reduction algorithms or techniques.
infrastructure (AMI) that includes a smart meter, a bidirectional communication network, and a data management system. Using data analytics and machine learning to evaluate high-frequency smart meter data yields important insights into home power consumption trends, as well as improved load forecasting and demand response management implementation. In this study, Principal Component Analysis (PCA) is employed as a dimensionality reduction technique to extract features from a dataset collected from the UMassTrace repository. The K-means unsupervised partitional clustering algorithm uses three distance metrics to cluster reduced data: Euclidean, Manhattan, and Pearson correlation distances. MATLAB programming software is used to do feature computation and clustering. The clustering model is evaluated by obtaining the average silhouette coefficient. Euclidean distance is obtained to perform best with better average silhouette coefficient, indicating that data points in a cluster are compact and far apart from other clusters, making distance measurement preferable for clustering consumer load profiles for better demand side management.
Dimensionality reduction techniques convert high dimensional data to reduced dimension without the loss of significant information. These techniques when employed reduces the computational complexity associated with smart meter data, as every data obtained from smart meters are not helpful in drawing useful conclusions[4]. Once converted to lower dimensionality, these data can be used by consumers and utility operators to deduce important results and understand energy consumption trends, anomaly detection, energy theft and better demand side management. Energy consumption behaviors of individual consumers are used by utility for improving better demand side management. It selects the appropriate number of consumers to participate and present precise data on peak energy consumers. Clustering is used to group the load profile of different types of consumers in a distribution network. The main basis of clustering is to group load profile in different clusters with minimum intra-cluster distance or maximum intra-cluster similarity and maximum inter-cluster distance or minimum inter-cluster similarity. The two broad categories of clustering methods are hierarchical and partitional clustering methods. Hierarchical clustering groups the load profile into different clusters by generating nested partitions [5].In Partitional clustering method each cluster is represented by its center which summarizes all the load profile present in the cluster. The main focus is to optimize the objective function, which is the distance between the center and all the load profiles.
Key Words: Smart meters, Dimensionality reduction, PCA, K-means, Manhattan distance, Euclidean distance, Pearson correlation distance, Average silhouette coefficient, Demand response management
1. INTRODUCTION Advanced metering infrastructure (AMI) which comprises of smart meter, bidirectional communication network and data management system are being increasingly deployed in recent years. They have significant role by providing benefits to end consumers, network operators and energy suppliers. Smart meters offer range of functions such as advance metering, control, data storage and communication technologies .It helps consumers by providing them near real time consumption patterns which help them to manage their energy usage, reduce greenhouse gases emission and save money[1].It improves demand management, network planning and operation by providing accurate demand forecast ,locate outages and shorten supply restoration time, reduce operational and maintenance costs of network and improve asset utilization in distribution[2][3]. Smart meters generate
© 2024, IRJET
|
Impact Factor value: 8.226
In this paper, PCA has been used for dimensionality reduction and k -means partitional clustering method for clustering of different consumer profile. An evaluation index, silhouette coefficient is used to compare the
|
ISO 9001:2008 Certified Journal
|
Page 2559