DEVELOPMENT OF ANOMALY BEHAVIOR DETECTION SYSTEM ON BANDUNG INSTITUTE OF TECHNOLOGY (ITB) LABTEK VI BUILDINGâS ELECTRICITY COMSUMPTION DATA USING K-MEANS ALGORITHM-BASED CLUSTERING METHOD
The occurance of Non-Technical Losses (NTLs) is a real impact of the prevalence of non-tchnical errors both in the process of installation, transmission, and distribution of electricity. This can be seen in the precentage of losses in the electricity network in Indonesia in 2020 wich reached 9.2%. T...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Project |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/71803 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | The occurance of Non-Technical Losses (NTLs) is a real impact of the prevalence of non-tchnical errors both in the process of installation, transmission, and distribution of electricity. This can be seen in the precentage of losses in the electricity network in Indonesia in 2020 wich reached 9.2%. The most commong factor encountered in the field is the high number of cases of electricity theft. Detection of anomalous behavior by extracting load profiles and patterns of electricity usage is a first step that has the potential to uncover acts of theft of electricity power. In this final project research, a system for detecting anomaly behavior will be developed using the clustering method based on the k-means algorithm. The input data was obtained from the energy management system of the Labtek VI Building, Bandung Institute of Technology (ITB), which has carried out a series of data processing in the form of Exploratory Data Analysis (EDA) and normalization data.
From a series of processing and model simulations, the research results obtained in the form of features used as model input features are power (P), voltages on each phase (V1, V2, V3), currents on each phase (A1, A2, A3), power factor on each phase (PF1, PF2, PF3), and On_hours. Besides that, the timestamps feature is also used for graphic plot to identify the patterns of electric power usage. From the simulations also results, the visualization of the elbow method is generated optimum degree value of k = 3 which indicates the number of clusters that will be formed. From three clusters, then it is generated to a scatter graph that showed sluster 2 is the cluster with the most anomalies. And then, the simulations process also generates a graphical plot of daily electricity usage which clearly shows a jump in power usage or is called anomaly.
|
|---|