In traditional electric power metering systems, concerns about high electricity bills, potential theft, incorrect price calculations, and excessive load consumption have prompted the need for innovative solutions. This paper introduces a novel IoT-based Smart Meter, leveraging an ESP8266 Wi-Fi module to enable real-time monitoring of appliance load consumption. The Smart Meter seamlessly integrates with the Thingspeak IoT platform via the Internet. However, the reliability of data collected over the internet is often compromised, necessitating a robust anomaly detection mechanism. This study focuses on investigating anomaly detection through machine learning techniques, with a particular emphasis on the selection of datasets crucial to the efficacy of the ML algorithm. The research showcases a comparative analysis of various anomaly detection algorithms, including KNN, Autoencoder, OCSVM, and kernel PCA. Notably, the Autoencoder method stands out, achieving a maximum accuracy of 79%. The study delves into the intricacies of each algorithm, shedding light on their comparative performance and offering insights into optimizing anomaly detection for smart energy meters.