Sacrificial Anode Cathodic Protection (SACP) is a commonly used technique for corrosion protection of under-ground steel structures. The short-circuit current flowing between the anode and the cathode is measured to model the lifetime of the sacrificial anode (SA) for its eventual replacement. We have designed a hardware system to harvest energy from an SACP cell to power an underground wireless sensor node to send data (that is, the short-circuit current) to the cloud to model the lifetime of the SA. We also measure the ambient temperature, humidity, soil temperature, and soil moisture, because those parameters affect the short-circuit current. In this paper, we propose a methodology to detect anomalous behavior of SACP cells by processing the aforementioned data collected from SACP cells. The methodology applies both changepoint detection theory and One-Class SVM (Support Vector Machine) on the same dataset. For the One-Class SVM (OCSVM) approach, the features consist of the principal components of the normalized input data. In the changepoint analysis approach: (i) the features consist of the statistical properties of the changepoints of the input data; and (ii) an optimal classifier is found among Decision Trees, Random Forests, Naive Bayesian, Logistic Regression, and SVM models. The irregular flow of current between anode and cathode, changes in the charging and discharging rate of the capacitor, or the changes in circuit operations of the SACP cell due to changes in environmental conditions indicate the anomalous behaviour of the SACP cell. In this work, we show how feature extraction improves the overall efficiency of the model to detect anomalous behavior. Moreover, comparing both the approaches proposed above shows that changepoint detection outperforms OCSVM by 7% in terms of accuracy.