With ambitious policies in place internationally for producing electricity by renewables, the development and installation of sustainable, safe power sources are already underway. Solar energy is one such power source; readily available, reliable and safe. Photovoltaic (PV) generally operate in harsh environments, which exposes them to numerous stresses that can affect performance and lead to component failure. Therefore, continuous monitoring of the health of photovoltaic systems is necessary for early fault detection and diagnosis to ensure continuing effective performance. The existing techniques for continuous condition monitoring (CM) of PV power systems are largely based on monitoring a system’s electrical, thermal and environmental parameters. These relatively simple techniques have been shown to be limited in their diagnostic reliability. Intelligent PV CM systems that can substantially improve awareness of the system’s health and have an early fault diagnosis capability would be hugely beneficial. To assist in the achievement of intelligent CM, this PhD research program emphasises the Internet of Things (IoT) as a means to access monitored health data available in large quantities for accurately tracking and predicting the health of a PV power systems in real-time. The main objective of this research has been to utilise machine learning algorithms, effective data transfer and management methods, including various IoT techniques for in-depth observation of the health of a PV power systems. This objective has been achieved, and the techniques used evaluated, by investigating the data characteristics and control of an in- house, purpose-designed small scale PV system. To improve PV performance three separate and distinct means of tracking the maximum power point (MPPT) have been examined, namely, Fuzzy logic, perturb and observe, and modified perturb and observe algorithm with varying step-size. Data yielded by real-time tracking using the IoT has been used for teaching a neural network to detect and classify defects in a PV power system. The dataset includes as input data measured values of environmental temperature, humidity, irradiance, voltage, current, and power. Also, measured data extracted from the PV system, healthy and with electrical faults and partial and full shading of the PV modules seeded into the system are included. Optimum values of the bias vectors and weights for the Artificial Neural Network (ANN) were determined using Cuckoo Search (CSA) and Genetic Algorithms (GA) to accelerate convergence of the ANN and boost its accuracy. Three optimization scenarios, ANN without optimization, ANN optimized using the CSA, and ANN optimized using the GA, have been compared to evaluate the performance parameters. The reported results include comparison of the MPPT tracking methods when monitored via the IoT, with detection and classification based on the three optimization scenarios listed above. The accuracy of the respective predictions have been evaluated based on regression analysis of predicted and measured data.