In this thesis a transformer fault diagnosis system using probabilistic neural network (PNN) is proposed and implemented. Many artificial neural networks (ANN) have been proposed to deal with the transformer fault diagnosis. However, when dissolved gas records change, adaptation capability becomes a problem in ANN applications. PNN analyzes the dissolved gas contents in the oil-immersed transformer to identify various fault types. Numerical gas ratios of oil and cellulose decomposition were used to create the training examples. Retraining can also be done by adding new examples without any iteration. With diagnostic gas records, computer simulations were conducted to show the effectiveness of the proposed system. The Internet based power transformer monitoring system was also proposed in this thesis . LabVIEW was used to develop the Man-Machine Interface (MMI), and DataSocket tool was used to share the information on Internet. Application of the harmonic load flow based on the Equivalent- Current Injection was used to solve harmonic problems. There are two sub-models including the fundamental and harmonic frequency models. The standard Fourier analysis was used to deal with the harmonic loads to get injection currents. A passive filter was also developed to improve harmonics to satisfy restriction standards of the Taiwan Power Company.