Search for: All records

Timely and accurate detection of events affecting the stability and reliability of power transmission systems is crucial for safe grid operation. This paper presents an efficient unsupervised machine-learning algorithm for event detection using a combination of discrete wavelet transform (DWT) and convolutional autoencoders (CAE) with synchrophasor phasor measurements. These measurements are collected from a hardware-in-the-loop testbed setup equipped with a digital real-time simulator. Using DWT, the detail coefficients of measurements are obtained. Next, the decomposed data is then fed into the CAE that captures the underlying structure of the transformed data. Anomalies are identified when significant errors are detected between input samples and their reconstructed outputs. We demonstrate our approach on the IEEE-14 bus system considering different events such as generator faults, line-to-line faults, line-to-ground faults, load shedding, and line outages simulated on a real-time digital simulator (RTDS). The proposed implementation achieves a classification accuracy of 97.7%, precision of 98.0%, recall of 99.5%, F1 Score of 98.7%, and proves to be efficient in both time and space requirements compared to baseline approaches. 

Impacting drops are ubiquitous and the corresponding impact force is their most studied dynamic quantity. However, impact forces arising from collisions with curved surfaces are understudied. In this study, we impact small cups with falling drops across drop Reynolds number 2975–12 800, isolating five dominant parameters influencing impact force: drop height and diameter, surface curvature and wettability, and impact eccentricity. These parameters are effectively continuous in their domain and have stochastic variability. The unpredictable dynamics of the system incentivize the implementation of tools that can unearth relationships between parameters and make predictions about impact force for parameter values for which there is not explicit experimental data. We predict force due to the impacting drop in a concave target using an ensemble learning algorithm comprised of four base algorithms: a random forest regressor, k-nearest neighbor, a gradient boosting regressor, and a multi-layer perceptron. We train and test our algorithm with original experimental data comprising 387 total trials using four cup radii with two wetting conditions each. Our approach permits the determination of relative importance of the input features in producing impact force and force predictions which can be compared to scaling relations modified from those for flat targets. Algorithmic predictions indicate that deformation of the drop and surface wettability, often neglected in scaling for impact force on flat surfaces, are important for concave targets. Finally, our approach provides another opportunity for the application of machine learning to characterize complex systems' fluid mechanics for which experimental variables are numerous and vary independently.