This paper proposes a collaborative machine learning-based framework to detect cyber-attacks in a power system, leading to deviation in the state variable behavior. Based on the proposed architecture, three different machine learning-based methods, i.e., visualization, classification, and clustering, are employed and compared to find the best one in the FDIA detection process. To this end, pre-processing is employed in the first stage. In the second stage, the patterns of the state vectors are transferred into features. Hence, 24 statistical features, including measures of central tendency, variability, measures of shape, and position, are extracted to find various properties. Then, in the third stage, a supervised algorithm is employed to rank and find the most crucial features in FDIA. In the fourth stage, an unsupervised dimensionality reduction technique (PCA) is applied to reduce the feature space. In the fifth and last stage, visualization, classification, and clustering-based methods are developed to detect FDIA. To simulate an attack, it is assumed that an intruder decreases or increases the state vectors at different buses with various attack parameters (i.e., 0.90, 0.95, 0.96, 0.97, 0.98, 1, 1.02, 1.03, 1.04, 1.05, and 1.10). The proposed method effectiveness is assessed on the New York Independent System Operator (NYISO) data applied to the IEEE 14-bus system. The results presented in the paper from different scenarios (i.e., phase angle (θ), voltage magnitude (Vm), measurements, and multiple attacks) on a real-world dataset demonstrate that the collaborative optimized PCA-Density-based machine learning technique can detect most of the attack samples with good performance scores (i.e., recall, precision, F1) and outperforms the other investigated methods. Moreover, it is general and adaptable enough to cover the situation where either the system characteristics or the attack behavior changes.

Electrical load forecasting is one of the critical tasks that helps power utility companies in planning and operation as well as the energy managementsystem (EMS) in controlling and optimizing the power grid’s performance. It guarantees system adequacy and reliability by reducing the gap between supply and demand, and it bolsters cybersecurity by enabling false data detection and system security by predicting system peak load. This article aims to build an accurate long-term load forecasting model based on deep learning (DL) algorithms. Initially, big-data analytics and feature engineering are applied to the power load and weather historical data. Next, three different load forecasting models are developed based on three corresponding deep neural network models (DNN): the simple RNN, the long short term memory (LSTM), and the gated recurrent unit (GRU), respectively. To find an accurate DNN model, a two-step hyperparameter optimization is performed consisting of cross-validation. In the first step, the random search (RS) algorithm is employed to narrow the parameter search space to a smaller local one. In the second step, the grid search (GS) algorithm is used to find the best hyperparameters in the local area located by RS. However, the computational burden of the DNN training process is too high and may lead to out-of-memory (OOM) issues, especially when we are dealing with a big volume of data. To overcome this problem, we developed our Python program based on Tensorflow framework and implemented the framework on a high-performance computer cluster (HPCC). Then, different jobs are defined to employ parallel distributed computing (PDC) capability on the SIU BigDawg HPCC. To this end, all DNN hyperparameters (e.g., number of layers, neurons, dropout, batch size, epoch, learning rate, etc.) can be optimized and result in a comprehensive and accurate load forecasting model. Our proposed method, along with the programming framework, is evaluated on a real-world dataset. Among the three models developed and tested, the GRU, with load and weather as features, has demonstrated excellent performance and outperformed the other two models for long-term forecasting with the best metrics (i.e., MSE, MAPE, $R^2$ ) and daily error almost less than 1%. Also, our best model (optimized GRU) is compared with the other reported state-of-the-art models such as DBN, SVR, RM-LSTM, DM-GCNN, TCMS-CNN. It is demonstrated that the proposed framework yields a superior performance than the others.

The paper presents a novel approach for treating the harmonic torque in permanent magnet motors at selected harmonic orders. The method is based on the online estimation of the motor harmonic flux. For this purpose, a novel Minimal Parameter Harmonic Flux Estimator (MPHFE) is developed. The latter is formulated such that the inductance, resistance, and stator current and its derivative are not necessary for the estimation of the harmonic flux. This was achieved by forcing the harmonic current induced by the harmonic flux component to zero through the combined action of a field-oriented controller (FOC) and a feed-forward controller. Subsequently, the harmonic flux can be obtained directly from the estimated harmonic back-EMF without the involvement of other motor parameters. Finally, the estimated flux is used in conjunction with a comprehensive analysis of the motor harmonic torque to determine the stator current compensation to eliminate the torque harmonic. Experimental results presented from a permanent magnet motor confirm the theoretical claims in the paper as well as demonstrate the effectiveness of the proposed method to reduce the magnitude of the harmonic torque significantly.

A two-level optimization method is presented to find the optimal number and location of conventional protective devices to be upgraded to remote-controlled switches (RCSs) for an existing distribution network (DN). The effect of distributed generation (DG) on this problem is considered. In the first level, a nonlinear binary program is proposed to maximize the restored customers subject to technical and topological constraints. All feasible interchanges between protective devices and ties involved in the restoration, when a fault occurs at all possible locations are found considering switching dependencies. In the second level, a nonlinear cost function, combining the expected cost of interruptions (ECOST) and the switch cost, is minimized with respect to the location of RCSs. The expected cost function is computed based on the optimum restoration policies obtained from the first level. The optimum placement of RCSs using the proposed algorithm is tested on a 4-feeder 1069-node test system and compared to the solution obtained with a genetic algorithm (GA) on the same system.

Monitoring of abnormal events in a distribution feeder by using a single technique is a challenging task. A number of abnormal events can cause unsafe operation, including a high impedance fault (HIF), a partial breakdown to a cable insulation, and a circuit breaker (CB) malfunction due to capacitor bank de-energization. These abnormal events are not detectable by conventional protection schemes. In this paper, a new technique to identify distribution feeder events is proposed based on the complex Morlet wavelet (CMW) and on a decision tree (DT) classifier. First, the event is detected using CMW. Subsequently, a DT using event signatures classifies the event as normal operation, continuous and non-continuous arcing events (C.A.E. and N.C.A.E.). Additional information from the supervisory control and data acquisition (SCADA) can be used to precisely identify the event. The proposed method is meticulously tested on the IEEE 13- and IEEE 34-bus systems and has shown to correctly classify those events. Furthermore, the proposed method is capable of detecting very high impedance incipient faults (IFs) and CB restrikes at the substation level with relatively short detection time. The proposed method uses only current measurements at a low sampling rate of 1440 Hz yielding an improvement of existing methods that require much higher sampling rates.

This article proposes a real-time multistage data-driven method to detect false data injected by an intruder. The proposed method is based on an accurate load model constructed by deep neural networks (long short-term memory and gated recurrent unit) where historical demand along with extra weather features are employed to formulate the problem as a sequence of predictions and build an appropriate DNN model. A distinct feature of the proposed method that dramatically improves accuracy is hyperparameter optimization (e.g., dropout rate, batch size, $\beta _{1}$ , $\beta _{2}$ , $\alpha$ ) based on cross-validation. Since hyperparameter optimization consumes enormous resources to analyze the different combinations of hyperparameters and causes out-of-memory problems, we have implemented our framework based on a parallel distributed computing technique on high-performance computer clusters. A variety of case studies are examined to find the best model. At the end of this step, the best model is employed as a base reference to a semisupervised scoring algorithm for the purpose of finding sequence of injected false data. To improve the detection efficiency, inter-range, pruning, and minimum score are added to the proposed methodology that suppress high false-positive rates. Several scenarios from a real-world dataset have been studied in this article, and the results demonstrate that the proposed method can be used to detect various types of attacks, including scaling, ramping, professional ramping, and random attacks, with a good performance score (i.e., recall, precision, and $F1$ score) for all of them. Furthermore, it can mitigate false data with the forecasted load values.