Siby Jose Plathottam

Transient loss minimization in induction machine drives using optimal control theory

Loss minimization during transient operation is an often neglected in most induction machine applications. However, loss minimization can greatly improve efficiency in case of fast changing load torques or reference speed profiles. Optimal control theory has been successfully applied previously to develop control laws that minimize losses during transients. All previous works were based on the field oriented induction machine model. In this paper, however, an optimal control problem based on the non-field oriented induction machine model is developed and simulated. Additionally, a more comprehensive cost functional is developed and utilized that would guarantee a stable steady state operation of the machine. Offline optimal control histories are generated using the conjugate gradient method. The performance of the proposed control law is verified through simulation and the results are compared with those from an indirect field oriented controller.

Induction machine transient energy loss minimization using neural networks

Reducing losses in an induction machine during transient periods is highly desirable in applications where the torque-speed operating point is continuously changing. Optimal control theory can be used to find control trajectories that minimize the cost function defining the losses. However, solving optimal control problems is generally impractical to realize in real time, especially if the underlying system dynamics are nonlinear. In this work, a feed forward neural network is trained so as to emulate the solution of the loss minimizing optimal control problem in real time. Simulation results using a 6th order model are presented comparing the performance of the proposed neural network control with that of standard field oriented control. It is found that transient energy losses are reduced.

Sub kW Wind Turbine Emulator (WTE)

Wind power is among the fastest growing renewable sources of energy. A Wind Turbine Emulator (WTE) is a tool that emulates an actual wind turbine and enables the study of different control topologies for electric power generators. This paper presents the development of a sub kW WTE and demonstrates its performance under a highly fluctuating wind profile. The laboratory prototype consists of a permanent magnet dc (PMDC) motor, controlled by a DC-DC converter, coupled with an induction generator that converts the dc motor generated torque to electrical power. System performance and results are discussed.

Convolutional Neural Networks (CNNs) for power system big data analysis

The concept of automated power system data analysis using Deep Neural Networks (as part of the routine tasks normally performed by Independent System Operators) is explored and developed in this paper. Specifically, we propose to use the widely-used Deep neural network architecture known as Convolutional Neural Networks (CNNs). To this end, a 2-D representation of power system data is developed and proposed. To show the relevance of the proposed concept, a multi-class multi-label classification problem is presented as an application example. Midcontinent ISO (MISO) data sets on wind power and load is used for this purpose. TensorFlow, an open source machine learning platform is used to construct the CNN and train the network. The results are discussed and compared with those from standard Feed Forward Networks for the same data.

Short-term load forecasting using deep neural networks (DNN)

Load forecasting is an important electric utility task for planning resources in Smart grid. This function also aids in predicting the behavior of energy systems in reducing dynamic uncertainties. The efficiency of the entire grid operation depends on accurate load forecasting. This paper proposes and investigates the application of a multi-layered deep neural network to the Iberian electric market (MIBEL) forecasting task. Ninety days of energy demand data are used to train the proposed model. The ninety-day period is treated as a historical dataset to train and predict the demand for day-ahead markets. The network structure is implemented using Googles machine learning Tensor-flow platform. Various combinations of activation functions were tested to achieve a better Mean Absolute percentage error (MAPE) considering the weekday and weekend variations. The tested functions include Sigmoid, Rectifier linear unit (ReLU), and Exponential linear unit (ELU). The preliminary results are promising. and show significant savings in the MAPE values using the ELU function over the other activation functions.

Trajectory training of feedforward neural networks for DC motor speed control

This work discusses the use of template trained neural networks for DC motor speed control. It proposes the use of input-output trajectories to train the neural network instead of using data points as is normally done for feedforward neural networks. Experimental results show that the trajectory trained neural network can successfully perform speed control of PMDC motors when compared to neural networks trained using conventional training methods.

Evaluating the transient stability impact of inertia less renewable generation

An analysis of the impact on transient stability of adding renewable generation capacity to a power system is performed and new approach to calculating the stability limit is proposed. Wind generation is being treated as inertia-less generation source. The results show that addition of wind generation has dissimilar impact depending on the transient event occurring in the system.

Equitable wind power scheduling in power system with multiple wind farms

Increasing penetration of wind power into the generation mix has necessitated the integration of wind farms in the daily planning and dispatching operations of independent system operators. This paper models a simple and intuitive cost function that can dispatch power among multiple wind farms in an economically equitable manner. The cost function is used in an optimal power flow problem for the IEEE 14 bus system with three wind farms. The results indicate that the proposed cost function can prevent the underutilization of certain wind farms due to their location while satisfying the system operating constraints.