The computational cost of topology optimization based on binary particle swarm optimization is greatly reduced by the use of Deep Neural Networks (DNN). A first Convolutional Neural Network (CNN) is trained with data coming from Finite Element Analysis (FEA) with the aim of correctly estimating the output quantity (a motor torque in the proposed case study). A second CNN is trained to give as output the boundary conditions (BC, expressed in terms of fields or potentials) to be used as BC of a reduced Finite Element Method (FEM) model, created in order to still be able to give the correct value of the output quantity. In the optimization phase, the torque properties are evaluated by the trained CNN, and only a reduced percentage of cases are re-evaluated by either the full FEM model or the reduced FEM model. The overall computational time of the optimization procedure is significantly reduced.

In this work a topology optimization procedure is proposed and applied to the TEAM 25 problem, i.e. a model of a die press with electromagnet for orientation of magnetic powder. The shape of the press is described as a free discretized profile, and its relation to the flux density in the cavity is simulated by Finite Element Analysis (FEA) and learned by a Deep Neural Network (DNN) model. The DNN is used as a surrogate model for optimization, aiming to obtain a desired flux density distribution in the cavity. Results are promising, as better accuracy is obtained with respect to full FEA-based optimization approach with reduced time cost. Once trained, surrogate model can be used to efficiently solve a whole family of problems where a different target field distribution is defined.

Many centralized and distributed power sharing algorithms have been proposed in the literature for de-loading operations in wind farms with variable speed wind turbines. Typically, in these strategies, two-way communications are required between the control center and the single turbines, or among the turbines. This paper solves the same problem in a truly decentralized fashion, which only requires a greatly reduced amount of one-way communications, without exchanging information among the turbines, and shows that an optimal solution can be obtained to minimize utility functions of interest of single wind turbines. In particular, we consider utility functions that take into account mechanical fluctuations and rotor over-speeds during transient, while balancing the utilization of wind turbines at steady-state operations. This is achieved by adopting the so-called Additive Increase Multiplicative Decrease (AIMD) algorithms, which are frequently used in communication applications, for solving the power sharing problem in a decentralized fashion. Extensive simulations under different working conditions, on wind farms consisting of wind turbines of different mechanical characteristics, are provided to illustrate the potential and the efficiency of the proposed methodology.

In this paper, a novel and flexible solution for fault prediction based on data collected from Supervisory Control and Data Acquisition (SCADA) system is presented. Generic fault/status prediction is offered by means of a data driven approach based on a self-organizing map (SOM) and the definition of an original Key Performance Indicator (KPI). The model has been assessed on a park of three photovoltaic (PV) plants with installed capacity up to 10 MW, and on more than sixty inverter modules of three different technology brands. The results indicate that the proposed method is effective in predicting incipient generic faults in average up to 7 days in advance with true positives rate up to 95%. The model is easily deployable for on-line monitoring of anomalies on new PV plants and technologies, requiring only the availability of historical SCADA data, fault taxonomy and inverter electrical datasheet.

The use of behavioral models based on deep learning (DL) to accelerate electromagnetic field computations has recently been proposed to solve complex electromagnetic problems. Such problems usually require time-consuming numerical analysis, while DL allows achieving the topologically optimized design of electromagnetic devices using desktop class computers and reasonable computation times. An unparametrized bitmap representation of the geometries to be optimized, which is a highly desirable feature needed to discover completely new solutions, is perfectly managed by DL models. On the other hand, optimization algorithms do not easily cope with high dimensional input data, particularly because it is difficult to enforce the searched solutions as feasible and make them belong to expected manifolds. In this work, we propose the use of a variational autoencoder as a data regularization/augmentation tool in the context of topology optimization. The optimization was carried out using a gradient descent algorithm, and the DL neural network was used as a surrogate model to accelerate the resolution of single trial cases in the due course of optimization. The variational autoencoder and the surrogate model were simultaneously trained in a multi-model custom training loop that minimizes total loss—which is the combination of the two models’ losses. In this paper, using the TEAM 25 problem (a benchmark problem for the assessment of electromagnetic numerical field analysis) as a test bench, we will provide a comparison between the computational times and design quality for a “classical” approach and the DL-based approach. Preliminary results show that the variational autoencoder manages regularizing the resolution process and transforms a constrained optimization into an unconstrained one, improving both the quality of the final solution and the performance of the resolution process.