8x8-Helical Antenna Array WPT Beamforming Using the Deep Learning Method

The CASSIOPeiA array utilises a helical array to beam steer evenly and constantly across the azimuth plane using fewer elements than a circular array. This property is beneficial for constructing a large-scale, 360-degree steerable antenna array without compromising performance. One of the challenges of this large-scale antenna array is the individual elements phase adjustment to provide optimal beam steering. A full-sized CASSIOPeiA antenna array on a space solar satellite can include up to two billion antenna elements, requiring extensive data to support beam steering. However, using traditional methods such as lookup table (LUT) to predict beam steering will take a relatively long time and require large storage space. Compared with LUT, the Deep Neural Networks (DNN) learning model reported in this paper, has the advantage of using a smaller storage size and accounts for mutual coupling. In this paper, a deep-learning model is trained to control an 8x8 Helical array for WPT. The training data is based on the array simulation results from CST microwave studio. The model can perform accurate beam steering with 0.1-degree accuracy. In the simulation stage, the model is trained using the Long Short Term Memory (LSTM) model. The data is extracted from MATLAB and CST, then, the model results are verified in CST simulation.