Optical communication is an integral part of the modern economy, having all but replaced electronic communication systems. Future growth in bandwidth appears to be on the horizon using structured light, encoding information into the spatial modes of light, and transmitting them down fibre and free-space, the latter crucial for addressing last mile and digitally disconnected communities. Unfortunately, patterns of light are easily distorted, and in the case of free-space optical communication, turbulence is a significant barrier. Here we review recent progress in structured light in turbulence, first with a tutorial style summary of the core concepts, before highlighting the present state-of-the-art in the field. We support our review with new experimental studies that reveal which types of structured light are best in turbulence, the behaviour of vector versus scalar light in turbulence, the trade-off of diversity and multiplexing, and how turbulence models can be exploited for enhanced optical signal processing protocols. This comprehensive treatise will be invaluable to the large communities interested in free-space optical communication with spatial modes of light.

Electricity theft is a global problem that negatively affects both utility companies and electricity users. It destabilizes the economic development of utility companies, causes electric hazards and impacts the high cost of energy for users. The development of smart grids plays an important role in electricity theft detection since they generate massive data that includes customer consumption data which, through machine learning and deep learning techniques, can be utilized to detect electricity theft. This paper introduces the theft detection method which uses comprehensive features in time and frequency domains in a deep neural network-based classification approach. We address dataset weaknesses such as missing data and class imbalance problems through data interpolation and synthetic data generation processes. We analyze and compare the contribution of features from both time and frequency domains, run experiments in combined and reduced feature space using principal component analysis and finally incorporate minimum redundancy maximum relevance scheme for validating the most important features. We improve the electricity theft detection performance by optimizing hyperparameters using a Bayesian optimizer and we employ an adaptive moment estimation optimizer to carry out experiments using different values of key parameters to determine the optimal settings that achieve the best accuracy. Lastly, we show the competitiveness of our method in comparison with other methods evaluated on the same dataset. On validation, we obtained 97% area under the curve (AUC), which is 1% higher than the best AUC in existing works, and 91.8% accuracy, which is the second-best on the benchmark.

Vast geographical distances in Africa are a leading cause for the so-called “digital divide” due to the high cost of installing fibre. Free-Space Optical (FSO) communications offer a convenient and higher bandwidth alternative to point-to-point radio microwave links, with the possibility of re-purposing existing infrastructure. Unfortunately, the range of high bandwidth FSO remains limited. While there has been extensive research into an optimal mode set for FSO to achieve maximum data throughput by mode division multiplexing, there has been relatively little work investigating optical modes to improve the resilience of FSO links. Here we experimentally show that a carefully chosen subset of Hermite-Gaussian modes is more resilient to atmospheric turbulence than similar Laguerre-Gauss beams, with a predicted upper bound increase in propagation distance of 167% at a mode dependent loss of 50%.

The optimal integration of electrical units, such as distributed generation units, power electronic devices, and electric vehicles, is a significant development of smart grids. This development has effectively transformed the traditional grid system, promising numerous advantages for economic values and autonomous energy source control. In smart grids development, metaheuristic algorithms are one of the optimization algorithms that have been applied extensively to mitigate the accompanying problems such as voltage instability, power loss, and high installation cost. This paper presents a comprehensive review of metaheuristic techniques for the optimal integration of electrical units in distribution networks, considering different phases of the optimization process. These include the approaches for handling of crucial objective functions and the optimal integration methods for different electrical units. This review shows a need for more research on developing efficient metaheuristic algorithms and the effective handling of multiple objective functions.

Electric vehicles (EVs) are crucial to the reduction of carbon emissions. However, their charging poses a threat to power system networks. Hence, EV charging control strategies are developed to curb this challenge, using charging prices to incentivize EV drivers to choose EV charging stations (EVCS) favourable to the grid's stability. The challenge of this strategy is the likelihood of EV drivers accepting EVCS suggestions. To increase the probability of accepting EVCS suggestions, we introduce a two-tailed incentive pricing (TTIP) scheme in an EV charging coordination model, where incentives are offered as charging prices and parking time. We formalized the EV charging problem as a multiobjective Markov decision process and proposed a deep deterministic policy gradient (DDPG) to solve it. To tackle the challenge of continuous action space that leads to the dimensionality curse, the proposed DDPG models the action space using a metaheuristic-based technique. The proposed scheme implements a multiple reward system to generate Pareto optimal solutions and a decision-making technique to choose the compromise reward. Using real-world electricity prices and the IEEE 33-bus distribution network, numerical simulations show that our proposed TTIP scheme yields an average of 18% improvement in grid stability than the sustainable policy following, random, and price-greedy algorithms. It also improves the EV charging profit margins by an average of 28%.

Although there has been a rapid development of practical applications, theoretical explanations of deep learning are in their infancy. Deep learning performs a sophisticated coarse graining. Since coarse graining is a key ingredient of the renormalization group (RG), RG may provide a useful theoretical framework directly relevant to deep learning. In this study we pursue this possibility. A statistical mechanics model for a magnet, the Ising model, is used to train an unsupervised restricted Boltzmann machine (RBM). The patterns generated by the trained RBM are compared to the configurations generated through an RG treatment of the Ising model. Although we are motivated by the connection between deep learning and RG flow, in this study we focus mainly on comparing a single layer of a deep network to a single step in the RG flow. We argue that correlation functions between hidden and visible neurons are capable of diagnosing RG-like coarse graining. Numerical experiments show the presence of RG-like patterns in correlators computed using the trained RBMs. The observables we consider are also able to exhibit important differences between RG and deep learning.