Abstract Effective control of Energy Storage Systems (ESS) is crucial for the secure and profitable operation of microgrids. In this context, ESSs are essential for enhancing the overall grid resilience, balancing supply, and mitigating voltage and frequency variations. This paper presents a novel neuroevolutionary method, coupling a modified version of the Multi-Objective Evolutionary Policy Search (MEPS) algorithm with the Cross-Entropy method, aimed at optimizing an ESS control problem. The modified MEPS, named Cascade-MEPS, employs a cascade weights mutation operator to refine policies by focusing on the most recent hidden node, ensuring localized and non-disruptive adjustments. The resulting algorithm, referred to as cross-entropy Cascade-MEPS (CE-CMEPS), utilizes the cross-entropy method as a depth initialization strategy, conducting an initial exploration of the weights space to initialize the population prior to Cascade-MEPS execution. Experimental validation on a newly proposed multi-objective ESS control problem demonstrates the efficacy of CE-CMEPS, showcasing performance improvements and reduced variation compared to standalone MEPS. Our results show that CE-CMEPS is an effective ESS discharge controller and a sustainable multi-objective reinforcement learning solution.

Abstract This paper presents a novel implementation of the Coral Reef Optimization with Substrate Layers (CRO‐SL) algorithm. Our approach, which we call TensorCRO, takes advantage of the TensorFlow framework to represent CRO‐SL as a series of tensor operations, allowing it to run on GPU and search for solutions in a faster and more efficient way. We evaluate the performance of the proposed implementation across a wide range of benchmark functions commonly used in optimization research (such as the Rastrigin, Rosenbrock, Ackley, and Griewank functions), and we show that GPU execution leads to considerable speedups when compared to its CPU counterpart. Then, when comparing TensorCRO to other state‐of‐the‐art optimization algorithms (such as the Genetic Algorithm, Simulated Annealing, and Particle Swarm Optimization), the results show that TensorCRO can achieve better convergence rates and solutions than other algorithms within a fixed execution time, given that the fitness functions are also implemented on TensorFlow. Furthermore, we also evaluate the proposed approach in a real‐world problem of optimizing power production in wind farms by selecting the locations of turbines; in every evaluated scenario, TensorCRO outperformed the other meta‐heuristics and achieved solutions close to the best known in the literature. Overall, our implementation of the CRO‐SL algorithm in TensorFlow GPU provides a new, fast, and efficient approach to solving optimization problems, and we believe that the proposed implementation has significant potential to be applied in various domains, such as engineering, finance, and machine learning, where optimization is often used to solve complex problems. Furthermore, we propose that this implementation can be used to optimize models that cannot propagate an error gradient, which is an excellent choice for non‐gradient‐based optimizers.

Microgrids have become popular candidates for integrating diverse energy sources into the power grid as means of reducing fossil fuel usage. Energy Resource Management (ERM) is a type of Unit Commitment problem, where a player operates a microgrid with diverse renewable generators integrated with an external supplier. Calculating the economic dispatch of each committed unit on a planning horizon is an NP-hard problem, and therefore, finding an exact solution is difficult. This paper presents a multi-objective solution to the ERM problem from the perspective of battery operation and external supplier dispatch. First, a novel multi-objective decision problem modeling is proposed that considers three objectives: cost, greenhouse gas emissions, and battery degradation. This framework involves a learning agent that controls the depth of discharge of a Lithium-Ion battery. To address the proposed problem, a new multi-objective algorithm called Multi-Objective Evolutionary Policy Search (MEPS) is introduced. The proposed algorithm uses NeuroEvolution of Augmenting Topologies structure to evolve artificial neural networks for estimating action-preference values considering multi-objective rewards. The MEPS performance is evaluated on both standard and newly-proposed benchmark problems, using the hypervolume as the evaluation metric. When compared to standard deep reinforcement learning, results showed that MEPS provides cost-effective, environmentally friendly, and efficient energy storage management solutions. Furthermore, MEPS effectively solves the proposed ERM problem by finding neural networks with a small number of nodes and connections, which are suitable for use in embedded control systems. Overall, MEPS proved to be a promising multi-objective approach in the transition to clean energy resources.

In this work we present a pattern classification approach coupling the Neighbourhood Component Analysis (NCA) classifier with the Canonical Differential Evolutionary Particle Swarm Optimization (C-DEEPSO). The standard NCA uses the conjugate gradient method to minimize the classification error. Here we propose an approach using the C-DEEPSO instead. In the experimental design, the coupled approach is applied to 20 benchmark data sets, and its performance is compared with the standard NCA using the conjugate gradient. The experimental analysis shows the usage of an evolutionary approach to enhance the performance of a machine learning algorithm can be competitive when compared to well-known iterative optimization techniques, and even outperform them in some problems. A real-world problem classifying cyber-attacks to an industrial control system of gas pipelines is also solved by the proposed approach. The results obtained indicate the proposed approach can successfully identify possible cyber-attacks to the control system. In this way, the NCA coupled to C-DEEPSO can work as an Intrusion Detection Systems (IDS), being able to guarantee an acceptable security level.

The Wilkie, Stonham, and Aleksander recognition device (WiSARD) [Formula: see text]-tuple classifier is a multiclass weightless neural network capable of learning a given pattern in a single step. Its architecture is determined by the number of classes it should discriminate. A target class is represented by a structure called a discriminator, which is composed of [Formula: see text] RAM nodes, each of them addressed by an [Formula: see text]-tuple. Previous studies were carried out in order to mitigate an important problem of the WiSARD [Formula: see text]-tuple classifier: having its RAM nodes saturated when trained by a large data set. Finding the VC dimension of the WiSARD [Formula: see text]-tuple classifier was one of those studies. Although no exact value was found, tight bounds were discovered. Later, the bleaching technique was proposed as a means to avoid saturation. Recent empirical results with the bleaching extension showed that the WiSARD [Formula: see text]-tuple classifier can achieve high accuracies with low variance in a great range of tasks. Theoretical studies had not been conducted with that extension previously. This work presents the exact VC dimension of the basic two-class WiSARD [Formula: see text]-tuple classifier, which is linearly proportional to the number of RAM nodes belonging to a discriminator, and exponentially to their addressing tuple length, precisely [Formula: see text]. The exact VC dimension of the bleaching extension to the WiSARD [Formula: see text]-tuple classifier, whose value is the same as that of the basic model, is also produced. Such a result confirms that the bleaching technique is indeed an enhancement to the basic WiSARD [Formula: see text]-tuple classifier as it does no harm to the generalization capability of the original paradigm.

In this paper, a hybrid single-objective metaheuristic, named as C-DEEPSO (Canonical Differential Evolutionary Particle Swarm Optimization), is proposed to solve large-scale optimization problems. C-DEEPSO can be viewed as an evolutionary algorithm with recombination rules borrowed from PSO or an swarm optimization method with selection and self-adaptiveness properties. To assess the algorithm performance, the algorithm is run over 15 benchmark continuous problems presented in CEC'2015. The algorithm is also applied over a real world large-scale problem. The results indicate that the proposed algorithm is an efficient and competitive method to handle such problems. The experimental results also show that the new approach reaches competitive results when compared to the reference algorithm DECC-G. An application of C-DEEPSO were perform to solve the electric dispatch in a large scale energy network, the IEEE 57 Bus-System. The results show that the algorithm is a good way to solve nonlinear problems respecting many constraints associated.

In this paper, we extend the risk zone concept by creating the Generalized Risk Zone. The Generalized Risk Zone is a model-independent scheme to select key observations in a sample set. The observations belonging to the Generalized Risk Zone have shown comparable, in some experiments even better, classification performance when compared to the use of the whole sample. The main tool that allows this extension is the Cauchy-Schwartz divergence, used as a measure of dissimilarity between probability densities. To overcome the setback concerning pdf's estimation, we used the ideas provided by the Information Theoretic Learning, allowing the calculation to be performed on the available observations only. We used the proposed methodology with Learning Vector Quantization, feedforward Neural Networks, Support Vector Machines, and Nearest Neighbors.

Acute lymphoblastic leukemia (ALL), the most common cancer in childhood, has its treatment modulated by the risk of relapse. An appropriate estimation of this risk is the most important factor for the definition of treatment strategy. In this paper, we build up a new decision support tool to improve treatment intensity choice in childhood ALL. Our procedure was applied to a significant cohort of Brazilian children with ALL, the majority of the cases treated in the last decade in the two main University Hospitals of Rio de Janeiro. Some intrinsically difficulties of this dataset introduce an assortment of challenges, among those the need of a proper selection of features, clinical and laboratorial data. We apply a mutual information-based methodology for this purpose and a Neural Network to estimate the risk. Among the relapsed patients, 98.2% would have been identified as high-risk by the proposed methodology. The proposed procedure showed significantly better results when compared to the BFM95, a widely used classification protocol.

In this paper we present a data selection methodology for vector quantization. The main goal is to identify, and possibly eliminate, noisy data in a supervised pattern classification context. We consider as 'noise' an inversion (in a two classes problem) of the class the data belongs to. The methodology is based on two mappings that bring the data selection problem into a R2decision space, independently of the data dimension. The proposed technique demands relatively low computational effort and is model independent. Numerical experiment has shown interesting performance enhancing the potentiality of the method.

In this paper, we propose a method that selects a subset of the training data points to update LVQ prototypes. The main goal is to conduct the prototypes to converge at a more convenient location, diminishing misclassification errors. The method selects an update set composed by a subset of points considered to be at the risk of being captured by another class prototype. We associate the proposed methodology to a weighted norm, instead of the Euclidean, in order to establish different levels of relevance for the input attributes. The technique was implemented on a controlled experiment and on Web available data sets.

In this paper we present an application of data and input selection to a heart disease diagnosis problem. We approach the problem by using a modified LVQ scheme that selects a subset of the training data points to update the prototypes. The main model goal is to identify patients with relevant coronary vessels obstruction. The selected subset provides an interesting interpretation. We associate this methodology with a weighted norm, instead of the Euclidean, in order to establish different levels of importance for the input attributes. Again, interesting interpretation arises concerning the relevance of the input attributes.

The goal of this paper is to test and model nonlinearities in several monthly exchange rates time series. We apply two different nonlinear alternatives, namely: the artificial neural-network time series model estimated with Bayesian regularization; and a flexible smooth transition specification, called the neuro-coefficient smooth transition autoregression. The linearity test rejects the null hypothesis of linearity in 10 out of 14 series. We compare, using different measures, the forecasting performance of the nonlinear specifications with the linear autoregression and the random walk models.

This paper proposes a hybrid forecasting system that combines linear models and multilayer neural networks to forecast hourly temperatures based on the past observed temperatures and the maximum and minimum forecast temperatures supplied by the weather service. First, a simple autoregressive model is built for the temperature series. Numerical results, using real data is reported.

In this paper, a new method for calculation of Zech's logarithm in GF(2/sup m/) is presented. For a given element, the logarithm is calculated by bit operations performed on its binary representation. No look-up tables are used. The proposed method makes feasible the implementation of an universal-type device for finite field arithmetic.

This paper proposes a new methodology to approximate functions by incorporating a priori information. The relationship between the proposed scheme and multilayer neural networks is explored theoretically and numerically. This approach is particularly interesting for the very relevant class of limited spectrum functions. The number of free parameters is smaller if compared to Back-Propagation Algorithm opening the way for better generalization results.

A new method for identifying multiplicative parameters (i.e. the matrix-valued coefficients of the multiplicative state feedback) appearing in discrete-time bilinear models is proposed. The system is supposed to operate in a stochastic environment where the noisy output is the only sequence available for measurements. Ergodicity is not assumed, and the probability distributions involved may be arbitrary and unknown. A correlative identification technique is presented which uses stochastic approximation algorithms for estimating (with probability one) the unknown parameters. Simulated results are included.