Non-Intrusive Load Monitoring (NILM) is a promising approach for energy disaggregation, enabling the identification of appliance-specific energy consumption from aggregated electrical signals. This study proposes a fast Sequence-to-Sequence (S2S) architecture that integrates a Convolutional Variational Auto-Encoder (CVAE) and a transformer model for accurate appliance disaggregation. Performing the reconstruction based on the CVAE is the key element for improving the S2S disaggregation model. The framework is validated on a residential dataset (UK-DALE) and a laboratory dataset collected in a university setting, which includes three distinct appliances: a laser cutting machine, a dust and fume extractor, and three 3D printers. Experimental results demonstrate the efficiency and accuracy of the proposed method in diverse environments, proving its potential for real-world applications in NILM.

The non-deterministic home appliances’ behaviour makes aggregated power consumption hard to be explored and to identify individual appliances’ consumption (disaggregation) in residential buildings. This paper presents a deep neural network learning scheme in order to disaggregate a main meter’s aggregated signal into 11 appliances’ signals and estimate their individual power consumption. A 1-Dimensional Convolution Neural Network (1D CNN) and Long Short-Term Memory (LSTM) layers are used together to form a sequence-to-point (S2P) and a sequence-to-sequence (S2S) Multi-Target Regressor (MTR) for learning and recognizing the loads. Our model is fed with the home total real power (P), total reactive power (Q) and total current (I) and outputs the disaggregated real power (P) for each appliance. The model was trained and evaluated on the AMPds2 public dataset which results in a global disaggregation accuracy of 93.27% for the S2P model and 87.79% for the S2S model. The S2P model outperforms the existing methods in terms of disaggregation accuracy and the number of disaggregated appliances (11 appliances instead of 9) on the used database.

A flatness-based control approach which is implemented in successive loops is used to solve the control problem for the multivariable and nonlinear dynamics of industrial robotic manipulators and autonomous vehicles. The state-space model of these robotic systems is separated into two subsystems, which are connected between them in cascading loops. Each one of these subsystems can be viewed independently as a differentially flat system and control about it can be performed with inversion of its dynamics as in the case of input-output linearized flat systems. The state variables of the second subsystem become virtual control inputs for the first subsystem. In turn exogenous control inputs are applied to the first subsystem. The whole control method is implemented in two successive loops and its global stability properties are also proven through Lyapunov stability analysis. The validity of the control method is confirmed in two case studies: (a) control of a 3-DOF industrial rigidlink robotic manipulator, (ii) control of a 3-DOF autonomous underwater vessel.

A nonlinear optimal (H-infinity) control method is developed for the Lotka-Volterra dynamical system. First, differential flatness properties are proven. The state-space description undergoes linearization, at each sampling instance, with the use of first-order Taylor series expansion and through the computation of the associated Jacobian matrices. Next, for the approximately linearized model of the system a stabilizing H-infinity feedback controller is designed. To compute the controller’s gains an algebraic Riccati equation has to be repetitively solved at each time-step of the control algorithm. Global stability properties are proven through Lyapunov analysis. Finally, the nonlinear optimal control method is compared against a flatness-based control approach implemented in successive loops.

This paper presents an improved current sensor fault-tolerant control for vector control of induction motors. This work aims to reconstruct the three current phases even under the failure of two or three current sensors and make the vector control in service under this situation. Therefore, the proposed scheme is made up of three blocks, vector control, sliding mode observers for the diagnosis process, and the reconfiguration algorithm. In the case of the failure of two or three current sensors, we will use the DC-link current sensor of the inverter to reconstruct the three phases currents. In addition, a synchronization technique named as PFCE (Positive Fundamental Component Estimator) is used to obtain results similar to real currents. Simulation results are presented discussing the studied approach.

Artificial intelligence technique based on fuzzy logic is increasingly used for the design of controller for maximum power point tracking (MPPT) in order to harvest maximum energy from photovoltaic (PV) generators. In this paper, a detailed analysis of the literature review of relevant works of fuzzy logic based MPPT algorithms for PV applications is elaborated first. Fuzzy Logic (FL) based MPPT techniques available in the literature are classified into three categories: An adaptive FL-based MPPT, FL combined with a classical technique (HC, InC, P&O) and FL combined with another intelligent technique (PSO, ANFIS, GA). Then a Fuzzy Logic Control (FLC) based MPPT algorithm for PV system is proposed. The FLC scheme design and rule table are presented in detail and based on the concept of asymmetric membership functions. The proposed scheme is implemented in real time using a dSPACE DS1104 board. From the experimental results, it is found that the proposed MPPT algorithm is simple, accurate and provides faster convergence to the maximum power point compared to the FL methods available in the literature.

Voltage-source inverter-fed Permanent Magnet Synchronous Machines are widely used in industry (for instance for the actuation of robotic and mechatronic systems, of cranes, in water pumping stations) as well as in transportation systems (for the traction of trains and electric vehicles). The present article proposes a nonlinear optimal control approach for voltage source inverter-fed Permanent Magnet Synchronous Machines (VSI-PMSMs). The nonlinear dynamic model of VSI-PMSMs undergoes approximate linearization around a temporary operating point which is recomputed at each iteration of the control method. This temporary operating point is defined by the present value of the voltage source inverter-fed PMSM state vector and by the last sampled value of the machine’s control inputs vector. The linearization relies on Taylor series expansion and on the calculation of the system’s Jacobian matrices. For the approximately linearized model of the voltage source inverter-fed PMSM an H-infinity feedback controller is designed. This controller stands for the solution of the nonlinear optimal control problem for the voltage source inverter-fed PMSM under model uncertainty and external perturbations. For the computation of the controller’s feedback gain an algebraic Riccati equation is iteratively solved at each time-step the control method. The global asymptotic stability properties of the control method are proven through Lyapunov analysis.

Open-loop synchronization techniques OLSs can be classified into True (TOLS) and Pseudo or Quasi-OLSs (POLS/QOLS). The TOLS means a system that has no feedback in its structures which results in unconditional stability. Roughly speaking the advanced TOLSs suffers from two main drawbakcks:1) compromising performances under off-nominal frequencies in faulty conditions, 2) inefficiency under large frequency drifts and high computation time. To tackle the problem of frequency adaptivity, the POLS uses a frequency estimator. This technique works efficiently under large frequency drifts even in faulty conditions and benefits from low computation time. However, one of the main challenges in OLS techniques is how to improve dynamic performance without compromising the ability to reject disturbances. In this paper, an enhancement is made to the standard POLS to tackle this problem. The use of a cascade positive fundamental components estimator (PFCE) will allow us to enhance the response time while keeping the phase and magnitude error at their lowest. The mathematical model is presented then simulated. The simulation results validate this proposal.

Fault diagnosis in energy conversion systems is performed with the use of neural networks and statistical decision making. An energy conversion system comprising a solar power unit, a DC-DC converter and a DC motor is considered and the related condition monitoring problem is solved. A neural network is used to model the dynamics of this energy conversion system after processing its input and output measurements, being accumulated at different operating conditions. The considered neural model is trained with the use of first-order gradient algorithms and consists of a hidden layer of Gauss-Hermite polynomial activation functions and of an output layer with linear weights. The neural network and the resulting model represents the fault-free functioning of the energy conversion system. At a next stage, the measurements of the real output of the energy conversion system are compared against the estimated outputs which are provided by the neural model. This provides, the residuals sequence. It holds that the sum of the squares of the residuals' vectors, multiplied with the inverse of the associated covariance matrix, stands for a stochastic variable (statistical test) which follows the χ2distribution. One can have a precise and almost infallible decision making tool about the appearance of faults in the energy conversion system, by selecting the 96% or the 98% confidence intervals of this distribution. When the upper or lower bound of the confidence interval are persistently exceeded one can conclude that the system has been subject to a fault. Finally, fault isolation can be also accomplished, by applying the statistical test into subspaces of the energy conversion system's state-space model.

1A nonlinear H-infinity (optimal) control approach is developed for the problem of the control of the spherical rolling robot. The solution of such a control problem is a nontrivial case due to underactuation and strong nonlinearities in the system's state-space description. The dynamic model of the robot undergoes approximate linearization around a temporary operating point which is recomputed at each timestep of the control method. The linearization relies on Taylor series expansion and on the computation of the system's Jacobian matrices. For the linearized dynamics of the spherical robot an H-infinity controller is designed. To compute the controller's feedback gains an algebraic Riccati equation in solved at each iteration of the control algorithm. The global asymptotic stability properties of the control method are proven through Lyapunov analysis. Finally, for the implementation of sensorless control for the spherical rolling robot, the H-infinity Kalman Filter is used as a robust state estimator.

A nonlinear optimal control approach is developed for the UAV and suspended load system. The dynamic model of the UAV and payload system undergoes approximate linearization. This makes use of Taylor series expansion around a temporary operating point which recomputed at each iteration of the control method. The linearization procedure relies on the computation of the Jacobian matrices of the state-space model of the system. Next, an H-infinity feedback controller is designed for the approximately linearized model. The proposed control method stands for the solution of the optimal control problem for the nonlinear and multivariable dynamics of the UAV and payload system, under model uncertainties and external perturbations. To compute the controller's feedback gains an algebraic Riccati equation is solved at each time-step of the control algorithm. The new nonlinear optimal control approach achieves fast and accurate tracking for all state variables of the UAV and payload system, under moderate variations of the control inputs. Finally, Lyapunov analysis is used to prove the global stability properties of the control scheme.

Control of the milling process of mining products (ore milling) is a non-trivial problem due to being related with a strongly nonlinear and multivariable state-space model. To provide an efficient solution to this problem, in this article a nonlinear optimal (H-infinity) control method is developed. In the considered nonlinear optimal control method, the dynamic model of the mining products' mill undergoes first approximate linearization with the use of Taylor series expansion and with the computation of the associated Jacobian matrices. The linearization point (temporary equilibrium) is recomputed at each time step of the control method and comprises the present value of the system's state vector and the last value of the control inputs' vector that was exerted on it. For the linearized description of the mill's functioning the optimal control problem is solved by applying an H-infinity controller. The feedback gain is computed again at each iteration of the control algorithm through the solution of an algebraic Riccati equation. The stability of the control scheme is confirmed through Lyapunov analysis. First, it is shown that the control method satisfies the H-infinity tracking performance, and this signifies elevated robustness against model uncertainty and external perturbations. Next, under moderate conditions, it is proven that the control loop is globally asymptotically stable.

An experimental comparative study of data compression algorithms is investigated for enhancing energy efficiency in low-powered smart meters. Data compression is able to reduce the RF communication time. We also propose a new lossless compression algorithm to achieve the best tradeoff between the compression ratio and computational costs. The performance of our proposed Run-Length Binary Encoding (RLBE) algorithm is compared to those obtained with other lossless compression algorithms: Huffman coding, Even-Rodeh, Exponential-Golomb, Lempel-Ziv Welch, Fibonacci coding, and the hybrid Bzip2 algorithm. The energy optimization in data transmission has been achieved under different operating conditions. The performance of each compression algorithms has been verified experimentally with real metering datasets from industrial and domestic cases.

A nonlinear optimal H-infinity control approach is proposed for bioreactors aiming at improved biofuels production. The dynamic model of the bioprocess taking place in the bioreactor undergoes approximate linearization round temporary equilibria which are recomputed at each iteration of the control method. The linearization makes use of Taylor series expansion and of the computation of the system's Jacobian matrices. For the approximately linearized model of the bioprocess an H-infinity feedback controller is designed. The feedback gain of the controller is found from the repetitive solution of an algebraic Riccati equation, taking place at each iteration of the control method. The stability of the proposed control scheme is evaluated through Lyapunov analysis. First, it is demonstrated that the control system satisfies the H-infinity tracking performance criterion, which signifies robustness against modelling uncertainty and external perturbations. Moreover, under moderate conditions it is proven that the control loop is globally asymptotically stable. The proposed control method solves finally the nonlinear optimal control problem for bioreactors in a computational efficient and of proven convergence manner.

A novel adaptive neurofuzzy H-infinity control approach to feedback control and stabilization of the nonlinear dynamical model of bioreactors used in biofuels production is developed. The form and the parameters of the differential equations that constitute the dynamic model of the bioreactor are considered to be unknown, while there is only knowledge about the order of the system. The model of the controlled system undergoes approximate linearization round a temporary equilibrium which is recomputed at each iteration of the control algorithm. The linearization procedure makes use of Taylor series expansion and the computation of Jacobian matrices. For the approximately linearized model of the bioreactor it is possible to design a stabilizing H-infinity feedback controller, provided that knowledge about the matrices of the linearized state-space description is available. Neurofuzy networks are used to estimate the unknown dynamics of the system and its Jacobians. The computation of the feedback controller's gain comes from the solution of an algebraic Riccati equation taking place at each iteration of the control method, and this allows the implementation of the H-infinity feedback controller. The learning rate of the neurofuzzy approximators is chosen from the requirement the first derivative of the system's Lyapunov function to be always a negative one, thus assuring the stability of the control loop. The global asymptotic stability and the robustness properties of the control method are proven through Lyapunov stability analysis.

The reality of Internet of Things (IoT), with its growing number of devices and their diversity is challenging current approaches and technologies for a smarter integration of their data, applications and services. While the Web is seen as a convenient platform for integrating things, the Semantic Web can further improve its capacity to understand things' data and facilitate their interoperability. In this paper we present an overview of some of the Semantic Web technologies used in IoT systems, as well as some of the well accepted ontologies used to develop applications and services for the IoT. We finally present the Semantic Web Stack for the Internet of Things pointing out some of its shortcomings in the development of an IoT application or service.

The PDE model of the commodities price dynamics is shown to be equivalent to a multi-asset Black-Scholes PDE. Actually it is a diffusion process evolving in a 2D assets space, where the first asset is the commodity's spot price and the second asset is the convenience yield. By applying semi-discretization and a finite differences scheme this multi-asset PDE is transformed into a state-space model consisting of ordinary nonlinear differential equations. For the local subsystems, into which the commodities PDE is decomposed, it becomes possible to apply boundary-based feedback control. The controller design proceeds by showing that the state-space model of the commodities PDE stands for a differentially flat system. Next, for each subsystem which is related to a nonlinear ODE, a virtual control input is computed, that can invert the subsystem's dynamics and can eliminate the subsystem's tracking error. From the last row of the state-space description, the control input (boundary condition) that is actually applied to the multi-factor commodities' PDE system is found. This control input contains recursively all virtual control inputs which were computed for the individual ODE subsystems associated with the previous rows of the state-space equation. Thus, by tracing the rows of the state-space model backwards, at each iteration of the control algorithm, one can finally obtain the control input that should be applied to the commodities PDE system so as to assure that all its state variables will converge to the desirable setpoints.

A new method of adaptive control for active power filters is developed in this article. By proving that the active power filter is a differentially flat system, its transformation to the linear canonical (Brunovsky) form becomes possible. In this new description the control input of the active power filter comprises unknown nonlinear terms which are identified by neurofuzzy networks and through an adaptation / learning procedure. These estimated parts of the system's dynamics are used in an indirect adaptive control scheme, which finally makes the outputs of the active power filter converge to the desirable setpoints. The learning rate in the aforementioned adaptation procedure is given a value which assures that a suitably chosen Lyapunov function will remain negative definite. Under the proposed control method, the closed loop of the active power filter is shown to satisfy the H-infinity tracking criterion, which implies a maximum capability for rejection of external perturbations as well as of modelling errors. flatness-based adaptive fuzzy control based on differential flatness theory is a completely model-free control method. When designing the controller, there is no need for prior knowledge of the system's parameters and state-space equations.

An experimental investigation of three adaptive algorithms for tracking the fundamental frequency in electric transmission grids is reported. An adaptive Prony's method, an adaptive notch filter and an extended Kaiman filter have been presented in this paper. The design principles and the validity of the models have been sketched. For each method, appropriate models are presented and parameter adjustment guidelines are also proposed. The algorithms were developed in simple digital implementations compliant to real-time applications, therefore with low computational burden. Their performance have been evaluated and compared with the zero-crossing technique which serves as a reference. They have been designed especially for estimating frequency changes like small jumps and when the signal is corrupted with noise and other disturbances due to harmonics. The frequency estimators are compared on the basis of precision, transient response, and degree of noise and harmonics immunity.

In this paper, three optimized neural harmonics extraction methods are presented and compared in terms of simulation results, FPGA implementation and practical considerations. Those distortion identification schemes are used in nonlinear loads compensation with Active Power Filters (APF). This optimization is performed in order reduce the number of hardware resources required for digital implementations. The given approaches tend to use only one Adaline and remain powerful even under unbalanced conditions of voltage. In this way, the implementation of all the functionalities of the active filter control has been realized by means of a unique FPGA chip. Moreover, even the most consuming method (i.e., the ITM) uses less than 52% of hardware resources. Even though the mp-q technique (based on the instantaneous reactive power theory) is not the fastest in terms of hardware response time, it stills appear the most powerful for its filtering aptitudes with an experimental source-side current THD of 3.3% after compensation.

A new approach based on a linear Multi Layer Perceptron (MLP) is introduced for harmonics identification. This neural approach uses linear neurons and inputs composed of synthetic harmonic terms in order to fit Fourier series of periodic signals. The amplitudes of the fundamental and high-order harmonics are deduced from a combination of the weights. The effectiveness of the approach is evaluated and compared to an Adaline-based method. Results show that the linear MLP is able to identify in real-time the amplitudes of harmonic terms from measured signals under noisy conditions. The approach can therefore be inserted in compensation strategies to ensure power quality in electrical grids.

This work reviews Adaline-based techniques for estimating Fourier series. The Adaline, with its linear structure and learning, fits a Fourier series by expressing any periodic signal as a sum of harmonic terms. The learning with elementary harmonic inputs enforces the weights to converge to the amplitudes. The Adaline therefore individually identifies the amplitudes of the harmonic terms present in the measured signal in real-time. Relevant study cases are provided. Performances are evaluated and show that harmonic terms of the signals are efficiently estimated.

A new strategy is presented for the implementation of threshold logic functions with binary-output Cellular Neural Networks (CNNs). The objective is to optimize the CNNs weights to develop a robust implementation. Hence, the concept of generative set is introduced as a convenient representation of any linearly separable Boolean function. Our analysis of threshold logic functions leads to a complete algorithm that automatically provides an optimized generative set. New weights are deduced and a more robust CNN template assuming the same function can thus be implemented. The strategy is illustrated by a detailed example.

This paper shows the advantage of using neural network modularity over conventional learning schemes to approximate complex functions. Indeed, it is difficult for artificial neural networks like Kohonen extended maps to converge toward an efficient and adequate solution when the dimensionality of the input and output spaces are high. Associated to an appropriate learning technique, modularity in neural networks is able to overcome the high dimensionality of the input/input space by decomposing it into different intermediate spaces of reduced dimensionality. The decomposition results in independent neural modules. The efficiency of this learning technique will be enlightened with a visual servoing application. In this application, the relationship between the visual features issued from a stereoscopic vision system and the angles of a 5 DOF-robot will be learned and approximated. Simulations have been conducted and clearly show that this complex, nonlinear, and high dimensional function can be learned efficiently with the neural network modularity approach. Moreover, we show through these simulations that neural modules can be re-utilized, thus reducing the convergence time of the learning and the memory requirements.

Prediction in real-time image sequences is a key-feature for visual servoing applications. It is used to compensate for the time-delay introduced by the image feature extraction process in the visual feedback loop. In order to track targets in a three-dimensional space in real-time with a robot arm, the target's movement and the robot end-effector's next position are predicted from the previous movements. A modular prediction architecture is presented, which is based on the Kalman filtering principle. The Kalman filter is an optimal stochastic estimation technique which needs an accurate system model and which is particularly sensitive to noise. The performances of this filter diminish with nonlinear systems and with time-varying environments. Therefore, we propose an adaptive Kalman filter using the modular framework of mixture of experts regulated by a gating network. The proposed filter has an adaptive state model to represent the system around its current state as close as possible. Different realizations of these state model adaptive Kalman filters are organized according to the divide-and-conquer principle: they all participate to the global estimation and a neural network mediates their different outputs in an unsupervised manner and tunes their parameters. The performances of the proposed approach are evaluated in terms of precision, capability to estimate and compensate abrupt changes in targets trajectories, as well as to adapt to time-variant parameters. The experiments prove that, without the use of models (e.g. the camera model, kinematic robot model, and system parameters) and without any prior knowledge about the targets movements, the predictions allow to compensate for the time-delay and to reduce the tracking error.

Vision has extensively expanded robots' capabilities, making the robot control problem more complex. To track a target with a robot arm in a three-dimensional space involves the use of precise commands. We propose to insert a hierarchical neurocontroller based on CMAC (cerebellar model articulation controller) networks in a visual servoing loop. This hierarchical structure splits the robot's workspace and assigns different CMAC controllers imposing thus a specialized region CMAC. Compared to a single CMAC with the same number of weights, the neurocontroller's sensitivity and precision is increased. Robot positioning and target tracking with visual feedback can then be done with a better precision.

A new adaptive Kalman filter is proposed to address the problem of nonlinear systems that cannot be linearized or where the model is unavailable. Using a correlation function of the output vector of a state model system, the transition matrix of the Kalman filter is adjusted to the current situation. This adaptive transition matrix, associated to Kalman gain compensation, produces efficient state estimation. The performance of this predictor has been evaluated on a visual servoing application.