Search

International audience

International audience; Distributed generators (DGs) reactive powers are controlled to mitigate voltage overshoots in distribution grids with stochastic power production and consumption. Classical DGs controllers may embed piecewise affine laws with dead-band terms. Their settings are usually tuned using a decentralized method which uses local data and optimizes only the DG node behavior. It is shown that when short-term forecasts of stochastic powers are Gaussian and the grid model is assumed to be linear, nodes voltages can either be approximated by Gaussian or sums of truncated Gaussian variables. In the latter case, the voltages probability density functions (pdf) that are needed to compute the overvoltage risks or DG control effort are less straightforward than for normal distributions. These pdf are used into a centralized optimization problem which tunes all DGs control parameters. The objectives consist in maximizing the confidence levels for which voltages and powers remain in prescribed domains and minimizing voltage variances and DG efforts. Simulations on a real distribution grid model show that the truncated Gaussian representation is relevant and that control parameters can easily be updated even when extra DGs are added to the grid. The DG reactive power can be reduced down to 50% or node voltages variances can be reduced down to 30%.

The insertion of stochastic renewable energies in distribution grids generates important voltage fluctuations and reactive power exchange between distribution grids and transmission grids. These variations need to comply with European grid codes. The control strategy involves the actuation of the On Load Tap Changer and Capacitor Banks tap position, and the reactive powers of the DGs. This paper focuses on the stochastic estimation of the powers at the interface and the influence of the Capacitor Bank tap position. The main results are that, using a simplified stochastic model of the grid, one can obtain the distribution of the powers, and select, on a daily basis, the tap position which minimizes the level of the prescribed PQ diagram violation risk. This study paves the way for a volt-var stochastic optimization of the distribution grid.

Volatile productions and consumptions generate a stochastic behavior of distribution grids and make its supervision difficult to achieve. Usually, the Distributed Generators reactive powers are adjusted to perform decentralized voltage control. Industrial controllers are generally equipped with a local affine feedback law, which settings are tuned at early stage using local data. A centralized and more efficient tuning method should aim to maximize the probability that all the node voltages of distribution grids remain within prescribed bounds. When the characteristics of the stochastic power forecasts are known, the centralized algorithm allows to update the settings on a regular time basis. However, the method requires to solve stochastic optimization problem. Assuming that stochastic variables have Gaussian distributions, a procedure is given which guarantees the convergence of the stochastic optimization. Convex problems drastically reduce the difficulty and the computational time required to reach the global minimum, compared to nonconvex optimal power flow problems. The linear controllers with optimized parameters are compared to traditional control laws using simulations of a real distribution grid model. The results show that the algorithm is reliable and moreover fast enough. Hence, the proposed method can be used to update periodically the control parameters.

This paper proposes a two-layer microgrid supervisor based on Model Predictive Control (MPC). The supervisor in the upper layer relies on an economical optimization that considers the cost of energy and the load and production forecasts to define the State of Charge (SoC) targets for each storage device on a timescale of 15 minutes. The lower layer displays a shorter timescale and aims to control the equipment to ensure the stability of the overall system and SoC tracking while satisfying the economic constraints specified by the upper layer. These two layers require an uniformization of the timestep and of the references in order to behave properly. The main contributions of this paper are the microgrid network modelling embedded in the optimization routine of the lower layer and a discretization for integrating upper-layer references.

This article deals with a joint distribution system operator–transmission system operator (DSO–TSO) reactive power management to support TSO operations at high-voltage level. This management is based on the coordination of several volt-var control schemes of medium-voltage networks able to control the reactive power at the DSO–TSO interface. The coordination is achieved through the resolution of an OPF problem. The definition of this latter is further discussed and recommendations are raised.

The control of continuous stirred bioreactors using productivity signal feedback consists of using the direct relation between the optimal static dilution rate and the optimal static productivity. This strategy drives the bioprocess to its optimal productivity point, irrespective of the feeding substrate concentration, and specific growth rate or substrate concentration estimates. While the algorithm covers a wide class of bioprocesses, explicit relations are given for the Monod and Haldane cases.

At low speeds, wind energy conversion systems (WECS) can extract a maximum output power for a reference wind turbine-wind speed ratio. While the classical control strategy aims at tracking this Maximum Power Point using Power Signal Feedback, the transmission shaft flexibilities may yield undesirable mechanical vibrations. In this paper, a controller embedding a delayed power signal is proposed which is able to damp vibrations, when the time-delay is adequately chosen, while tracking the optimal set-point. Other strategies involving time-delayed feedback are proposed and simulations show the relevance of the proposed approach.

A new and fast registration algorithm has been proposed to update the tool-path of a deburring robot, intended to machining composite workpieces under gravity and clamping deformations. A Subdivision Iterative Closest Point algorithm, which considersdifferent parts of the contour with respect to curvature, allows to obtain far better results than classical methods, without complicated assumptions or computations.The procedure has shown to be effective for porthole and nose-cone deburring. Experimental tests conducted on robotic milling workcell demonstrated the efficiency of the registration method.; International audience; A new and fast registration algorithm has been proposed to update the tool-path of a deburring robot, intended to machining composite workpieces under gravity and clamping deformations. A Subdivision Iterative Closest Point algorithm, which considersdifferent parts of the contour with respect to curvature, allows to obtain far better results than classical methods, without complicated assumptions or computations.The procedure has shown to be effective for porthole and nose-cone deburring. Experimental tests conducted on robotic milling workcell demonstrated the efficiency of the registration method.

This paper proposes a case study of the design of robust control for flexible axes. Robust parametric methods are worked out, and an heuristic initial tuning of the PID controller is proposed from additional pole-placement constraints on the rigid mode. Experimental validations highlight the relevance of adequate closed-loop control and trajectory planning design.

While improving the dynamical performance of a flexible motion control system, additional stresses are generated which induce undesirable vibrations. The design of adequate control algorithms makes it possible to enhance the overall behaviour of the machine while limiting its deformation modes vibrations. This paper presents a comparative study of the influence of different control algorithms on the trade-off between the cycle time and the amplitude of the residual vibrations, which have been applied to a Cartesian robot. While this selection of control algorithms is not exhaustive, these can all easily be implemented in current industrial systems. Each selected control algorithm is analysed along with experimental validations.

An indirect iterative learning control is proposed to update the gains of a Parallel Distributed Compensation controller when the trajectory is repetitive. The corresponding Takagi-Sugeno fuzzy system, to be controlled, is supposed to exhibit time-varying matrices in the consequent part. When the membership functions are fixed, it will be possible to estimate the consequent system for every time, using different control gains for each trial. Finally, when the system parameters are considered as properly estimated, it will be possible to derive a controller which will not only be stable but will also exhibit good tracking properties.

A controller using acceleration feedback has been applied to a flexible robot for which the position and velocity of the load are not measured. It is shown that acceleration feedback allows an exact tracking of the motor position, irrespective of the non-linear flexibilities of the axes and of the measurement disturbances. This easy-to-tune algorithm whose main control parameters are the modal masses of the motor and load part, and only consists of a positive acceleration feedback plus a PD controller, has been validated on an industrial robot with orthogonal axes.

This paper reviews a class of tools which can help to check the stability of fuzzy linguistic controllers which is a difficult task due to their regionwise structure. The application of conventional methods when considering a fuzzy controller as a global nonlinear controller often leads to conservative stability results. The use of appropriate continuous or piecewise continuous Lyapunov functions allows to address the stability of nonlinear systems under fuzzy control while taking into account the local structure of the fuzzy controllers. Some methods are presented which enforce the global Lyapunov function to decrease when leaving one subspace for another closer to the equilibrium, using concentric domains and providing less conservative conditions.

This paper investigates the way modelling mixing phenomena occur in unsteady stirring conditions in agitated vessels. In particular, a new model of torus reactor including a well-mixed zone and a transport zone is proposed. The originality of the arrangement of ideal reactors developed here lies in the time-dependent location of the boundaries between the two zones. This concept is applied to model the positive influence of unsteady stirring conditions on homogenization process: the model avoids a mass balance discontinuity when the transition from steady to unsteady stirring conditions is performed. To ascertain the reliability of the model proposed, experimental runs with highly viscous fluids have been carried out in an agitated tank. The impeller used was a non-standard helical ribbon impeller, fitted with an anchor at the bottom. The degree of homogeneity in the tank was observed using a conductivity method after a tracer injection. It is shown that for a given agitated fluid and mixing system, model parameters are easy to estimate and that modelling results are in close agreement with experimental ones. Moreover, it would appear that this model allows the easy derivation of a control law, which is a great advantage when optimizing the dynamics of a mixing process. 2005 Elsevier Ltd. All rights reserved.

International audience; A method for detecting in real time structure changes of a food product during a heat transfer process in an agitated vessel is proposed. The method is based on electrical conductivity measurements and a modified Generalized Likelihood Ratio (GLR) algorithm was applied to the first derivative of the electrical conductivity signal. The paper describes in details protocol for carrying out such method. Examples of structure changes obtained for a starch mixture and a gelatin solution are presented. Finally, by monitoring apparent viscosity of the solution during heating, reliability of change detection in real time is ascertained.

A class of nonlinear systems with input-dependent parameters and transport delays is considered, for which the output is oscillatory. The input-dependent transport delay is given by an implicit equation. As for δ-flat systems, the open-loop motion planning problem can be solved using an explicit parametrization of the trajectories. It is shown that, using an approximate solution of the time delay function, the control input can be found explicitly, but the structure of the system generates some constraints on the input and therefore restricts the performances of the admissible trajectories. The motion planning problem with a transport delay is solved using a discretisation of the implicit equation which defines this delay. An example is provided, in which the desired trajectory is piecewise polynomial: the open loop control and the corresponding constraints are explicitly given. It is shown that, in both cases, the control signal presents discontinuities or peaks that can be of high amplitude; however, the performances are not significantly altered when the control is saturated.

In this paper, it is shown that a fuzzy controller with singleton deffuzzification can be considered as the association of a regionwise constant term and of a regionwise non linear term, the latter being bounded by a linear controller. A systematic design method is presented that allows to design stable controlled systems. Stability conditions are given under the form of a set of inequalities which provides an admissible domain for the controller parameters.

This paper proposes a chemical Port-Hamiltonian formulation of a well-mixed CSTR model considering that the chemical reaction is taking place at constant pressure and temperature. More focus is put on chemical reaction network theory and its inclusion in the formulation to achieve synchronization between concentration space and reaction space. It is made clear that Gibbs free energy is an apt Hamiltonian function for such cases. The same concept is applied on a basic enzyme reaction. The Bond Graph models related to Hamiltonian formulation for both types of reactions are given in order to show its ability of pictorial representation and intuitive solution.

The analytical expression of a fuzzy controller has been studied when the output fuzzy membership functions are equally spaced and triangular or trapezoidal. The cases, when a fuzzy set of the output is extreme - with respect to the universe of discourse - or not extreme have been considered. When considering extreme sets, the controller can be approximated with a good precision by a second-degree polynomial and, in the contrary, by a first-degree polynomial. The modelling of the output by a second-degree multivariable polynomial has been shown to affect strongly the convergence rate, which is enlightened by an example for which the closed-loop stability is not exponential.

Several methods have been compared to estimate the dominant frequency of a flexible manipulator. Fast algebraic estimation techniques have proven to be faster than time-domain or FFT methods and, introducing a simplified model, quite easy to compute. While these methods are traditionally used to improve closed-loop algorithms, they have allowed the fast calibration of an open-loop trajectory planning algorithm that decreases significantly the level of vibrations at the end effector.

The non-minimum phase phenomena of a robot's arm modelled by a set of Euler-Bernoulli beams clamped on a moving cart is demonstrated. Modal analysis is used to derive the exact eigensolutions which verify the described geometric boundary conditions. The equations of motion are determined using the Assumed Mode Method. The influence of geometrical and mechanical parameters such as the thickness of the cross sections of the beams, or the beam length, is discussed using numerical simulations. The overall results, with an application to a real robot, show that one has to take the non-minimum phase problem into consideration when designing a flexible mechanical structure.

Two kinds of hybrid multimodel-neural network controllers are presented, which either realize a fusion of local controllers or a fusion of local models which will generate an appropriate controller. The hybrid controller using control fusion improves the tracking ability of controllers using multiple models, by taking into account a better approximation of the adequate control. Simulation results are shown on an underdamped system.

A class of nonlinear systems with input dependent parameters and time delays has been considered, for which the output is oscillatory. As for /spl delta/-flat systems, the open-loop motion planning problem can be solved using an explicit parametrization of trajectories; however, it is shown that the structure of the system brings out some constraints on the input and therefore limits the possible choice of trajectories. An example is provided, in which the trajectory is piecewise polynomial, where the open loop control and the corresponding constraints are explicitly given.

In this paper, a new multimodel approach for complex systems modeling based on classification algorithms is presented. It requires firstly the determination of the model-base. For this, the number of models is selected via a neural network and a rival penalized competitive learning (RPCL), and the operating clusters are identified by using the fuzzy K-means algorithm. The obtained results are then exploited for the parametric identification of the models. The second step consists in validating the proposed model-base by using the adequate method of validity computation. Two examples are presented in this paper which show the efficiency of the proposed ap- proach.

Adequate Path Planning design is an important stage for controlling flexible axes because it may allow to cancel vibrations induced by oscillating modes. Among bang-bang profiles which are linked to optimal control, jerk assignment (acceleration derivative) and input shapers have been investigated. Theoretical results show the performance and robustness with respect to natural frequency mismatch. Practical validations on a real robot arm show the relevance of the jerk algorithm which is more robust with the same productivity performances as input shaping techniques.