Martin Hennebel

Introducing decentralized EV charging coordination for the voltage regulation

This paper investigates a decentralized optimization methodology to coordinate Electric Vehicles (EV) charging in order to contribute to the voltage control on a residential electrical distribution feeder. This aims to maintain the voltage level in function of the EVs power injection using the sensitivity matrix approach. The decentralized optimization is tested with two different methods, respectively global and local, when EV take into account their impact on all the nodes of the network or only on a local neighborhood of their connection point. EV can also update their decisions asynchronously or synchronously. While only the global approach with asynchronous update is theoretically proven to converge, using results from game theory, simulations show the potential of other algorithms for which fewer iterations or fewer informations are necessary. Finally, using Monte Carlo simulations over a wide range of EV localization configurations, the first analysis have also shown a promising performance in comparison with uncoordinated charging or with a “voltage droop charging control” recently proposed in the literature.

Reducing the Impact of EV Charging Operations on the Distribution Network

A key assumption made in this paper is that electric vehicle (EV) battery charging profiles are rectangular. This requires a specific and new formulation of the charging problem, involving discrete action sets for the EVs in particular. The considered cost function comprises of three components: 1) the distribution transformer aging; 2) the distribution energy losses; and 3) a component inherent to the EV itself (e.g., the battery charging monetary cost). Charging start times are determined by the proposed distributed algorithm, whose analysis is conducted by using game-theoretic tools such as ordinal potential games. Convergence of the proposed algorithm is shown to be guaranteed for some important special cases. Remarkably, the performance loss with respect to the centralized solution is shown to be small. Simulations, based on realistic public data, allow one to gain further insights on the issues of convergence and optimality loss, and provide clear messages about the tradeoff associated with the presence of the three components in the considered cost function. While simulations show that the proposed charging policy performs quite similarly to existing (continuous) charging policies such as valley-filling-type solutions when the non-EV demand forecast is perfect, they reveal an additional asset of rectangular profiles in presence of forecasting errors.

Minimizing the impact of EV charging on the electricity distribution network

The main objective of this paper is to design electric vehicle (EV) charging policies which minimize the impact of charging on the electricity distribution network (DN). More precisely, the considered cost function results from a linear combination of two parts: a cost with memory and a memoryless cost. In this paper, the first component is identified to be the transformer ageing while the second one corresponds to distribution Joule losses. First, we formulate the problem as a non-trivial discrete-time optimal control problem with finite time horizon. It is non-trivial because of the presence of saturation constraints and a non-quadratic cost. It turns out that the system state, which is the transformer hot-spot (HS) temperature here, can be expressed as a function of the sequence of control variables; the cost function is then seen to be convex in the control for typical values for the model parameters. The problem of interest thus becomes a standard optimization problem. While the corresponding problem can be solved by using available numerical routines, three distributed charging policies are provided. The motivation is threefold: to decrease the computational complexity; to model the important scenario where the charging profile is chosen by the EV itself; to circumvent the allocation problem which arises with the proposed formulation. Remarkably, the performance loss induced by decentralization is verified to be small through simulations. Numerical results show the importance of the choice of the charging policies. For instance, the gain in terms of transformer lifetime can be very significant when implementing advanced charging policies instead of plug-and-charge policies. The impact of the accuracy of the non-EV demand forecasting is equally assessed.

Sensitivity analysis of a simulation model for evaluating renewable distributed generation on a power network

We present a sensitivity analysis of a simulation model for the evaluation of the performance of a renewable distributed generation (DG) network. Uncertainties in renewable energy sources, components failure and repair events, loads and grid power supply are taken into account. The sensitivity analysis is performed with respect to the characteristic uncertain variables associated to each type of DG technology available. The impact of these uncertain variables is evaluated in terms of two performance functions, global cost (Cg) and energy not supplied (ENS). The results show the trends of performance of the DG-integrated network under different conditions. This allows evaluating the impact of the different technologies.

Probabilistic load flow methods to estimate impacts of distributed generators on a LV unbalanced distribution grid

The aim of this paper is to apply probabilistic load flow methods on a three phases, unbalanced low voltage distribution network. We use a point estimate method and a Monte Carlo simulation based method to estimate the electrical characteristics (buses voltage, phases and neutral conductors currents) of a distribution grid in presence of a large number of small size photovoltaic generators. Probabilistic load flow allows us to take into account the uncertainty of photovoltaic production and load consumption in load flow computation. The literature shows that PEM method gives good accuracy results while requiring less time simulation than Monte Carlo simulation. In this paper, we aim to check if this assumption is still right with different kinds of probability density function and for a large size electrical network. Usually, random parameters are modeled as a normal distribution. In this work, a generalized extreme value is used to model load consumption behaviour instead of a normal one. The uncertainty of photovoltaic production is supposed to be directly linked to the sky clear index which is modeled as a beta distribution.

Network reduction based on multiple scenarios

Simulation of large power systems require the reduction of the complexity of the network models. Different works propose methodologies to aggregate buses and to model the connections between those clusters, that are computationally expensive and developed to work around a specific operating point, presenting significant errors when that set-point changes. This paper proposes a methodology to define a reduced network model consisting in aggregating network buses, using clustering algorithms, and modelling the connections between them as an optimization problem as well as a performance assessment index to evaluate its accuracy. It differentiates from state of the art by the use of multiple operation scenarios, that help to improve its robustness to the system operation set-point, and through its application to a realistic case-study of the European transmission network.