Fiche individuelle

Low voltage distribution networks have not been traditionally designed to accommodate the large-scale integration of decentralized photovoltaic (PV) generations. The bidirectional power flows in existing networks resulting from the load demand and PV generation changes as well as the influence of ambient temperature led to voltage variations and increased the leakage current through the cable insulation. In this paper, a machine learning-based framework is implemented for the identification of cable degradation by using data from deployed smart meter (SM) measurements. Nodal voltage variations are supposed to be related to cable conditions (reduction of cable insulation thickness due to insulation wear) and to client net demand changes. Various machine learning techniques are applied for classification of nodal voltages according to the cable insulation conditions. Once trained according to the comprehensive generated datasets, the implemented techniques can classify new network operating points into a healthy or degraded cable condition with high accuracy in their predictions. The simulation results reveal that logistic regression and decision tree algorithms lead to a better prediction (with a 97.9% and 99.9% accuracy, respectively) result than the k-nearest neighbors (which reach only 76.7%). The proposed framework offers promising perspectives for the early identification of LV cable conditions by using SM measurements.

The structure of the Low Voltage distribution networks is not always accurately known by the Distribution System Operators, especially with the significant meteorological variations observed in recent years and the growth of decentralized PV production sources. In this paper a probabilistic Load Flow algorithm has been developed for radial Low Voltage network considering the line resistance distribution as an uncertainty depending on the temperature. The single-phase network model is therefore associated to the temperature variation in the network deployment area. The Load demand and the PV production generally used in classical Load Flow calculation are computed using Smart Meter data with a quarter of an hour resolution time. Both power values are considered to be time varying. Either the resistance value or the network to client exchanged power are randomly selected at each iteration using a Monte Carlo method. Both annual and seasonal dependencies of the line resistance have been implemented in the developed Probabilistic Load Flow. The simulation results have shown that integrating the resistance distribution in a seasonal probabilistic tool can impact the collected reliability indices up to 10.4% depending on the season. In a context of upgrading the Low Voltage electrical network knowledge by the Distribution System Operator, and with an accordance to the requirements of the EN50160 standard, this tool can be presented as an efficient algorithm for quantifying the impact of the line resistance statistical distribution on a Probabilistic Load Flow computation.

Voltage variations that occur in the Low Voltage (LV) distribution networks are caused by different factors such as the load demand changes, intermittent powers of photovoltaic (PV) generation, the phenomena related to the operating environment of the network and their influence on the physical structure of the network, network faults, etc. The voltage variations can have important adverse impacts on the connected equipment. This paper aims to analyze the electrical conductance variation of the cable insulation, due to the degradation of the insulating material, and its impact on the nodal voltages. To this end, a probabilistic framework is proposed based on the Monte Carlo (MC) simulations and load flow calculations. The MC simulations are used here in order to characterize the insulation degradation of lines through scenarios and to model the resistance variation of lines. Considering the load demand and photovoltaic generation in a single-phase network as well as line the impedance data, in different degradation conditions, load flow calculations are conducted in order to calculate the nodal voltages. Simulation results reveal voltage drops over 50% depending on the position, number and degree of the cable insulation degradation as well as on the network working conditions. This result offers promising perspectives for the detection of cable degradations by use of Smart Meter (SM) voltage measurements.

Ce projet de recherche explore des approches scientifiques pour la modélisation des réseaux de distribution électrique à partir des données énergétiques mesurées par les compteurs communiquant. Ces compteurs ont été initialement déployés pour permettre la mise en place de tarifs dynamiques et ainsi une participation active des consommateurs sur le marché de l’électricité. Dans ces travaux de recherche, on explore la possibilité d’augmenter l’observabilité sur le fonctionnement du réseau et de déterminer l’état de dégradation physique des lignes et câbles électriques. En exploitant les données mesurées, une méthode heuristique permet de retrouver une architecture du réseau satisfaisant les points de fonctionnement mesurés. Les impédances entre nœuds sont ensuite identifiées. A partir de l’analyse statistique de ces impédances au cours d’une année, l’impact de la température extérieure sur les lignes et câbles du réseau est analysé ainsi que l’impact sur le plan de tension. Les données sont ensuite utilisées pour investiguer l’influence de la dégradation des isolants sur les variations de tensions aux différents nœuds du réseau. Plusieurs techniques d’intelligence artificielle, Machine Learning sont évalués et comparées pour détecter préventivement ces défauts dans les câbles Basse Tension. Pour les réseaux de distribution, les méthodes et outils développés dans cette thèse peuvent aider à maintenir leur fonctionnement, élargir leur capacité d’hébergement (augmentation de la demande pour les ménages, nouveaux consommateurs et sources renouvelables) et permettre la planification rentable des opérations de maintenances.