Skip to Main content Skip to Navigation
Theses

Une nouvelle approche pour la programmation distribuée dans les smartgrids

Abstract : The main challenge of smart grids control and management is the amount of data to be processed. Traditional, centralized techniques, even if they offer the advantage of the ease of management by their global grid vision, do not support in practice the continuous growth of data volumes (limited bandwidth, bottleneck, amount of calculations, etc.). The transition to decentralized(distributed)control and management, where the system is made up of a multitude of co-operating computing units, offers very good prospects (robustness, calculation close to the producers and consumers of data, exploitation of data in all available resources), but remains challenging to implement. In fact, the programming of distributed algorithms requires taking into account the data exchanges and the synchronization of the participating units; this complexity increases with the number of units. In this thesis, we propose an innovative approach of programming of a high level of abstraction masking these difficulties.First, we propose to abstract all Smartgrid computing units (smart meters, sensors, data concentrators, etc.) as a distributed database. Each computing unit hosts a local database and only the data needed to continue the calculation are exchanged with other units, which decreases the use of the available bandwidth. The use of a declarative data handling language will simplify the programming of control and management applications. Besides, we also propose SmartLog, a rule-based language (based on the Datalog language and its derivatives dedicated to these applications. It facilitates distributed programming of Smartgrid applications by immediately responding to any changes in the data.Even with a language such as SmartLog, it is necessary to take into account the data exchange and the synchronization of the participants. This is why we then propose an approach that simplifies distributed programming. This approach, named CPDE for Centralized Programming and Distributed Execution, consists of two steps: (i) programming the centralized application in SmartLog, as this is easier, and (ii) translating the centralized program into a distributed program based on the actual location of the data. To do this, we propose a semi-automatic Smartlog rule distribution algorithm.In order to demonstrate the interest of CPDE, we conducted a comprehensive experiment using applications and algorithms actually used in Smartgrids, such as secondary control in isolated micro-grids or fair voltage regulation. The experiment was carried out on a real-time electrical network simulation platform, with an OPAL-RT simulation machine, and a Raspberry-Pi network representing the computing units (their performances are quite comparable to the real equipment). This experiment allowed validating the behaviours and the performances of the distributed programs conceived with CPDE, and comparing to their centralized versions in SmartLog and their reference versions implanted in Java. The impact of different parameters, such as the number of calculation units or different data distribution alternatives, is studied as well.
Document type :
Theses
Complete list of metadatas

Cited literature [225 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-02536617
Contributor : Abes Star :  Contact
Submitted on : Wednesday, April 8, 2020 - 11:44:39 AM
Last modification on : Wednesday, October 14, 2020 - 4:17:59 AM

File

NGUYEN_2019_diffusion.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-02536617, version 1

Collections

Citation

Thi Thanh Quynh Nguyen. Une nouvelle approche pour la programmation distribuée dans les smartgrids. Energie électrique. Université Grenoble Alpes, 2019. Français. ⟨NNT : 2019GREAT079⟩. ⟨tel-02536617⟩

Share

Metrics

Record views

146

Files downloads

156