Application of Long Short-Term Memory (LSTM) Neural Network for the estimation of communication network delay in smart grid applications

Vast integration of new technologies to enable smart control of the power grid requires a reliable, efficient and resilient communication infrastructure. Today, many communication protocols (e.g., IEC 61850, OPC UA, Modbus, Internet, WiMAX, 4G, Wi-Fi, etc.) and technologies (e.g., PLC, GSM, Optic fiber, RF radio mesh, Cellular, etc.) are established for the smart grid applications. In case of the stability guarantee of smart grid, the Quality of Service (QoS) is a challenge to be considered. One of the major concerns in data delivery over the network is a low latency message transmission to ensure the time critical tasks, e.g., control and protection tasks. In this context, the main contribution of this paper is to propose a model methodology for the communication network delay in smart grid applications. To be compatible with the further goal of delay predictive compensation method, the present paper proposes a message transmission delay estimation method using Long Short-Term Memory (LSTM) neural network. To this end, Python is the chosen programming language, including its required libraries for the considered application. Delay values measured on a real-time HV/MV substation application are used as input data for validation purpose.

I. Introduction

Current power grid is moving towards a modern grid, which benefits from innovative products and services such as: Distributed/Renewable Energy Resources, energy storage technologies, and intelligent control-protection system. Modern power system properly named “Smart Grid” benefits of the Information and Communication Technologies (ICTs) to deliver real-time information and near-instantaneously balance the demand-response on the electrical grid. Thus, smart grids provide: 1) self-healing technologies, 2) distributed control and protection, 3) self-monitoring, 4) two-way communication.