Development of Advanced Smart Energy Management Framework Integrated with Optimization Techniques and Prediction Models for Demand Side Consumers Based on IoT Platform

Abstract

Smart grid framework plays an indispensable role in dealing with the usage of available electrical energy proficiently. However, to manage power supply effectively, electrical appliances and devices at home and building environment should have smart energy management capability. Further, Smart Energy Management System (SEMS) can be unified with the smart grid for effective power consumption. SEMS can be used to control the status of the electrical appliances and devices by monitoring environmental conditions with the associated sensors and the context in which the appliance is being operated. In addition, SEMS can be used to reduce the standby power consumption of the appliances by turning off the supply to it. The SEMS system can be associated with a Grid or the distributed generation, and thus power negotiation techniques can be applied depending on the availability of the power or tariff information. In this research work, the emphasis is given to the design of a smart energy management system and deployment of power negotiating algorithms for effective power utilization. The proposed SEMS replaces the scenario of a complete power outage in a particular region with partial load shedding in a controlled manner as per consumer’s priority. The hardware experiments are demonstrated assuming a demand response event, taking into account the constraints of maximum demand limits in various cases of changing priorities. The cost optimization algorithms are deployed by scheduling the appliances, considering the Time of Usage (ToU) and minimum slab rate. Sensory information’s and indicators are used to control the loads with user comfort settings and alarm the user during peak hour usage, respectively. Reliable ZigBee communication is established in the Application Transparent (AT) mode of configuration with a self-diagnostic mechanism. Internet of Things (IoT) environment is created for uploading the data, storing it in the database with load wise data analysis daily and monthly basis with Graphical User Interface (GUI). The challenge of energy shortages requires an optimized solution to demandside consumer issues. Energy demand factors contribute to the implementation of variable tariff’s and reward consumer’s electricity usage during off-hours rather than during peak hours. In addition, the surge in tariffs iiiand price volatility emphasize the need to carefully schedule the operation of large devices to minimize power consumption. In this task, a genetic algorithm is used to find the optimal load schedule that minimizes the cost spent on power according to considerations such as user comfort, maximum allowable demand, load characteristic’s, environmental factors and so on. Further, this work will focus on testing the Binary Backtracking Search and Artificial Bee Colony algorithms against the Binary Particle Swarm algorithm benchmark to find the optimal load scheduling in terms of complexity, cost optimization, and execution time. On the other hand, the challenge in energy management lays focus on the efficient utilization of available power sources without limiting power consumption. Above issue seeks for design and development of an intelligent system with day-ahead planning and prior forecasting of energy availability. Hence there is a need for accurate energy prediction technique to minimize imbalance in the power sector. In this context, an Intelligent Smart Energy Management Systems (ISEMS) is proposed to handle energy demand in a smart grid environment with penetration of renewable sources. The proposed scheme compares several prediction models for accurate forecasting of energy for hourly and day-ahead planning. Based on the predicted information, ISEMS negotiates the available power and dispatch the control action depending on the consumer assigned priority for an appliance. Several energy prediction models are evaluated and it is found that the Particle Swarm Optimization (PSO) based Support Vector Regressors (SVR) outperforms over other prediction models in terms of performance accuracy.