Cognitive Power Metering and prediction using Edge AI

Project Title

Cognitive Power Metering and prediction using Edge AI

Project Area of Specialization Internet of ThingsProject Summary

In this project, a lab testbed for a smart IoT-based solar system will be developed. On top of a conventional solar system, additional capabilities will be added for efficient energy usage. In this context, it is envisaged to devise a proficient approach wherein the system would automatically monitor and control the current, voltage, and other associated parameters of the solar system and provide real-time statistics to the users. Accordingly, Hardware will be developed for monitoring the power generation, battery storage, load consumption along with temperature at the solar panels. Arduino would be used as the main system controller that controls the charging and usage of electricity and transmits these parameters to the cloud for data analytics.

The data analytics and prediction will be performed at the Edge and on the cloud using appropriate ML algorithms. We shall also deploy a reinforcement learning algorithm to predict the future usage and electricity production of solar panels. An optimal load schedule/demand-side response will be generated as a result. The load schedule will be shared with the homeowner on a mobile app and upon his permission will be automatically implemented by providing control input to the load switches. Users would have the capability to track, monitor, and control their solar panels remotely and program the system to optimize the production, storage, and usage of electricity.

Project Objectives

• To develop a smart IoT based solar panel system
• System would automatically monitor and control the solar system's current, voltage, and other associated parameters and provide real-time statistics.
• Hardware development for monitoring of power generation and load consumption, battery storage.
• Development of a machine learning algorithm for prediction of future usage and electricity production of solar panels.
• Deploying the machine learning algorithm on EDGE AI devices and cloud solutions.
• Optimal load schedule/demand-side response will be generated.
• Development of a mobile app or a user interface (UI).

Project Implementation Method

1. Testbed

We may go with a 1 kVA testbed system. Solar panels, batteries, and inverters may be sized accordingly. If off-the-shelf Inverter and charge controller are to be procured, they must allow user control inputs, so that we can implement demand-side management, etc. The solar panels would be provided by the university.

2.Hardware

• Sensors

Voltage, Current, Temperature, Humidity, etc will be used to monitor different parameters.

• Microcontroller

It is used to control each peripheral of the slave node and act as a central processing unit for the node. In addition to this, it will control the charging and usage of electricity and transmit these parameters to the cloud for data analytics. 

• Wireless Communication Module

To enable slave nodes to wirelessly communicate with the master nodes.

• Power Sensors 

The purpose of the slave node is to monitor the power of the appliances, the power sensors will detect the powers being consumed.

• Inverter

The inverter will be used to convert the direct current (DC) to Alternating Current (AC) generated by the solar panels.

• Solar Panel

Multiple Solar panels will be used to generate the DC (direct current).

• Relay

The relays will enable the slave nodes to switch the appliances.

• Main Processor

This model has an inbuilt Wi-Fi module to transfer the data to the IoT platform for data analytics and wirelessly communicate with the cloud. 

          Machine learning

1. Initially, data to be collected for a month regarding solar power generation and consumption will be analyzed using machine learning algorithms.
2. The algorithms will be used to identify the cluster of peak hours when the energy has been generated and consumed to the maximum. This data will be used to perform trend analysis and extract patterns of peak energy generation and consumption.
3. If the produced energy is more than the building requirements the surplus energy would be transmitted to WAPDA. The model would automatically switch the system to power saving mode and turn off some electrical appliances.
4. Implement ML algorithms on already provided data.

1. User Interface and Load Schedule

The user interface and the control of the system are given by a dashboard. Appliances can be switched off from anywhere over the internet. Real-time predictions of power generation and consumption are displayed. An automated demand-side response will be generated. The load schedule will be shared with the user via a dashboard. Users would have the capability to track, monitor, and control their solar panels remotely and to optimize the production, storage, and usage of electricity.

Benefits of the Project

• The solar light concentration varies for every home as it depends on how many hours the home is facing the sun. In this context, the proposed system is useful in predicting the peak hours of energy generation to optimize power consumption.
• Solar energy has various advantages in terms of increasing energy efficiency and reducing greenhouse gas emissions. 
• Using solar energy as a distributed energy resource, it is possible to minimize transmission loss and supply energy to the consumer more efficiently.
• On the remote-based IoT platform, users can perform plant monitoring, generation monitoring, and fault detection remotely. Minimum human intervention is needed.
• The IoT network is scalable, and more nodes can be made and connected to other electrical nodes of the house.
• Machine learning algorithms are applied to gather data for solar forecasting and load scheduling. These algorithms give insights into the future power consumption, and solar generation and useful suggestions are provided to the user to shed loads efficiently.
• The IoT network can efficiently monitor and control electricity bills by using the suggestions provided by the web application.

Technical Details of Final Deliverable

We will design a prototype of a smart energy monitoring system based on IoT and machine learning. The application will predict the energy consumption bill and solar PV output based on the trained machine learning models. The design of the hardware will incorporate an actual PV array with an inverter, charge controller, and battery to make a product-based solution.

Final Deliverable of the Project HW/SW integrated systemCore Industry Energy Other Industries Others Core Technology Internet of Things (IoT)Other Technologies Artificial Intelligence(AI)Sustainable Development Goals Affordable and Clean Energy, Industry, Innovation and Infrastructure, Responsible Consumption and ProductionRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	73300
Inverter	Equipment	1	43000	43000
Arduino	Equipment	2	1500	3000
Arm Cortex M4	Equipment	1	12500	12500
Relay	Equipment	4	100	400
NodeMCU	Equipment	1	1200	1200
Sensors	Equipment	4	1500	6000
Project Report Binding	Miscellaneous	1	3000	3000
Project Report Printing	Miscellaneous	1	200	200
Stationary	Miscellaneous	1	4000	4000