Artificial Intelligence based flood monitoring system
Floods are the most sudden and damaging natural hazards causing huge loss of lives, properties, and infrastructure. Acccording to NDMA Damages & Recovery Needs Assessment Report (2013), floods cause $0.44 Billion worth of losses in Pakistan. Timely prediction and alerts
2025-06-28 16:30:18 - Adil Khan
Artificial Intelligence based flood monitoring system
Project Area of Specialization Artificial IntelligenceProject SummaryFloods are the most sudden and damaging natural hazards causing huge loss of lives, properties, and infrastructure. Acccording to NDMA Damages & Recovery Needs Assessment Report (2013), floods cause $0.44 Billion worth of losses in Pakistan. Timely prediction and alerts of upcoming floods can be criticial in protecting phyiscal assets and human lives in Pakistan. Timely prediction of floods is complex due to abrupt change in climate conditions and other manmade factors. Flood prediction and forecast depends on a number of hydrological and meteorological factors. Understanding the interrelated processes involved in occurrence time, magnitude and duration of floods is very important for the design, construction and operation of hydraulic structures and flood control projects.
The damage caused by floods can be reduced if not entirely eliminated. Physical models were used previously for flood prediction but the limitation of physical models are that it requires in depth knowledge of underlying structure and are not accurate enough for short term flood prediction like flash floods. These short comings encourage use of data driven models like machine learning models. Machine learning which is a subfield of Artificial intelligence amid to find the underlying pattern within given data and develop a generalized model. These models do not require knowledge of underlying patterns and can be used on unseen data of same type.
Our aim here is to developed a machine learning model for early and timely flood prediction model, to save human lives and infrastructure. Our model will consist of wireless sensor network (WSN) for collection of data related to flood occurrence. These sensors include water level sensor for rivers water level, Water flow sensor for river flow speed, Temperature sensor, Humidity sensor, Air pressure sensor and Rain drop sensor. Microcontroller module Arduino will be used with these sensors for collection of data from sensors. After collecting data communication modules such as GSM or Zigbee will be used for transmitting data from sensors to database where data will be stored and used further. GSM will be used in areas where we have mobile signals coverage. In case no covrage Zigbee will be used. Data cleansing will be performed on stored data to remove inaccurate data.
Different machine learning models will be trained and tested for accurate early flood prediction. Machine learning model with high accuracy will be selected to be used for prediction. Data coming from sensors will be fed into machine learning model as input for prediction.
Website and application will be developed which will show separate data coming from every sensor and a graph showing the current, past and predicted water levels. An alert notification will be sent on the website and mobile application of any upcoming flood in the user's nearby area, as predicted by the machine learning model.
Project Objectives- To devolop sensors for monitoring the following: a) water level b) water flow c) environment temperature d) air pressure e) humidity f) rain
- Develop a Wireless Sensor Network for data accumulation. Both GSM to Central Database, and Radio based Zigbee ad-hoc network will be experimented with.
- To develop and compare performance of Machine learning algorithms for early flood event pridection.
- To develop website and mobile application for alert system and data visualization.
This project will implement machine learning method for prediction of flood. The cause of floods and impact of cause of flood will be studied. Wireless sensors will be used for the collection of hydrological and meteorological parameters data collection. The wireless sensor network will have three major components; Sensor nodes, Communication nodes and Central Database.
Sensors node include water level sensor, water flow sensor, temperature sensor, humidity sensor, air pressure sensor, rain sensor, microcontroller and a communication module. Arduino uno microcontroller will be used to aquire data from sensors within the node, perform data compression operations, and will be interfaced with communication modules (GSM or Zigbee) to route the data to a Central Database. Multiple sensor nodes will installed across water streams in flood prone areas.
Intermediate communication nodes will be used in case of remote areas, where GSM signal strength is weak. Here Zigbee modules to develop a ad-hoc network with Zigbee-to-Zigbee and Zigbee-to-GSM communication to transmit data to a node where there are strong GSM signals. Nodes with strong GSM signal strength will be responsible for communication with the Central Database. Thus intermediate communication nodes will comprise of a Arduino Uno microcontroller, power supply, and a communcation module.
After data collection on a Central Database, data cleansing or cleaning will be performed on stored data to remove corrupt and in accurate records. This data will be then used for training machine learning algorithm and then live data will be used as input to algorithm for early flood prediction. Different machine learning algorithms will be trained and tested and the one with better accuracy will be selected.
After developing machine learning technique website and mobile application will be developed for our alert system of early warning notification to concern authorities and public. Website and mobile application will show the data coming from every sensor individually and graphs showing current water levels and predicted.
Benefits of the ProjectThe following NDMA Damages & Recovery Needs Assessment Report (2013) shows damages caused by floods:
Our project which aims to deliver timely alerts of upcoming floods can help reduce these damages by:
- Alerting communities to protect valuables.
- Timely evacuation of areas to protect human lives.
- Crop and Livestock protection, by informing farmers to move their harvested goods to higher grounds.
- Reduce cost of Disaster management by taking precautionary measures instead of only damage reimbursements.
- Detecting flood-prone areas for future urban and rural planning.
The final deliverable will comprise of the following:
- Sensor Nodes: The sensor nodes will be able to measure aforementioned features correlated to flood events, process the data, and transmit data to another communication node or directly to Central Database.
- Communication Node: Modules with GSM or Zigbee that can recieve and transmit data.
- Machine Learning Algorithm for Flood Detection: Code that implements a pretrained flood detection model that can be embedded into Web applications.
- Web and Mobile application: For data visualization and alert system.
| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 79860 | |||
| GSM module | Equipment | 3 | 3500 | 10500 |
| arduino | Equipment | 6 | 800 | 4800 |
| water level sensor (JSNSR04) | Equipment | 4 | 1200 | 4800 |
| water flow (YFS201) | Equipment | 4 | 550 | 2200 |
| Temperature Sensor | Equipment | 4 | 300 | 1200 |
| Pressure sensor | Equipment | 4 | 320 | 1280 |
| Humidity sensor | Equipment | 4 | 570 | 2280 |
| Zigbee | Equipment | 3 | 3600 | 10800 |
| Domian and hosting for web app | Equipment | 1 | 6000 | 6000 |
| Robust IP67 Physical Structure of Nodes | Equipment | 1 | 15000 | 15000 |
| Electronic prototyping(solders, wires etc) | Equipment | 1 | 8000 | 8000 |
| 5V-2A Batteries and Power Supply | Equipment | 5 | 600 | 3000 |
| Shipment, Papers, Stationary etc | Miscellaneous | 1 | 10000 | 10000 |