Utility: Automatic forecasting and prediction of water and hazard level based on Artificial Intelligence

Case study: Artificial Intelligence based real time forecasting of a Dam's Water Level and prediction of its Hazard Level

South Africa (SA) has an extensive infrastructure of dams. Publications from the Water and Sanitation Department of SA illustrate that in 2016, there were 5,226 registered dams in the country. Investment in such huge infrastructure is necessary as the country receives one of the least rainfalls in the world and with abundant sunshine, it is able to hold very little water in its ground. Thus these dams are required for conserving water that can be used for industry, agriculture and domestic purposes.

Given the present shortage of technical skill in SA, these APPS are very few in number. One estimate shows that currently there are less than 100 APPs in the country, and with such a low number of APPs, safety inspection of dams is a challenge. Additional statistics, due to the less number of APPs, in 2014-2015, only 58% of the targeted numbers of dams could be inspected.

Our work addresses the problem of low number of APPs by capturing a dam’s water level at regular intervals using ‘Internet of Things’ devices and then storing the data on a server. Once the developed system is setup, it is fully automatic and needs no human intervention in terms of calibration, measurement, etc.

The data on the server is next used to forecast in real time, the future water level of a dam using two families of time series models. Considering large fluctuations in the mean and variance due to (multiple) seasonality and trend in the data, two families of algorithms are compared to get the best results. They are the conventional statistical time series (Auto Regressive Integrated Moving Average and Exponential Smoothing) algorithms and the neural network (Recurrent Neural Network – Elman, Jordan and Long Short Term Memory) based algorithms. The forecasts (for a considerable time window) can help APPs in reviewing safety protocols of the concerned dam and take precautionary steps, if any are required.

Using available historical data, single and ensemble supervised decision tree based AI models are next developed. The AI model takes inputs like the basic characteristics of the dam such as (wall height, crest length, surface area etc.) inputs into the model. The model then predicts a dam hazard level. This AI based software system will in real time predict a dam’s hazard level (High, Significant, and Low). Robustness of the solution is ensured by choosing the algorithm which gives the minimum test error over a large number of folds of cross validation.

Also as the larger dams are more than 30 years of age, infrastructural integrity of the dams have to be ensured through adherence to a long list of safety regulations. Safety regulation checks are conducted by Approved Professional Persons (APPs). These APPs are mostly professionally certified engineers, technologists and technicians. One of the primary functions of these individuals is to regularly check the various parameters of the dams, and also classify its hazard level (High, Significant, and Low).

Nature-mountain-river-green-trees-green-wallpaper-hd

Want To Be Ahead In The 4th Industrial Revolution?

Home

Artificial Intelligence Group (AI Group) is a developer and marketer of AI and big data solutions company.

Big Data

Big Data is the proliferation of structured and unstructured data that floods your organization on a daily basis – and if managed well, it can deliver powerful insights.

Artificial Intelligence

Artificial Intelligence is the branch of computer science concerned with making computers behave like humans, i.e. with automation of intelligent behaviour.

Contacts

Tell: +27 10 035 2335

Address:
150 Rivonia Road
Sandton
2057
South Africa

© Copyright Artificial Intelligence Group 2019

Close Menu