Flood prevention using AI
Machine learning, if used correctly can solve an array of problems including flood prevention.
We at ML Sense, combined our knowledge and expertise to produce a data product that can successfully predict floods on UK highways. Here’s the complete success story.
Floods have been a consistent problem for Highways in the UK. Not only do they occur regularly, but they also occur without warning and are often reported late. This is extremely dangerous and cannot be ignored.
Currently, there are two ways of flood prediction; guessing or watching through a camera feed attached to some motorways. Both of these methods are outdated to accurately predict floods.
Therefore, to solve this problem, we used state-of-the-art methods in data engineering and artificial intelligence to gather three different data sources:
- Conditions of pipes data.
- Historical weather data.
- Historical flood index data.
By combining these data sources, we created a week-in-advance flood warning system. This helped us predict the flooding hotspots and report them before there was likely to be any flood. Recurring hotspots or the ones most likely to flood were dealt with first, which gave our clients a grip on a staggering problem.
Our client had been collecting data on pipe conditions and the historical flood index for many years now. This data was used for a different purpose altogether. However, with the realization that this gold mine of data could be used for something much more powerful, our client was thrilled and immediately saw greater value in this dataset and other similar internal datasets.
We went through a set of steps to arrive at a fruitful solution, here is the breakdown:
- We first scanned through all the data and identified the data that was useful. Thereafter, we started the cleansing and transforming part.
- Once the data was ready, we applied the machine learning algorithms.
- After training the algorithms, we were ready to use the solution for predicting flooding hotspots.
- The next step was to find the flooding hotspots. The way this works is we put the latest weather forecast into our machine learning model. This model returns results of dangerous hotspots. Therefore, the most dangerous and the most occurring hotspots are attended to.
- The final step was to set up the whole system for continuous improvement. Essentially, after a flooding hotspot has been cleaned, we need to update our model. This allows us to consistently deliver top results.
You cannot control the weather but you can surely predict it and take the necessary steps to minimize the weather disasters. We have the proof right here:
- 80% more floods were predicted than before at UK highways.
- Flood predictions were received 7 days in advance.
- Zero devices were installed on the roadside for monitoring.
Highway agencies need a smart solution for the flood prevention problem. Moreover, this solution needs to be as lightweight as possible in terms of attaching sensors.
Data Engineering and Artificial Intelligence have proven to address many business problems by using data sets. First, we identify the problem and then ask the right questions from the data to arrive at a solution to the given problem. As a matter of fact, AI algorithms have great predictive powers if provided enough data.
To sum up; we combined the domain knowledge, data already owned by Highways England, and our Artificial Intelligence expertise to produce a data product. This data product takes weather as an input and flooding hotspots as an output, which helps our client stay prepared and equipped for any future flood. What’s great is that it is designed to stay up to date; which essentially means, daily weather data is consumed automatically.