banner

Selecting AI Use Cases in Energy and Commodity Industry

Artificial Intelligence (AI) has come of age in the past few years, doing several tasks from predicting frauds to driving cars. There is barely any industry left which hasn’t been touched by it – from Banking and Financial Services to Recruitment, Education, Retail and several others. Energy and Commodity industry too has started adoption of various Machine Learning & Deep Learning methods and technologies which will enable them to use AI in solving their problems better.

Majority of the companies in the Energy and Commodity Industry struggle with issues which are not very common with other industries – like which use cases are right for predictive analytics, and how to derive the most benefit out of those projects. In this post, we deal with this critical question, which is the process of selecting Use Cases for AI.

A company should ideally segregate the use cases it has picked for applying AI into the following 2 categories:

  • The Go Zone: These are those use cases which have proven results published by the industry leaders / AI vendors. The company need not do much research on whether to go ahead with these use cases or not, rather, just figure out how to go about implementing these. Some examples of such use cases include Yield Prediction, Supply and Demand Prediction, Machine Failure Prediction, Journal Anomaly detection, Supply Chain and Logistics, Scheduling, etc.

The data around such use cases is usually quite good in terms of volume, though in some cases it may be unorganized or even unstructured. Risk Edge has built and deployed many solutions around such use cases for large global companies, and we’ve found that the accuracy and acceptance of such use cases within the organization is quite good. In some cases where we found the data to be unorganized, we have devised separate algorithms to organize and structure the data, before applying machine learning / deep learning algorithms to them.

A list of successful case studies in AI for Energy / Commodity companies done by Risk Edge can be downloaded as an E-Book from here: E-Book – AI Use Cases for Energy and Commodity Trading Companies

  • The Test Zone: There are certain use cases which have still not seen much success in organizations, but might be well worth trying. Examples of such use cases include Price Prediction, optimizing profitability in trading which includes real-time updates on logistics, prices, weather, etc. Some of these use cases can be tricky and the organization must be crystal clear in terms of what the desired output / accuracy level is from such a use case. These kinds of use cases must be initiated and monitored under a controlled POC (Proof-of-Concept) environment which can be set-up easily and at a low cost. The data too in some cases like these is sparse, but can be overcome to some extent by using statistical sampling techniques or by deploying other sources of data.

Once the POC stage is set and the results verified and found acceptable, the POC can be put into a full-blown system development project. However, if the results are not found to be satisfactory as per management’s expectations, the company can then stop this POC and move on to the next POC.

We’ve seen that companies that segregate their Predictive Analytics initiative in such buckets, see much faster adoption and better RoI for these projects. Over time, the team gets accustomed to an assembly-line style of project execution, wherein, the internal research team points towards potential POCs, the Test Zone team works with vendors to build POCs, successful AI POCs are moved to next stage of system development and deployment, and join hands with already under-execution AI projects to bring about greater synergies across divisions for the entire organization.