Neural Model in Energy Management. An Industry 4.0 Component

Many different types of digital technology available nowadays for processing big data create an illusion that any issue can be resolved as it arises. In reality, though, this creates even more issues.

Production companies operating many different types of equipment face a challenging problem: How much electric power, steam, argon, nitrogen, blast furnace and natural gases, compressed air and hot water a site needs to maintain continuous production?

A production site houses hundreds and thousands of machines consuming different types of utilities. Proper forecasting and planning are directly related to the cost of the finished products. It is an important and challenging task. Artificial neural networks (ANNs) could be one of the solutions here. An artificial neuron is a function that converts a number of inputs into one output. ANN is a mathematical model that functions similarly to biological neural networks. The key advantage of neural networks over regular computing algorithms is their learning capacity. The more information they process the better they learn.  

At the moment, KONSOM GROUP in collaboration with Higher School of Economics (HSE) are engaged in the implementation of a prediction model that can analyze how much fuel and power a production site will need for short-term planning. As part of this work, KONSOM GROUP carried out an inspection, and HSE’s Laboratory of Big Data led by A. E. Ustyuzhanin developed a programming library using a neural model at KONSOM GROUP’s request. Right now, a prototype prediction model is being developed.

A connectionist model developed by our partner will become a major part of our energy management system. This particular module will be used to estimate a short-term need (for one to two days) for the utility.

The main contributors to this project include Yu. N. Volshchukov from KONSOM GROUP and A. E. Ustyuzhanin, a leader of HSE’s Laboratory of Big Data.