Researchers have now developed a simple and inexpensive approach to deep learning with the “model predictive control” method – which can reduce energy consumption in buildings by 50 percent.
It is an advanced method for controlling a process within the framework of a series of set restrictions. “Model predictive control” (MPC) has been used in the chemical industry and by oil refineries since the 1980s – but now the solution is on the rise in other areas.
With the help of MPC, for example, it would be possible to reduce the energy consumption of buildings by up to half, but previously this required large investments in the form of new software and increased computer capacity. Now, however, the Pacific Northwest National Laboratory, PNNL, has developed a new approach to deep learning. It writes PNNL on its website.
An artificial neural network usually requires large amounts of data and powerful computers for the deep learning needed to build artificial intelligence. But the method that PNNL uses to train the MPC is based on the data and physical knowledge that the building already has. The algorithm learns to make optimal decisions much faster and based on a much smaller amount of data than the conventional neural networks available.
The method trains a neural control policy that can be installed on existing hardware. You do not have to invest in new software or increased capacity – and the solution means that the person who handles the optimization of the building’s energy consumption does not need to be an expert in the field at all.
The method can be used on all types of buildings. But so far, researchers have only released a prototype code for the purpose.
– The MPC method for deep learning reduces costs and contributes to a faster design and installation of control systems compared to traditional methods, says Draguna Vrabie at PNNL’s Advanced Computing, Mathematics, and Data Division, on the research institute’s website.
And there are large amounts of energy that can be saved. Residential and commercial buildings account for close to 40 percent of the United States’ total energy use. Of that amount, 40 percent goes to heating, ventilation and air conditioning, areas where there may be good margins for adjustments.