Scientists from South Korea have developed a heat pump optimization technique that controls the secondary refrigerant flow. The proposed technology offers the advantage of bypassing the need to control the compressor of the heat pump, which is a function that is commonly not available in commercial devices.
Article Source: pv magazine
Photo Credit: Nenad Stojkovic (Flickr)
Scientists from the Pusan National University in South Korea have developed an artificial neural network (ANN)-based optimum control logic (OCL) system for optimizing performance and operation of air source heat pumps (ASHPs). The proposed technology offers the advantage of bypassing the need to control the compressor of the heat pump.
In the paper “Performance improvement of air-source heat pump via optimum control based on artificial neural network,” published in Energy Reports, the Korean group explained that the OCL system is able to identify the parameters that increase the system performance, without changing the main components of the system, which in turn reduces system costs.
“An ANN model was developed to prevent overfitting errors, by constructing a training dataset in the form of synthetic time-series load data,” they added, noting that overfitting errors occur when the model fits too closely to the training dataset instead. “Machine learning algorithms can render good predictions on the data used to train the model but may be limited by overfitting and not generalizing adequately according to new data.”
Previous models for optimizing ASHPs were based on controlling the compressor, which the scientists said is not an in-field valid solution, as heat pump manufacturers tend to limit this function in their products. Their optimization technique, by contrast, is based on the secondary-side working fluid of the heat pump, which can be controlled after conducting minor changes to a standard system.
In a heat pump, the secondary refrigerant is the fluid that acts as an intermediary to transfer heat between the primary refrigerant and the space that needs to be heated or cooled. The researchers used water as the secondary refrigerant controlled by the ANN.
In order to train the model, the scientists provided it with some 762,000 data points based on a real ASHP system located in a building facility in the South Korean city of Busan. The data was then verified using some 254,000 data points from the same facility, and was then tested against an identical number of data points.
The academics compared the performance of the AI-assisted ASHP with that of a benchmark ASHP without the ANN. They found that novel control technique can help improve significantly the heat pump coefficient of performance (COP). “The improvement in the energy performance of the optimum models with respect to the conventional models was: 1.52% and 3.58% for the cooling and heating system COP, respectively, and 0.76% and 0.81% for the heat pump COP,” they specified.
The researchers also conducted an economic analysis of the AI-modified heat pump system. According to their results, the payback time of that novel system is 11.6 years, while the system’s lifetime is 15 years, which they claim makes it “possible to recover the initial investment during the life cycle.”