Data-driven operation performance evaluation of multi-chiller system using self-organizing maps

Data Analytics for Performance Evaluation Under Uncertainties Applied to an Industrial Refrigeration Plant

“…An initial work addressing some of these problems was previously presented [22], nevertheless many deficiencies were detected that led to the development of this improved proposed methodology, including a new reduced set of analyzed variables, a different preprocessing, the introduction of an outlier detection filtering and an uncertainty detection stage. Therefore, the potential savings can be more accurately measured instead of expecting misleading maximum COP gain values which would benefit any posterior control modification by proposing a reliable target.…”

“…Industrial refrigeration plants are mainly composed by compressors, condensers, expansion valves and evaporators in order to perform the vapor-compression cycle. As previously presented in [22], the tested installation, from the company Corporación Alimentaria Guissona, S.A., contains all the aforementioned characteristics to remove the heat from their facilities. The refrigerant (R717), in the state of saturated vapor, is compressed employing three compressors in parallel, two of them are the same model and the third has less capacity.…”

“…Data when the plant was stopped or out of range values was not taken into account. In comparison with the previous works [22], a reduced set of three relevant variables in order to analyze the plant performance is selected according to the plant experts. The magnitudes studied from the refrigeration plant are the compressors cooling capacity (Qc) in kW, the suction pressure (Sp) in bar and the wet bulb temperature (Twb) in • C, obtained according to [24].…”

Machine learning for anomaly detection and process phase classification to improve safety and maintenance activities

“…The energy sector is deeply studied because a lot of data is collected from resources distributed over large areas, such as in wind turbines [36,37] and photovoltaic plants [38], or in production plants [39,40]. Some other recent industrial applications of anomaly detection can be found in the chemical sector [41][42][43][44], in the food industry [45], ultraprecision machining processes [46] or for automated quality control in manufacturing processes [47]. Recently, an anomaly detection method has been successfully applied in detecting industrial control system cyber-attacks [48], linking security to the monitoring of industrial process data.…”