Integration of machine learning to increase steam turbine condenser vacuum and efficiency through gasket resealing and higher heat extraction into the atmosphere

Abstract

In large thermal power facilities, thermal machines mainly fitted with steam turbines are used in thermodynamic process of converting thermal energy into mechanical work. In order to improve the efficiency of heat energy conversion into mechanical work, the increase of vacuum level in the condenser of a steam turbine was analyzed using the machine learning method. The vacuum in the steam turbine condenser increases after replacing the labyrinth gaskets with honeycomb ones, resealing the vacuum system and increasing the amount of cooling water through the condenser. The latter is controlled by algorithm, which is integrated into the steam turbine distribution control system. As a result of the increased vacuum in the steam turbine condenser, the enthalpy difference of the steam turbine low-pressure cylinder increases, and the steam turbine performs more mechanical work resulting in higher conversion efficiency. The drawback of the increased vacuum in the turbine condenser is that the flow of cooling water through the steam turbine condenser increases as well as heat dissipation from the condenser into the environment to be replaced by regenerative condensate heating. The results of the analysis show that the overall balance of the vacuum increase in the steam turbine condenser is positive, the steam turbine generates more power, and consequently the process efficiency increases by as much as 2.48 %.

Design Considerations of Low Bypass Ratio Mixed Flow Turbofan Engines with Large Power Extraction

“…One example is provided in [33]. The use of machine learning and neural networks to create more accurate estimations and models is becoming more common, both for propulsion purposes [34] and for other fields of power generation [35].…”

Octovalve Thermal Management Control for Electric Vehicle

“…This is acceptable as the purpose of this paper is to present the pros and cons of the architecture and the way it operates. Keeping parameters similar makes for a better comparison and mitigates the need for validation as used in the literature [26,27].…”

Remaining Useful Life Prediction Method for Bearings Based on LSTM with Uncertainty Quantification

“…Ali et al [10] introduced the WD (Weibull distribution) to the SFAM (simplified fuzzy adaptive resonance theory mapping) neural network to predict the RUL of rolling bearings. Using an artificial neural network unit (ANN) with multiple hidden layers, Strušnik et al [11] constructed a simulation model to predict system performance based on real data. Kim et al [12] proposed a CNN-based (convolutional neural network, CNN) prediction model to reflect the correlation between RUL estimation and a health status detection process.…”