Search

Combined Cycle Power Plants (CCPPs) generate electrical power through gas turbines and use the exhaust heat from those turbines to power steam turbines, resulting in 50% more power output compared to traditional simple cycle power plants. Predicting the full-load electrical power output ( P E ) of a CCPP is crucial for efficient operation and sustainable development. Previous studies have used machine learning models, such as the Bagging and Boosting models to predict P E . In this study, we propose employing Super Learner (SL), an ensemble machine learning algorithm, to enhance the accuracy and robustness of predictions. SL utilizes cross-validation to estimate the performance of diverse machine learning models and generates an optimal weighted average based on their respective predictions. It may provide information on the relative contributions of each base learner to the overall prediction skill. For constructing the SL, we consider six individual and ensemble machine learning models as base learners and assess their performances compared to the SL. The dataset used in this study was collected over six years from an operational CCPP. It contains one output variable and four input variables: ambient temperature, atmospheric pressure, relative humidity, and vacuum. The results show that the Boosting algorithms significantly influence the performance of the SL in comparison to the other base learners. The SL outperforms the six individual and ensemble machine learning models used as base learners. It indicates that the SL improves the generalization performance of predictions by combining the predictions of various machine learning models. [ABSTRACT FROM AUTHOR]

In order to maximize the profit of a combined cycle power plant, it is important to accurately predict its full load power output. When a combined cycle power plant operates, the exhaust gas produced by the previous stage is used to drive the next stage heat engine to drive the generator, and its full-load electrical output is affected by ambient temperature, atmospheric pressure, relative humidity and exhaust pressure. Firstly, the kernel principle component analysis (KPCA) was used to perform feature combination dimensionality reduction for power generation related features to complete feature extraction. Then, the extreme gradient boosting algorithm (XGBoost) was used to score feature importance and the optimal feature subset was obtained by forward selection (FS). Finally, a KPCA-XGB-FS model was constructed for the prediction of hourly power output under full load of combined cycle power plants. By experimenting with real data from a combined power station and comparing with existing research methods using the same data, it is found that the method proposed in this paper can effectively predict the power output, and is better than the existing research methods.

Combined Cycle Power Plants (CCPPs) generate electrical power through gas turbines and use the exhaust heat from those turbines to power steam turbines, resulting in 50% more power output compared to traditional simple cycle power plants. Predicting the full-load electrical power output (PE) of a CCPP is crucial for efficient operation and sustainable development. Previous studies have used machine learning models, such as the Bagging and Boosting models to predict PE. In this study, we propose employing Super Learner (SL), an ensemble machine learning algorithm, to enhance the accuracy and robustness of predictions. SL utilizes cross-validation to estimate the performance of diverse machine learning models and generates an optimal weighted average based on their respective predictions. It may provide information on the relative contributions of each base learner to the overall prediction skill. For constructing the SL, we consider six individual and ensemble machine learning models as base learners and assess their performances compared to the SL. The dataset used in this study was collected over six years from an operational CCPP. It contains one output variable and four input variables: ambient temperature, atmospheric pressure, relative humidity, and vacuum. The results show that the Boosting algorithms significantly influence the performance of the SL in comparison to the other base learners. The SL outperforms the six individual and ensemble machine learning models used as base learners. It indicates that the SL improves the generalization performance of predictions by combining the predictions of various machine learning models.

Copyright of Power Generation Technology is the property of Power Generation Technology Editorial Office and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Here, we present the performance of a thermoelectric (TE) module consisting of n-type (La0.12Sr0.88)0.95TiO3 and p-type Ca3Co4−xO9+δ materials. The main challenge in this investigation was operating the TE module in different atmospheric conditions, since n-type has optimum TE performance at reducing conditions, while p-type has optimum at oxidizing conditions. The TE module was exposed to two different atmospheres and demonstrated higher stability in N2 atmosphere than in air. The maximum electrical power output decreased after 40 h when the hot side was exposed to N2 at 600°C, while only 1 h at 400°C in ambient air was enough to oxidize (La0.12Sr0.88)0.95TiO3 followed by a reduced electrical power output. The module generated maximum electrical power of 0.9 mW (∼ 4.7 mW/cm2) at 600°C hot side and δT ∼ 570 K in N2, and 0.15 mW (∼ 0.8 mW/cm2) at 400°C hot side and δT ∼ 370 K in air. A stability limit of Ca3Co3.93O9+δ at ∼ 700°C in N2 was determined by in situ high-temperature x-ray diffraction. [ABSTRACT FROM AUTHOR]

Perfect power system voltage stability is not possible in practice. Generally, the power grid is continually exposed to changes in its load and operating conditions. Therefore, dynamic stability analysis is one the most important and effective elements for greater security, stability and reliability of planning, design, operation and economic aspects of electric power networks. This study investigates and reports on the dynamic stability of the power system with a large‐scale photovoltaic system (L‐S PV). Two different scenarios with centralised PV power plants are considered in the medium voltage level without voltage regulation capabilities. Simulation results with these scenarios will show how the voltage instability decreases with the L‐S PV based on the bus status, disturbance location, and disturbance duration. In addition, the study discusses network terminal voltage, generator's rotor angle, generator's terminal current, generator's reactive power and electrical power output. This study is an extension of the earlier published conference paper. [ABSTRACT FROM AUTHOR]

Partial shading of a photovoltaic (PV) installation has an inconsistent impact on power production. This study investigates the effect of partial shading on PV performance. The experiments were carried out with a 90‐W PV module under both variable and constant irradiations with shaded area increased from 0 to 80% to observe the effect of variable solar radiation at certain shading points. The effect of shading under irradiation levels from 300 to 800 W/m2 was investigated. At a 600 W/m2 irradiation level, the shading impact factor was 1.25 with a 25% shading, while at 75% shading the impact factor decreased to 0.86. Results also show that for every 100 W/m2 increase in irradiation level, the electrical power output was enhanced by 3.89, 3.37, 2.27, and 2.02 W at 0, 25, 50, and 75% shading, respectively. The efficiency level was increased by 0.29, 0.27, 0.25, and 0.22% at 0, 25, 50, and 75% shading, respectively. Increasing the shaded area by 10% causes a 12.41 W drop in power output and a 2.3% drop in electrical efficiency. Partial shading not only deteriorates the PV performance, but also causes long‐term degradation of the module. [ABSTRACT FROM AUTHOR]

In this study, a long-term assessment of the wave energy resource potential for the Australian southeast shelf is performed from deep to shallow water, based on a 31-year wave hindcast. The hindcast, covering the period from 1979 to 2010, has been performed at high spatio-temporal resolution with the wave energy transformation model SWAN using calibrated source-term parameters. The model has been applied with a variable spatial resolution of up to approximately 500 m and at 1 h temporal resolution and driven with high-resolution, non-stationary CFSR wind fields and full 2D spectral boundary conditions from WaveWatch III model. Model validation was conducted against wave measurements from multiple buoy sites covering 10–31 years and showed a relatively high correlation between hindcast and measured significant wave height ( H s ) and mean wave direction ( θ m ). Maps of wave power resource distribution for annual and seasonal mean potential were generated along with the maps of resource reliability and variability. The high resolution allowed us to perform in-depth analysis of wave power characteristics, providing resource knowledge on seasonal and longer-term variability necessary for reliable and optimal design of wave technology. The most promising area for wave power exploitation was found to be the central coast of New South Wales, where various high-energy hotspots were selected for a further analysis. For each of the considered hotspots, the wave power magnitude, variability and consistency were carefully assessed and characterized by means of sea state parameters and mean wave directions. Finally, estimates of electric power outputs from different types of pre-commercial wave energy converter devices were drawn for each hotspot based on the wave data hindcast and discussed. [ABSTRACT FROM AUTHOR]

• Numerical model for a four-layer CdTe-based VPV glazing was developed. • The model was validated experimentally with a U-value of 0.92 W/(m2⋅K). • About 60% heat loss and heat gain can be reduced by VPV in cold and hot climates. • More than 40 kWh/m2 average power generation of VPV windows can be achieved. Vacuum-photovoltaic (VPV) glazing has attracted much attention due to its excellent thermal insulation performance and its ability to utilize solar energy. However, few simulation models have been established based on actual products and rarely have been validated by experiments. In this paper, a four-layer CdTe-based VPV glazing was developed and the corresponding numerical heat transfer model was established with the integration of a dynamic power generation model. The numerical model was then validated against both the results from the WINDOW program and a guarded hot box experiment. Afterward, the validated model was employed to analyze the energy and power generation performance of the VPV glazing in diverse climate zones in China with Harbin, Beijing, Changsha, Guangzhou, and Kunming used as representative cities. The numerical simulation results indicate that the U-value of the proposed VPV glazing is 0.89 W/(m2⋅K), which is in good agreement with the experimental results. Compared with a normal double glazing, the average energy reductions achieved with VPV glazing in air conditioning seasons are 128 kWh/m2, 23 kWh/m2, 45 kWh/m2, and 52 kWh/m2 in Harbin, Beijing, Changsha, and Guangzhou, respectively. In addition, the average annual power outputs of VPV glazing in Harbin, Beijing, Changsha, Guangzhou, and Kunming are 47 kWh/m2, 48 kWh/m2, 34 kWh/m2, 36 kWh/m2, and 45 kWh/m2, respectively. The numerical model developed in this study can be used for energy-saving potential analysis and optimization of VPV glazing in different meteorological conditions, the results of which could provide guidance for the effective application of VPV glazing. [ABSTRACT FROM AUTHOR]