Article
Energy & Fuels
Martin Janos Mayer, Gyula Grof
Summary: Physical modeling plays a crucial role in forecasting the power production of grid-connected PV power plants. Different model chains can lead to significant differences in forecast accuracy, with irradiance separation and transposition modeling identified as the most critical calculation steps. Wind speed forecasts have only a marginal effect on PV power prediction accuracy.
Article
Energy & Fuels
Linfei Yin, Xinghui Cao, Dongduan Liu
Summary: This study proposes a weighted fully-connected regression network model, which reduces the photovoltaic power prediction errors by automatically selecting well-trained regression networks from multiple groups and configurations, without the need for additional sensors and data sources. Experimental results show that this method can reduce the mean absolute error of photovoltaic power prediction by at least 75.9954% compared to state-of-the-art methods and 68.2937% compared to 18 other famous convolutional neural network methods.
Article
Energy & Fuels
Bouchaib Zazoum
Summary: The study aimed to explore the relationship between input parameters and solar PV power using machine learning. Results showed that the Matern 5/2 GPR algorithm performed the best among the ML approaches considered. The proposed ML models were found to be suitable for predicting the power of different solar PV panels, with effectiveness and accuracy demonstrated through comparison using RMSE and MAE criteria.
Article
Engineering, Electrical & Electronic
Haojie Huang, Zhongmei Li, Xin Peng, Steven X. Ding, Weimin Zhong
Summary: This study introduces a method using modified likelihood to deal with output corrupted by noises, which is applied to both simulation experiments and practical industrial processes, demonstrating the effectiveness and stability of this approach.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2021)
Article
Computer Science, Artificial Intelligence
Jing Ren, Wei-Qin Li
Summary: Developing a robust forecasting method for time series is of great significance. The reliability and accuracy of the traditional model are reduced due to outliers in the series. This study proposes a robust maximum correntropy autoregressive (MCAR) forecasting model by examining actual power series of Hanzhong City, Shaanxi province, China. To reduce outlier interference, the local similarity between data is measured using the Gaussian kernel width of correlation entropy, and the semi-definite relaxation method is used to solve the parameters in the MCAR model. The results show that the MCAR model outperforms deep learning methods, with a 1.63% improvement in the average value of the mean absolute percentage error (MAPE). It was found that maximum correntropy is helpful for reducing the interference of outliers.
PEERJ COMPUTER SCIENCE
(2023)
Article
Green & Sustainable Science & Technology
Mehmet Yilmaz, Aliriza Kaleli, Muhammed Fatih Corapsiz
Summary: Photovoltaic systems are affected by changing irradiance, and it is important to use maximum power point tracking (MPPT) algorithms to ensure optimal operation. In this study, a two-stage MPPT method using gaussian process regression and super twisting sliding mode controller is proposed and compared with other algorithms. The results show that the proposed method outperforms other algorithms in different scenarios and leads to less oscillation at the maximum power point.
Article
Robotics
Lan Wu, Ki Myung Brian Lee, Liyang Liu, Teresa Vidal-Calleja
Summary: The Log-Gaussian Process Implicit Surface (Log-GPIS) is a novel continuous and probabilistic mapping representation suitable for surface reconstruction and local navigation. By applying a logarithmic transformation, accurate Euclidean distance field and implicit surface can be recovered through Log-GPIS. Experimental results show that Log-GPIS produces the most accurate results for the Euclidean distance field and comparable results for surface reconstruction.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2021)
Article
Energy & Fuels
Manel Marweni, Mansour Hajji, Majdi Mansouri, Mohamed Fouazi Mimouni
Summary: The majority of energy sources being used today are traditional types, which are limited in nature and quantity and continuously diminishing. Renewable energy is a compensation for this, but it is strongly dependent on climatic conditions. Energy management plays a crucial role in accurately calculating energy usage and understanding energy requirements and origins, with a focus on the forecasting phase using intelligent machine learning techniques.
Article
Thermodynamics
Connor Scott, Mominul Ahsan, Alhussein Albarbar
Summary: To mitigate carbon emissions from buildings, on-site renewable energy generation systems should be installed to supply energy without relying on the national grid. However, renewable generation sources are unreliable due to weather conditions. This research aims to compare benchmark machine learning algorithms for accurately forecasting photovoltaic (PV) generation at a university campus in central Manchester.
Article
Computer Science, Information Systems
Jinho Choi
Summary: This article explores the use of GPR and DAS for data collection in Internet-of-Things systems, demonstrating how active sensor selection and prediction feedback can improve data interpolation efficiency.
IEEE INTERNET OF THINGS JOURNAL
(2021)
Article
Engineering, Electrical & Electronic
Shengyuan Liu, Yicheng Jiang, Zhenzhi Lin, Fushuan Wen, Yi Ding, Li Yang
Summary: This paper proposes a two-step day-ahead electricity price forecasting algorithm based on the weighted $K$-nearest neighborhood (WKNN) method and the Gaussian process regression (GPR) approach. The algorithm detects price spikes in the first step and accurately forecasts electricity price in the second step. It is verified using actual market data and compared to existing algorithms to demonstrate its effectiveness.
JOURNAL OF MODERN POWER SYSTEMS AND CLEAN ENERGY
(2023)
Article
Automation & Control Systems
Haojie Huang, Yedong Song, Xin Peng, Steven X. Ding, Weimin Zhong, Wei Du
Summary: In this article, a sparse nonstationary GPR method is proposed to handle nonstationary relationships among samples, making the model more flexible and overcoming the stationarity assumption issues in existing methods. The performance of the proposed method is evaluated using public datasets and a sampled diesel engine dataset, showing its superiority in terms of accuracy.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2021)
Article
Mathematics
Pere Josep Pons-Vives, Mateu Morro-Ribot, Carles Mulet-Forteza, Oscar Valero
Summary: This paper proposes an improved algorithm, OWA-based K-means, for clustering customers based on their spending propensity. Experiments show that the use of OWA operator improves the performance of classical K-means significantly. The OWA-based K-means can be applied to classify customers in different seasons without requiring radical changes in the implementation of the classical method or additional implementation costs in real hotel management.
Article
Green & Sustainable Science & Technology
Ying Wang, Bo Feng, Qing-Song Hua, Li Sun
Summary: Solar power is considered a promising candidate for power generation to tackle climate change, and accurate short-term forecasting is crucial for its safe integration into the smart grid. The hybrid model combining LSTM and GPR, known as LSTM-GPR, shows superior performance in both point prediction accuracy and interval forecasting reliability compared to other algorithms.
Article
Energy & Fuels
L. Alfredo Fernandez-Jimenez, Claudio Monteiro, Ignacio J. Ramirez-Rosado
Summary: This article presents original probabilistic forecasting models for day-ahead hourly energy generation forecasts for a photovoltaic (PV) plant, based on a semi-parametric approach using three deterministic forecasts. The proposed models, which incorporate data from numerical weather prediction models and sun position variables, show better forecasting performance than the benchmark models in a real-life case study.
Article
Chemistry, Physical
Hanmin Sheng, Jian Xiao
JOURNAL OF POWER SOURCES
(2015)
Article
Automation & Control Systems
Hanmin Sheng, Jian Xiao, Peng Wang
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
(2017)
Article
Engineering, Electrical & Electronic
Wenjian Zhou, Sheng Yang, Li Wang, Hanmin Sheng, Yang Deng
Summary: The paper proposes a nonlinear calibration method based on sinusoidal excitation and DFT transformation, which achieves high calibration accuracy by using Fourier transform and cubic spline interpolation method to reduce calibration errors.
JOURNAL OF SENSORS
(2021)
Article
Engineering, Multidisciplinary
Lei Shi, Yuhua Cheng, Jinliang Shao, Xiaofan Wang, Hanmin Sheng
Summary: This paper introduces a new opinion dynamic model that explains the evolution of trust/distrust levels between agents due to opinion differences. It also conducts a theoretical analysis on opinion consensus and polarization with the presence of a stubborn leader, and further explores opinion dynamics in networks with multiple non-stubborn leaders. Numerical simulations are provided to illustrate the theoretical results.
IEEE TRANSACTIONS ON NETWORK SCIENCE AND ENGINEERING
(2022)
Article
Computer Science, Artificial Intelligence
Hanmin Sheng, Biplob Ray, Jinliang Shao, Dimuth Lasantha, Narottam Das
Summary: This paper proposes a Transfer Learning (TL) based PV energy yield model that can transfer knowledge from a source domain to a target domain, even when there are differences in time and location. The model improves the accuracy of solar energy forecasting and reduces dependence on large historical datasets.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Computer Science, Information Systems
Hanmin Sheng, Biplob Ray, Kai Chen, Yuhua Cheng