Research on Photovoltaic Power Prediction Method Based on Dynamic Similar Selection and Bidirectional Gated Recurrent Unit

Advanced Theory and Simulations, Journal Year: 2025, Volume and Issue: unknown

Published: March 8, 2025

Abstract Photovoltaic (PV) power generation is vital for sustainable energy development, yet its inherent randomness and volatility challenge grid stability. Accurate short‐term PV prediction essential reliable operation. This paper proposes an integrated method combining dynamic similar selection (DSS), variational mode decomposition (VMD), bidirectional gated recurrent unit (BiGRU), improved sparrow search algorithm (ISSA). First, DSS selects training data based on local meteorological similarity, reducing interference. VMD then decomposes into smooth components, mitigating volatility. The Pearson correlation coefficient used to filter highly relevant variables, enhancing input quality. BiGRU captures temporal evolution patterns, with ISSA optimizing key parameters robust forecasting. Validated historical Australian under diverse weather conditions, the proposed effectively reduces volatility, significantly improving accuracy reliability. These advancements support stable supply efficient

Language: Английский

---

Research on Precision Marketing and Smart Tourism Service Optimization of Online Marketing Driven E-commerce Platform Based on Big Data and Machine Learning

Applied Mathematics and Nonlinear Sciences, Journal Year: 2025, Volume and Issue: 10(1)

Published: Jan. 1, 2025

Abstract E-commerce platform occupies an important position in the modern economy, how to improve precision marketing effect of through online has become key development enterprises. At same time, smart tourism services also puts forward higher requirements for personalized recommendations and precise marketing. The widespread application big data technology machine learning methods provides new opportunities, making it possible optimize data-driven strategies. In this paper, we collect process from e-commerce users construct user profile model. model is used accurately categorize users, NSE strategies are developed implemented. Decision trees, random forests, support vector machines, LightGBM algorithms predict users’ purchasing behavior interest preferences. Meanwhile, service was generate a list Top-K attractions as recommendation provide services. empirical analysis results show that strategy based on these techniques can effectively conversion rate increase overall revenue platform, number orders after use increased 138 245, which 77.62%. Furthermore, level personalization been significantly enhanced. After structural equation analysis, be seen standardized coefficients influence experience behavioral intention perceived value 0.136 0.193 respectively while P-value less than 0.05, indicates positive tourists’ value. study shows combination strong technical optimization platforms services, help enterprises achieve objectives enhancement changing market environment.

Language: Английский

---

Compressive strength prediction of sleeve grouting materials in prefabricated structures using hybrid optimized XGBoost models

Construction and Building Materials, Journal Year: 2025, Volume and Issue: 476, P. 141319 - 141319

Published: April 15, 2025

Language: Английский

---

Concrete Carbonization Prediction Method Based on Bagging and Boosting Fusion Framework

Buildings, Journal Year: 2025, Volume and Issue: 15(8), P. 1349 - 1349

Published: April 18, 2025

Concrete carbonation is an important factor causing corrosion of steel reinforcement, which leads to damage reinforced concrete structures. To address the problem depth prediction, this paper proposes a prediction model. The framework synergistically integrates Bagging and Boosting algorithms, specifically replacing original Random Forest base learner with gradient variants (LightGBM (version 4.1.0), XGBoost 2.1.1), CatBoost 1.2.5)). This hybrid approach exploits strengths all three algorithms reduce variance bias, further improve accuracy, Bayesian optimization were used fine-tune hyperparameters, resulting in hybrid-integrated models: Forest–LightGBM Fusion Framework, Forest–XGBoost Forest–CatBoost Framework. These models trained on dataset containing 943 case sets six input variables (FA, t, w/b, B, RH, CO2). comprehensively evaluated using comprehensive scoring formula Taylor diagrams. results showed that model outperformed single model, RF–CatBoost fusion having highest test set performance (R2 = 0.9674, MAE 1.4199, RMSE 2.0648, VAF 96.78%). In addition, Framework identified exposure t CO2 concentration as most features. demonstrates applicability predictive based predicting carbonation, providing valuable insights into durability design concrete.

Language: Английский

---

Reliable protocol using gradient boosting decision tree with limited experimental data to modify membrane surface for enhanced resilience and nitrogen removal in biofilm system

Bioresource Technology, Journal Year: 2025, Volume and Issue: unknown, P. 132602 - 132602

Published: April 1, 2025

Language: Английский

---