Enhancing investment performance of Black-Litterman model with AI hybrid system: Can it be done?

Expert Systems with Applications, Journal Year: 2023, Volume and Issue: 244, P. 122924 - 122924

Published: Dec. 22, 2023

Language: Английский

---

Investigating Periodic Dependencies to Improve Short-Term Load Forecasting

Energy Engineering, Journal Year: 2024, Volume and Issue: 121(3), P. 789 - 806

Published: Jan. 1, 2024

With a further increase in energy flexibility for customers, short-term load forecasting is essential to provide benchmarks economic dispatch and real-time alerts power grids. The electrical series exhibit periodic patterns share high associations with metrological data. However, current studies have merely focused on point-wise models failed sufficiently investigate the of series, which hinders improvement accuracy. Therefore, this paper improved Autoformer extract learn representative feature from deep decomposition reconstruction. In addition, novel multi-factor attention mechanism was proposed handle multi-source numerical weather prediction data thus correct forecasted load. also compared model various competitive models. As experimental results reveal, outperforms benchmark maintains stability types consumers.

Language: Английский

---

A Short-Term Parking Demand Prediction Framework Integrating Overall and Internal Information

Sustainability, Journal Year: 2023, Volume and Issue: 15(9), P. 7096 - 7096

Published: April 24, 2023

With the development of smart cities and transportation, can gradually provide people with more information to facilitate their life travel, parking is also inseparable from both them. Accurate on-street demand prediction improve resource utilization management efficiency, as well potentially urban traffic conditions. Previous methods seldom consider correlation between a road section its surroundings. Therefore, in order capture temporal spatial dimensions carefully possible enrich relevant features model so achieve accurate results, we designed structure that considers different two perspectives: overall internal. We used gated recurrent units (GRU) extract influences dimension. The GRU combination graph convolutional neural network (GCN) influencing factors Additionally, detailed representation express dimensional features. Then, based on historical extracted using encoder–decoder, fuse spatio-temporal them finally obtain an combining internal information. By them, integrate prediction. performance evaluated by real data Xiufeng District Guilin. results show proposed achieves good compared other baselines. In addition, design feature ablation experiments. Through comparison find each considered important

Language: Английский

---

A Novel Air Pollutant Concentration Prediction System Based on Decomposition-Ensemble Mode and Multi-Objective Optimization for Environmental System Management

Systems, Journal Year: 2022, Volume and Issue: 10(5), P. 139 - 139

Published: Sept. 3, 2022

With the continuous expansion of industrial production scale and rapid promotion urbanization, more serious air pollution threatens people’s lives social development. To reduce losses caused by polluted weather, it is popular to predict concentration pollutants timely accurately, which also a research hotspot challenging issue in field systems engineering. However, most studies only pursue improvement prediction accuracy, ignoring function robustness. make up for this defect, novel pollutant (APCP) system proposed environmental management, constructed four modules, including time series reconstruction, submodel simulation, weight search, integration. It not realizes filtering reconstruction redundant based on decomposition-ensemble mode, but search mechanism designed trade off precision stability. Taking hourly PM2.5 Guangzhou, Shanghai, Chengdu, China as an example, simulation results show that APCP has perfect capacity superior stability performance, can be used effective tool guide early warning decision-making management

Language: Английский

---

An Intelligent Adversarial Deep Forecasting Model for Load Demand Using Hybrid Modified DA-GAN

2022 IEEE/IAS Industrial and Commercial Power System Asia (I&CPS Asia), Journal Year: 2023, Volume and Issue: unknown, P. 1340 - 1345

Published: July 7, 2023

Extreme weather conditions can have a significant impact on the electricity load demand and energy management programs thus cause unexpected blackouts in systems. To predict extreme conditions, it is important to consider different historical data analysis. This paper proposes an intelligent adversarial model for prediction of consumers' electric condition. By analyzing past events associated demand, new predictive deep learning developed that be used estimate future conditions. The proposed constructed based generative network (GAN) dragonfly algorithm (DA) make precise prediction. generator trained produce are similar data, while discriminator correctly classify real from generated ones. A modified DA suggested enhance GAN training through iterative process. dataset California over years 2015–2020 examine accuracy model.

Language: Английский

---