Energy forecasting of generation, demand, sources, and prices over short-time horizons is necessary for optimization of energy management. Given the increased use of developing technologies and reliance on renewable energy sources, strategic planning, management, and operational decision-making depend on accuracy and reliability of forecasting system. Complex interconnections reside among energy features in modern day power systems. Developing nations such as Bangladesh encounter challenges, including insufficient advanced tools for power planning and policy development. Previous studies have often focused on forecasting a single energy variable, like load or demand, with little attention on multiple energy parameters, and the interrelations among them. This study introduces a Hybrid Neural-SVM Ensemble (HNSE) model to simultaneously forecast day-ahead daily total energy generation, non-renewable energy generation, fuel cost, and evening peak demand of national grid of Bangladesh. Utilizing the Power Grid Company of Bangladesh’s (PGCB) data, HNSE went through processing and hyperparameter optimization. Performance evaluation based on five statistical indices demonstrated the model's predictive capabilities, with a coefficient of determination (R2) of 0.9744, and a mean squared error (MSE) of 0.0291. Additionally, the study utilizes the Kernel-based Changepoint Detection (KernelCPD) algorithm to detect structural shifts in residuals, and two explainable artificial intelligence (XAI) methods, Local Interpretable Model-Agnostic Explanations (LIME) and SHapley Additive exPlanations (SHAP), for feature contribution analysis to provide local and global interpretability. The aim is to offer actionable guidance for policymakers and stakeholders in mitigating current energy crises in Bangladesh through strategic decision-making and support the development of sustainable energy policies in emerging economies.
扫码关注我们
求助内容:
应助结果提醒方式:
