A novel hybrid photovoltaic current prediction model utilizing singular spectrum analysis, adaptive beluga whale optimization, and improved extreme learning machine

The precise and accurate prediction of the photovoltaic (PV) system can ensure the grid-connected system's constant operation and defect-free performance. Nonetheless, the PV current system's chaotic output is compounded by unpredictable weather patterns, which lead to poor predict stability, accuracy, and increasing costs. This paper introduces a novel prediction hybrid model based on singular spectrum analysis (SSA), adaptive beluga whale optimization (ABWO), and an improved extreme learning machine (IELM). The SSA is first employed for pre-processing long-term data input and output-time series. Then, the ABWO algorithm is enhanced to increase the exploitation tendency, simplify whale fall mechanism, and improve exploration/exploitation balace. The ABWO is validated by comparing it to well-published methods using a range of benchmark functions and then implemented to address real-world engineering challenges. Additionally, the inverse matrix of the output weights is adjusted after the training phase to optimize the ELM model. Ultimately, the hyperparameters of the IELM model are optimized utilizing ABWO algorithm. The numerical experimental results demonstrated that the ABWO algorithm beat all other optimization methods and could solve the majority of the benchmark functions. In contrast, the proposed SSA-ABWO-IELM model showed exceptional performance when compared to other hybrid deep machine learning models, as evidenced by numerous statistical assessments.

Text 122531.pdf - Published Version Restricted to Repository staff only Download (39MB) Official URL or Download Paper: https://linkinghub.elsevier.com/retrieve/pii/S0960...

Actions (login required)

View Item