Skip to main content
Article

Explainable hybrid forecasting model for a 4-node smart grid stability

Taher M. GhazalDepartment of Networks and Cybersecurity, Hourani Center for Applied Scientific Research, Al-Ahliyya Amman University, Amman, JordanMohammad Kamrul HasanFaculty of Information Science and Technology, Universiti Kebangsaan Malaysia, 43600 Bangi, Selangor, MalaysiaRosilah HassanFaculty of Information Science and Technology, Universiti Kebangsaan Malaysia, 43600 Bangi, Selangor, MalaysiaMustafa AbdullahElectric vehicles engineering department, Hourani Center for Applied Scientific Research, Al-Ahliyyaesearch, Al-Ahliyya Amman University, Amman, JordanShayla IslamInstitute of Computer Science and Digital Innovation, UCSI University, Kuala Lumpur 56000, MalaysiaMunir AhmadUniversity College, Korea University, Seoul, 02841, Republic of Korea
2025en
ABI

Abstract

The 4-Node architecture for smart grids (SGs) distinguishes the operational features to enhance its stability and interruption less power distribution. This architecture contains 1 generator node and 3 consumption nodes to retain the stability of SGs. However, the chances of grid component failure, deficient power distribution, or peak demands result in the 4-node architecture imbalance. To forecast such imbalance towards SG stability, this paper exploits the explainable artificial intelligence (AI) and deep learning models to verify the existence of bias in the 4-node architecture due to the above flaws. For this purpose, in this paper, the component functioning rate and demand-to-distribution ratio are computed and verified to compute the bias. The deep learning model trains the explainable AI model for the potential flaws from the distribution intervals. Using the training using the deep learning model, the 4-node architecture’s performance is evaluated to forecast the grid stability for future distributions. The 4-node architecture stability is verified under different loads, peak demands, and device performances using the integrated hybrid model with EAI. The proposed model improved the forecast accuracy by 14.53 % and reduced the loss in flaw detection by 13.41 % for the varying dissemination hours.

Identifiers

Citations and references

Cited by 30 references