EVN SPC breaks through digital transformation with high-precision electricity load forecasting model

The application of the Rolling SARIMAX model helps EVN SPC improve the accuracy of short-term electricity load forecasting, thereby increasing dispatch efficiency, ensuring safe and stable electricity supply for the Southern region.

EVN SPC staff inspect 110kV power grid project in the Southern region

In the context of the constantly increasing demand for electricity in the Southern region, increasingly affected by weather changes, socio -economic development and unusual factors, the requirement to improve the accuracy of electricity load forecasting is becoming more urgent than ever.

According to Deputy General Director of Southern Power Corporation (EVN SPC) Bui Quoc Hoan, electricity load forecasting plays a key role in planning, operating and ensuring the balance of supply and demand of the national power system. In particular, the Southern region is a dynamic economic region, with many industrial parks and export processing zones, so forecasting requirements must not only be accurate but also have the ability to adapt quickly to actual fluctuations. Meanwhile, electricity load data is a complex time series, simultaneously affected by long-term trends, seasonal rules and exogenous factors such as weather, holidays, epidemics, and socio-economic fluctuations.

Based on that practical requirement, EVN SPC has researched and proposed to deploy the Rolling SARIMAX model - a form of rolling forecasting, allowing the model to be continuously updated every day instead of static training like the traditional method. Exogenous variables such as temperature, calendar factors, and social mobility index are integrated to fully reflect the impact from the environment on electricity load. The research data includes daily electricity load of 21 southern provinces (excluding Ho Chi Minh City) in the period from 2021 to September 2024.

Conductor Phase Project of 110kV Vung Liem - Tra Vinh 2 Line

Experimental results show that the Rolling SARIMAX model achieves a mean absolute percentage error (MAPE) of only 1.79% — the lowest among all the models compared. This figure outperforms deep learning models such as LSTM, GRU, attLSTM with MAPE ranging from 4.3% to 7.36%; traditional statistical models SARIMA, Holt-Winters with MAPE ranging from 3.62% to 5.36%; as well as machine learning models such as Random Forest, XGBoost with MAPE around 2.8%.

This fact shows that a dynamically updated linear model can still outperform complex deep learning models, while having low operating costs and being easily integrated into the existing dispatching system. According to Deputy General Director Bui Quoc Hoan, this is a huge advantage that helps EVN SPC deploy the application into actual operation in a short time.

In the coming time, EVN SPC plans to continue researching advanced signal decomposition methods such as EEMD, VMD to optimize input data, and develop an automated forecasting system, allowing real-time model updates and direct connection to the power dispatch system. The application of the Rolling SARIMAX model is expected to contribute to improving the reliability of power supply, optimizing operating costs and minimizing risks in the dispatching of the power system in the Southern region.

EVN General Director Nguyen Anh Tuan visited EVN SPC's booth at Techshow 2025

Not only impressing through load forecasting research, at the Electricity Science and Technology Exhibition - Techshow 2025, EVN SPC also made a strong impression with a series of products and technological solutions serving digital transformation and modernization of the electricity industry. EVN SPC's booth displayed 10 typical products, from the model of the 110 kV Ha Tien - Phu Quoc underground cable line and the 220 kV Kien Binh - Phu Quoc line - the longest sea-crossing line in Southeast Asia, to BIM and Scan to BIM applications in grid construction investment.

In the technical field, the hotline remote control insulation cleaning device allows cleaning insulation with high pressure water without cutting off the power, doubling labor productivity and eliminating the risk of having to climb poles to do manual operations. Along with that is a series of artificial intelligence application solutions such as AI analyzing partial discharge signals, AI diagnosing transformer failures by analyzing dissolved gas, and AI monitoring labor safety.

EVN SPC also introduced management and customer service platforms in the digital age such as the E-Tivi internal communication application, the customer branch program displaying the current status of the power grid on a digital map, thereby improving management efficiency and service quality.

The research results and technological products at Techshow 2025 not only affirm EVN SPC's efforts to master technology but also demonstrate its determination to comprehensively transform digitally to meet increasingly high demands in the context of rapidly increasing load, strong energy transition and increasingly stringent requirements on energy security. Thereby, EVN SPC continues to play a core role in ensuring safe and stable power supply for socio-economic development in the Southern region and the whole country.

Source: https://daibieunhandan.vn/evn-spc-but-pha-chuyen-doi-so-voi-mo-hinh-du-bao-phu-tai-dien-do-chinh-ac-cao-10399479.html