Typesetting and Proofreading | Ke Yangming

On November 15, China's first large petrochemical model was released in Ningbo, Zhejiang. This model was collaboratively developed, validated, and optimized by Supcon, Zhenhai Refining & Chemical, and Ningbo Wanhua, among other petrochemical enterprises.

Ningbo is one of the country's seven major petrochemical industry bases and was successfully selected as a national advanced manufacturing cluster in November 2022, making it the only one in the national petrochemical industry. In 2022, the total industrial output value of Ningbo's petrochemical industry reached RMB 522.1 billion.

Prior to this, on October 14, CNOOC officially released the "Haineng" AI model, and the previous day, on October 15, Guoneng Rixin released the "Kuangming" large model. 2024 is a year for the industrial implementation of large models. Besides education, communications, finance, and healthcare, the energy sector is also gradually embracing large models.

In the nuclear power sector, on March 11, China National Nuclear Corporation's 8th Institute released China's first digital productivity platform in the nuclear field, "Longyin World," along with five general-purpose Nu Copilot™ series digital assistants. A month later, Tsinghua Tongfang Knowledge Network and Huawei jointly released the Huazhi large model 2.0 and a series of solutions, specifically designed for knowledge services and the scientific research industry.

Four months later, the research and development project for the "Knowledge Management Platform (i-Knowledge)" of Qinshan Nuclear Power, a subsidiary of China National Nuclear Power Co., Ltd., passed industry expert evaluation, marking the successful development of China's first AI-based large model knowledge management platform in the nuclear power industry.

In the State Grid sector, on May 9, the first large model product applied in the power industry was launched. This product was developed by the State Grid Information & Telecommunication Corporation's subsidiary, State Grid Xintong Industry Group. Two months later, Xiao Jian, a leading expert at State Grid Hunan Electric Power, released a 1 billion-node distribution network vision large model independently developed by the Electric Power Research Institute.

On October 11, the power safety supervision knowledge-enhanced large model independently developed by Anhui Jiyuan Software Co., Ltd., a subsidiary of State Grid Information & Telecommunication Industry Group Co., Ltd., went online.

In the power generation sector, in January, the Three Gorges Group released the self-developed industry-specific large model, "Dayu Large Model." In April, the China Coal Research Institute released the self-developed Taiyangshi Mining Large Model. In July, State Power Investment Corporation's Digital Intelligence and Technology Company released its self-developed Energy Channel Large Model.

Additionally, in June of this year, Shibeiyun released the CyberwLLM energy large model, and three months later, in September, Lingyang Industry iterated to version 2.0. Meanwhile, Lechuang Energy, which released an energy large model in 2023, also updated its energy storage AI-related solutions this year.

According to statistics from eeo.com, in the first three quarters of 2024, there were over 80 publicly bid energy industry large model projects. Besides general large model enterprises such as Alibaba, Baidu, and Zhipu, energy enterprises and industrial IT service providers were also actively involved in the research and development and investment in industry-specific large models.

Industry insiders state that as the global energy structure transitions towards clean energy, large models are gradually becoming a key driver in the energy sector. Large model technology can utilize deep learning and advanced data analysis methods to achieve high-precision power generation forecasts, grid scheduling optimization, and power market trading support. In the power sector, it not only changes the operation mode of traditional systems but also brings new commercial opportunities for the intelligent upgrading of the energy industry.

That's not all.

On September 26 last year, China Southern Power Grid released the industry's first self-innovated power large model, "Da Watt," marking an exploratory first step in the independent innovation and development of AI in the power industry. Simultaneously, the AI innovation platform for the power industry was launched, opening up and sharing China Southern Power Grid's data, computing power, and model resources with all sectors of society.

Long Yun, General Manager of China Southern Power Grid's Digitalization Department, once introduced that after nearly a year of development, "Da Watt" has incorporated 281 models, covering over ten fields such as grid production and operation, customer service, safety supervision, scheduling and command, and supply chain, with over 100 application scenarios and a cumulative call volume of 2.34 billion times, initially forming a good situation with rich application scenarios and significant application effects.

Energy Vision observed that the impetus behind this wave of AI in the energy industry stems from policy requirements.

Since 2023, the State-owned Assets Supervision and Administration Commission of the State Council (SASAC) has repeatedly made requests for central enterprises to develop AI. In particular, at the special promotion meeting on AI for central enterprises in February 2024, it was proposed that central enterprises should "carry out AI+ special actions." Requirements for computing power and application were also put forward for ten central enterprises, including State Grid, China Telecom, Sinopec, ChemChina, Baowu Steel, and China Merchants.

In July of the same year, the State Council Information Office held a press conference on the theme of "Promoting High-Quality Development," proposing that over the next five years, central enterprises are expected to invest over RMB 3 trillion in large-scale equipment upgrades and retrofits, updating and deploying a batch of advanced equipment with high technology, high efficiency, and high reliability.

Currently, state-owned enterprises are accelerating their efforts to promote the deployment of large models, becoming a pioneering force in driving the implementation of most domestic AI large model projects. According to incomplete statistics, from January to July 2024, the number of large model projects procured by central state-owned enterprises exceeded 950, evenly distributed across multiple directions such as smart computing centers, large model pre-training, agents, and industry applications.

"In the future, we will need more professionals in fields such as data science and AI, especially cross-disciplinary talents who understand both business and AI," Wei Qianhu, Director of the Human Resources Department of Shenzhen Power Supply Bureau, said in an interview with State-owned Assets Report. The evaluation criteria for job competence have changed, and the ability to apply AI needs continuous improvement. Therefore, the company actively recruits and introduces relevant talents on the one hand and coordinates employees to understand and learn AI technology on the other, encouraging everyone to actively adapt.

It is introduced that by utilizing AI to analyze grid operation data, Shenzhen Power Supply Bureau can make more scientific decisions on grid planning and maintenance, effectively manage risks, and significantly improve management efficiency and accuracy.

In fact, state-owned enterprises undertake different research and development tasks based on their own capabilities regarding investments in large models. Therefore, they prioritize scenarios with higher maturity and clearer benefits, currently focusing on applications that enhance employee productivity.

According to the "2024 Tracking Report on the Application of Large Models by State-owned Assets and Central Enterprises" released by the Dune Intelligence think tank, a survey of 36 technical executives from state-owned assets and central enterprises revealed that the preferred application scenarios for large models are knowledge assistants and data analysis, followed by smart customer service and employee office assistants.

Industry insiders state that while energy large models demonstrate huge market potential, their promotion and application also face many challenges.

Firstly, the high costs and computational resources required for model construction and operation require enterprises to carefully weigh the input and output during the implementation process. Secondly, data acquisition and data privacy issues have become important topics in the application of large models. Especially in sensitive scenarios such as power dispatch and market forecasting, ensuring data security and compliance is a key factor that enterprises must consider during implementation.

Looking ahead, large models are not only important tools for realizing energy intelligence but will also play an indispensable role in the future zero-carbon energy system. For example, in the power sector, through precise power generation forecasts and intelligent scheduling, large models help improve the consumption rate of renewable energy and optimize the overall efficiency of the power system.

In the future, with the continuous iteration and advancement of new energy large model technology, emerging industries such as smart grids and virtual power plants will further develop, promoting a more flexible and efficient power system and realizing true source-load interaction and real-time dynamic adjustment. Through the coordinated development of these technologies and industries, large models will play a crucial role in the global transition to clean energy.