清鹏智能创始人李中阳对《财经》说，前两年公司潜心做算法构架，打下技术基础。2024年开始将产品 实际运用到电力市场。今年以来，清鹏智能开始商业化运营。 除了清鹏智能这样的科技公司，还有诸多能源企业，包括发电商、售电商、用电企业等都在尝试自研电 力交易AI产品。这些公司在电力市场盈亏的关键都是预测新能源的发电量和用电量情况，有的公司， 一天的盈亏波动可能就超千万元。电力交易员需综合分析多维度的海量数据以做出交易决策，显然， AI处理数据的能力远高于人脑。 电力市场的交易决策依赖AI，既可能放大盈利也可能放大亏损。电力交易相关AI模型尚未成熟，企业 使用时必须具备自主操控能力 文｜《财经》记者 徐沛宇 编辑 | 马克 如同DeepSeek创始人梁文峰将AI（人工智能）引入金融市场做量化交易，一些创新者正在将AI运用于 电力市场交易。 成立于2021年的北京清鹏智能科技有限公司（下称"清鹏智能"），孵化自清华大学电子系的一个AI实验 室。2022年，清鹏智能聚焦到能源赛道，专注做电力交易AI智能体（Agent）。 国家电网公司和南方电网公司近期都举行了AI预测用电量、AI做电力交易等相关比赛；数百家发电 商、售电商 ...

Group 1 - The core viewpoint of the articles highlights the rapid development and application of large models in the energy sector, transitioning from general to specialized fields [1] - Several major energy companies, including China National Petroleum Corporation and State Power Investment Corporation, have launched large models aimed at enhancing efficiency in energy production and management [1][2] - The energy industry has begun to adopt large models for various applications, including grid scheduling, coal and nuclear power production, and renewable energy management [1][2] Group 2 - In the renewable energy sector, power forecasting using large models has become a critical application, addressing the challenges posed by the increasing share of renewable energy in the grid [2] - Traditional forecasting methods are becoming inadequate due to the complexity of weather conditions and the growing scale of renewable energy installations, necessitating the use of advanced large models [2][3] - Companies like Google DeepMind and Huawei are developing sophisticated weather prediction models that enhance the accuracy of renewable energy power forecasting [2] Group 3 - Large models can optimize the allocation of renewable energy in real-time, significantly reducing the waste of wind and solar power [3] - The integration of large models in equipment maintenance can improve operational efficiency by analyzing vast amounts of energy data and enabling predictive maintenance [3] - Collaboration with advanced technologies such as drones and robots can further enhance the application of large models in energy equipment inspection [3] Group 4 - Prior to the emergence of large models, the energy sector primarily utilized specialized small models for specific tasks, which had limited data requirements [4] - The introduction of large models has expanded the scope of applications in the energy sector, addressing more complex challenges such as grid stability and renewable energy integration [5] Group 5 - Various technical routes for large models exist, with time-series models showing significant potential in renewable energy power forecasting [6] - The integration of more meteorological data into time-series models can enhance predictive accuracy and improve energy dispatching [6] Group 6 - The maturity of language models in the energy sector is currently low due to the lack of available data compared to general language models [7] - The fragmentation of IT and OT systems in the energy industry complicates the effective integration of heterogeneous data, which is essential for AI applications [7] - Developing reliable and interpretable industrial AI models that combine expert knowledge with AI algorithms remains a challenge in the energy sector [7]

Core Insights - The 18th International Solar Photovoltaic and Smart Energy Conference (SNEC PV+) was held in Shanghai from June 10-12, where Ant Group's Ant Financial Technology launched the "Energy Service Intelligent Agent" based on the EnergyTS time-series model, which significantly enhances investment decision-making efficiency by over 60 times compared to manual processes [1][2] - The recently implemented "Document No. 136" mandates that starting June 1, 2025, new renewable energy projects will participate in market trading for electricity pricing, which will be determined by market supply and demand rather than fixed benchmarks, raising the bar for renewable energy companies in terms of forecasting and decision-making accuracy [1] - The Energy Service Intelligent Agent covers three main scenarios: pre-investment decision-making, smart operations, and asset finance, enabling automated task planning and multi-agent collaboration through simple text commands [1][2] Pre-Investment Decision-Making - The intelligent agent can automatically generate project proposals, conduct economic assessments, perform sensitivity analyses, and optimize plans, reducing the investment calculation cycle from 2-3 days to just over ten minutes [2] - It produces detailed investment decision reports, enhancing the efficiency of the pre-investment phase [2] Smart Operations - In the smart operations scenario, the agent can automatically create comprehensive operational strategies, risk management strategies, and trading strategies, improving management efficiency and mitigating electricity price volatility risks [2] Asset Finance - The intelligent agent provides asset profiling analysis and evaluations for investors, offering asset yield enhancement measures and financing suggestions for energy companies [2] - It is set to be applied across various fields, including commercial photovoltaic, residential photovoltaic, energy storage, and integrated energy, with a recent partnership established with JA Solar for pre-investment decision-making applications [2] Technology and Model Performance - In March, Ant Group released the EnergyTS time-series model, which integrates industry-specific knowledge and multimodal data, enhancing the accuracy and effectiveness of task planning and tool invocation [2] - The model has demonstrated superior electricity generation forecasting accuracy compared to mainstream general time-series models from Google and Amazon in industry evaluations [2]

Core Viewpoint - GCL-Poly Energy (协鑫能科) reported a significant increase in net profit for 2024, driven by asset optimization and favorable national "dual carbon" policies, with a focus on energy services for future growth [1][2][6] Financial Performance - In 2024, GCL-Poly achieved operating revenue of 9.796 billion yuan, a slight decrease of 5.42% year-on-year, but net profit attributable to shareholders reached 489 million yuan, with a non-recurring net profit of 294 million yuan, marking a substantial increase of 190.83% [1][2] - For Q1 2025, the company reported a net profit of 254 million yuan, a year-on-year growth of 35.15%, with a non-recurring net profit growth of 176.61% [1] Growth Drivers - The company's performance improvement is attributed to the elimination of inefficient coal-fired units and increased investment in renewable energy assets, raising the share of renewable energy installations from 24.94% in 2022 to 57.38% in 2024 [2] - GCL-Poly completed green electricity transactions of 4.42 billion kWh in 2024, with a corresponding green certificate volume of 1.224 billion kWh [2] Energy Storage and Services - The company is expanding its new energy storage business, achieving a grid-side storage capacity of 650 MW/1300 MWh and a user-side storage capacity of 11.75 GW/31.96 MWh [3] - GCL-Poly's dual-driven strategy of "energy assets + energy services" has created an ecological closed loop, participating in market transactions of 27.057 billion kWh [4] Technological Innovation - The company has partnered with Ant Group to complete the largest domestic RWA (Real World Asset tokenization) project, exceeding 200 million yuan, enhancing digital technology in the green industry [4][5] - GCL-Poly has developed a time-series model for photovoltaic scenarios, reducing curtailment rates by 18%, breaking the technological monopoly of Western companies in the energy AI sector [5] Strategic Goals - GCL-Poly aims for energy service revenue to exceed 50% within five years, with a current share of 12.18% and a gross margin of 59.03%, significantly higher than traditional energy sales [6] - The company is positioned to benefit from the green electricity and energy storage markets, with projections indicating a potential net profit exceeding 1.1 billion yuan by 2025 [6][7] Shareholder Returns - GCL-Poly plans to distribute a cash dividend of 1.00 yuan per share, totaling 158 million yuan, which represents 32.34% of the 2024 net profit [7]