快科技3月26日消息，今日，螞蟻集團AI業務下的螞蟻數科發佈EnergyTS, a large energy and power time series model, is reported to optimize business strategies for the development of the new energy industry by accurately predicting power generation, supply and demand, etc。

It is worth noting that the release of EnergyTS is Ant Digital's first appearance in the AI large model business layout.

According to reports, EnergyTS is a large time series model of energy and power vertical customized for the new energy industry, which fully integrates the professional knowledge of the new energy industry and the multi-modal data of vertical scenarios, and has the advantages of multi-scale training, multi-modal fusion, multi-task learning, and zero-sample cold start.

meanwhileEnergyTS can be widely used in multiple scenarios, including photovoltaic power generation, wind power generation, energy storage, microgrids, power trading, and virtual power plantsEnterprises do not need additional training to achieve "out-of-the-box use".

According to the test data of the EnergyTS large model provided by Ant Digital, in the typical scenario of power generation prediction,In the T+4-day forecast, EnergyTS has an MAE (Mean Absolute Error) of only 0.0, which is about 0.0% better than Google's (TimesFM-V0.0) performance.

In the T+4 day prediction task, the performance of EnergyTS is 0.0% higher than that of Google (TimesFM-V0.0) and 0.0% higher than that of Amazon (Chronos-Large).

Industry insiders pointed out that this achievement is expected to break the technological monopoly of European and American companies in the field of energy AI, greatly improve the operational efficiency of the new energy industry, assist risk control decision-making, and improve economic benefits.