石油和天然气运营速度很快。即使是“快速跟踪”的开发也可能需要数月甚至数年的时间。 NobleAI 在 S&P Global 的 CERAWeek 上表示，该公司正在寻求利用其“基于科学的人工智能”来加速行业运营。

NobleAI 能源销售主管 TC Zoboroski 在 3 月 18 日的会议上对观众表示：“从化学品的角度，我们能够让[客户]从材料到配方，加速他们的产品开发。”对我们来说，我们可以将五年的开发周期缩短到 90 天。”

Zoboroski 表示，NobleAI 是一家软件公司，将自己视为“力量倍增器”，可以大幅缩短上市时间。为此，NobleAI 为运营商提供了软件即服务 (SaaS) 模型，该模型针对水库管理、减排和预测性维护等问题进行定制。

“我们采用所有的科学方法、所有不同的物理学，并将其融入到我们的产品中。我们获取所有正在运行的统计数据和经验数据，并在其上分层使用我们的 ML [机器学习] AI 工具。”佐博罗斯基说道。

虽然在石油和天然气行业有很多人工智能应用的例子，但 NobleAI 由于其“基于科学的人工智能”方法，将自己定位为其他技术的改进。

典型的机器学习方法需要大量数据，但如果数据丢失或不完整，则很容易出错，从而导致有偏差的结果。 Zoboroski 表示，对于特定任务来说，大型语言模型通常也过于笼统。

当被问及传统的模拟方法是否可以产生与 NobleAI 相同的结果时，Zoboroski 表示，除了成本高昂之外，模拟速度也很慢，并且依赖于数学求解器和人类。这损害了模拟的一致性，因为“正在运行多个实验，给出多个不同的答案，并且有多个人接触它。”

NobleAI 的技术提供预测和见解，并比传统操作更快地直接生成设计。佐博罗斯基说，从“分子到材料、配方到完整系统”，其平台适用于任何材料。

尽管埃克森美孚是一家规模较小的公司，但它还是采用了 NobleAI 的服务来预测石油产量。 NobleAI 能够结合埃克森美孚提供的多个因素和输入，构建模型并准确预测哪些井会做出响应。埃克森美孚还在后来的应用中使用 NobleAI 为其油井选择最佳支撑剂，从而提高了产量。

“我们从他们这边获取了所有输入信息——多相流、油井干扰、裂缝特性的空间变化——我们推断他们在一个月内只会做少数几个，但做了数百万次推断，” ” 佐博罗斯基说道。 “突然之间，你加快了生产时间，也加快了决策时间。”

埃克森美孚现在使用 NobleAI 来帮助识别表面活性剂的新分子。 Zoboroski 说，自 1993 年以来，尚未发现一种新的表面活性剂分子。

随着世界其他地区采用人工智能，石油和天然气行业也可以利用 NobleAI 的技术产品做到同样的事情。

Oil and gas operations aren’t fast. Even a “fast-tracked” development can take months, if not years. NobleAI is looking to use its “science-based AI” to speed up the industry’s operations, the company said at CERAWeek by S&P Global.

“We enable [clients] from a chemicals perspective all the way through materials and into formulation to accelerate their products,” T.C. Zoboroski, NobleAI’s head of energy sales, told the audience at the conference on March 18. “For us, we can take a five-year development cycle down to 90 days.”

NobleAI is a software company that sees itself as a “force-multiplier” to drastically shrink time-to-market, said Zoboroski. To do this, NobleAI provides operators with a software as a service (SaaS) model, custom built for problems ranging from reservoir management to emissions reductions to predictive maintenance.

“We take all the scientific methodologies, all of the different physics and we put that into our product. We take all the statistical, all the empirical data that is being run and we layer our ML [machine learning] AI tools over it,” Zoboroski said.

While there are plenty examples of AI being used in the oil and gas industry, NobleAI positions itself as an improvement of other technologies because of its “science-based AI” approach.

A typical machine learning approach requires enormous amounts of data, but is error prone if data is missing or incomplete, leading to biased outcomes. Large language models are also oftentimes too general for specific tasks, said Zoboroski.

When asked about whether a traditional simulation approach could yield the same results as NobleAI, Zoboroski said that in addition to being costly, simulations are slow and rely on mathematical solvers and humans. This harms the consistency of simulations as “multiple experiments are being run, giving multiple different answers, with multiple people touching it.”

NobleAI’s technology delivers predictions and insights as well as directly generates designs faster than traditional operations. From “molecules to materials to formulations to complete systems,” its platform works for any material, Zoboroski said.

Despite being a smaller company, supermajor Exxon Mobil employed NobleAI’s services for oil production forecasting. NobleAI was able to combine multiple factors and inputs that Exxon provided, build a model and accurately predict which wells would be responsive. Exxon also used NobleAI in a later application to select the optimal proppant for their wells, which led to an increase in production.

“We took all of the inputs from their side—multi-phase flow, well interference, spatial variation of fracture properties—we took inferences that they would only do a handful of over a month, and did millions of inferences,” Zoboroski said. “Suddenly you're accelerating time to production and accelerating time to decision itself.”

Exxon is now using NobleAI to help identify new molecules for surfactants. A new surfactant molecule hasn’t been identified since 1993, Zoboroski said.

As the rest of the world adopts AI, the oil and gas industry can do the same with NobleAI’s technology offering.