Artificial Intelligence - Exploring its use in grid modernization

By Indu Nambiar, Principal, IT Architecture | 09/30/2024

Editor’s Note: Artificial intelligence is getting a great deal of attention these days in various parts of our economy. Industry leaders and others are working quickly to fully understand the opportunities and challenges. Indu Nambiar, a Principal in Information Technology in the ISO’s Architecture, Integration & Information Security Compliance unit, has been engaged in this topic and offers an overview of the ISO’s current efforts.

The thoughtful adoption of Artificial Intelligence (AI) can drive innovation across the electric energy industry, helping to achieve a variety of state and local energy goals. At the same time, the transformative nature of this technology requires a balanced approach to mitigate the risks associated with such changes. This blog post explores the role of AI in the context of grid modernization, and the potential risks that can accompany its widespread use.

Advances in AI, which involves creating systems or machines that can efficiently perform tasks that typically require human interaction, have the potential to empower electrical utilities and the ISO to build the next-generation smart grid – a grid that is fully integrated, flexible, resilient, interactive, and predictive.

Machine Learning (ML) and Generative AI (Gen AI) are both considered a subset of AI, and it is important to understand the difference between the two because they serve different purposes, have distinct applications, limitations, and implications.

Machine learning is used by systems to autonomously learn from and provide decisions based on large amounts of input data. It is typically used in applications such as image recognition, recommendation systems, and predictive analytics.

Generative artificial intelligence can generate new data or content based on patterns learned from existing data, which includes creating text, images, music, and other media. Technologies like GPT (for text) are examples of generative models.

The ISO is collaborating with multiple vendors to find the optimal and most economical artificial intelligence machine learning solution for many of the system operator’s core market and reliability-related operational processes.

There are opportunities to use off-the-shelf or customized generative artificial intelligence solutions for other corporate business functions, such as human resources, legal affairs customer care, and finance.

Broad set of opportunities to use AI technology for the Grid

Senate Bill 100, signed into law by Gov. Jerry Brown in 2018, established as California policy that renewable energy and zero-carbon resources supply 100 percent of electric retail sales to end-use customers by 2045, and a more modernized grid will be needed to reach that goal.

The ISO already plays a crucial role in facilitating the rapid integration of new carbon-free generation resources like wind and solar. This includes actively balancing supply and demand, integration of grid-connected distributed energy resources, and proactively mitigating the risks to the electrical grid associated with climate change and extreme events.

Modernizing the grid requires optimizing three key areas of grid management: planning, grid operations, and reliability and resilience. AI has the potential to improve and accelerate our work in grid management with a wide spectrum of potential uses.

These new tools not only need to perform the tasks and processes, but they need to offer relevant insights into the decision-making behind their use to sustain trust and reliability. Some general examples of areas in the industry that may potentially benefit from AI are:

Grid Planning
- Analyzing historical data to predict future demand and develop load and variable energy resource profiles for optimal future grid expansion plans.
- Optimizing the placement of new infrastructure.
Grid Operations
- Improving demand and variable supply forecasting, real-time monitoring of grid conditions, and automated responses to dynamic changes of grid conditions.
Reliability and Resilience
- Helping with predictive maintenance of equipment, fault detection, diagnosis, and anomaly detection.
- Assisting in stress testing by simulating extreme conditions, developing adaptive responses, and modeling recovery scenarios to help build a resilient grid that can withstand and recover from disruptions.

Potential risks with large scale data exposure

As we progress towards incorporating AI solutions, we of course take very seriously the obligation of protecting our critical business functions against the many risks associated with the use of this new technology.

Many of those risks are addressed in a report issued earlier this year by US Department of Energy.

Some of the most significant security challenges can be classified as follows:

Attacks against AI tools: This includes data leaks from the machine learning pipeline, altering the data used to train the model (data poisoning), generating adversarial input data to trick the model, etc.
Hostile use of AI: Bad actors could use AI to quickly generate and execute cyber or physical attacks, use a series of seemingly disparate low-risk vulnerabilities that when combined can yield an intrusion. In addition, autonomous control of devices for physical attacks, evasion of cyber detection measures, etc., could be exploited.
AI software supply chain cyber security risk: This can occur through the compromise of both proprietary and open-source software, which AI systems are heavily reliant on, including common tools, libraries, and in some cases, foundation models. As such, cybersecurity and energy system supply chain security best practices are critical to securing the AI software supply chain.

The ISO has implemented several safeguards, as part of compliance with the North American Electric Reliability Corporation (NERC) Critical Infrastructure Protection (CIP) standards, which have enhanced the cyber security posture and protection of the grid. Significant investments have been made in Defense in Depth (DID) cybersecurity strategy that employs multiple layers of security controls to protect systems and data such that if one fails, other layers remain to defend against threats.

Safeguards through policy and regulation

While the ISO is taking strong measures to protect and defend against cyber-attacks, we also recognize the need to consider further research and the development of potential regulations that would facilitate security of AI solutions deployed in energy infrastructure.

Discussions across the energy industry revealed several areas that regulatory and government agencies can explore to help maintain a robust grid that can withstand the threats associated with this evolving technology:

Collaboration with energy sector partners on research to support solutions regarding AI security and grid resilience.
Additional studies and research on AI risk mitigation techniques.
Investing in Privacy-Enhancing Technologies (PET) for AI tools and methods designed to maintain privacy and protect sensitive data, while allowing for data processing and analysis.
Replacing or regulating aging power system equipment which is vulnerable to potential attacks. AI-enabled mass attacks on decentralized vulnerable devices can also affect grid reliability. This is especially concerning as launching such attacks using hostile AI applications is now easier and quicker.
Data from grid operators, materials manufacturers, regulators, and many other entities might be combined using PETs to enable the development and optimization of robust energy systems while protecting private and proprietary information.
Evolving protocols to mitigate the risk of “deep fakes”, which are highly realistic but fake images, videos, or audio recordings generated by a special kind of machine learning called “deep” learning (hence the name). This is especially needed in control rooms and/or dispatch instructions for rural and/or small generation units, smaller water districts, solar farms etc. to protect against this kind of social engineering attacks.
Increasing the use of external penetration testing through hired hackers as an on-going activity to constantly boost the security layers.
Providing education and training programs to ensure that there is a constant supply of technically skilled supply of AI-related cyber security talent.

There is no question that artificial intelligence holds immense promise for advancing grid modernization, particularly in enhanced planning, operations, reliability, and resilience. The ISO is proactively exploring these opportunities and will collaborate closely with our customers, vendors and regulators as AI continues to evolve. However, AI also presents potential risks that must be managed very carefully. This involves implementing robust security measures, ensuring transparent data usage, and maintaining continuous dialogue among stakeholders to address emerging issues.

By thoughtfully navigating these challenges, grid operators can harness AI to create a more efficient, sustainable and resilient energy future for everyone.