Artificial intelligence in grid management A new level of dialogue - AI-supported interaction in grid management
Interacting with critical system data is often complex and requires in-depth specialist knowledge. But what would you say if complex analyses and calculations were as easy as having a conversation? The Model Context Protocol (MCP) promises precisely this paradigm shift.
The MCP connects applications and AI: Thanks to standardized communication, AI language models can be quickly integrated, understand the current system context, and directly support users in data analysis, processes, and informed decisions. It therefore enables intuitive, AI-supported communication with the data in your system.
- It is a standardized protocol for communication between applications and large language models (LLMs).
- It offers simplified use of AI therefore - special interfaces between applications and different LLMs are not necessary.
- Complex queries to the system can be executed using context-sensitive data exchange.
- In conjunction with a locally operated AI chat model that has strong reasoning capabilities, which has been selected by PSI Software SE, analyses can be performed and code can be executed directly. Local operation ensures high performance.
Overcoming Challenges
Chat to applications in a naturally way? Thanks to MCP, it's no longer a problem.
Skills shortage and high staff turnover
For grid operators, the exchange and user-friendly nature and evaluation of network data are becoming increasingly important. With the MCP, PSI Software SE enables AI-supported interaction with the software via natural language, without the need for in-depth prior knowledge or years of experience. This allows users to take advantage of the full range of functions offered by the software, even if they not been working with it for very long. This is pareticularly valuable in periods where there is shortage of skilled labor and high staff turnover, as it significantly reduces training costs and time.
MCP as an interface between AI and applications
The MCP is a standardized protocol for coupling applications with LLMs. Users formulate their queries in natural language, the LLM interprets them and communicates with the respective application via the MCP. The MCP performs several key tasks in this process. It simplifies the use of AI by standardizing communication, eliminating the need for applications to provide special interfaces for different LLMs.
At the same time, the MCP ensures that the system understands the context – the LLM not only receives individual queries, but also, for example, current statuses and ongoing processes, enabling it to respond appropriately. The MCP can be used not only to retrieve information, but also to execute actions within the applications, such as starting calculations or updating data.
MCP offers a high degree of flexibility. An LLM can be extended to support various applications and data sourcing using numerous MCPs.
General Conditions - basis for agentic AI systems
The MCP is more than just a communication standard. It forms the technical basis for so-called agentic AI systems - AI agents that can independently perform complex tasks.
A possible scenario
An AI agent evaluates PSIcontrol reports on unusual load distributions or voltage deviations on an hourly basis and creates a summary based on this information. At the same time, the responsible employees are informed via communication systems so that necessary measures can be initiated immediately – all completely done automatically.
Examples like this demonstrate the potential of this technology and why PSI Software SE is specifically developing this prototype further.
Application example
The MCP has been tested in conjuction with the CIM (Common Information Model) browser, an established tool for visualizing, analyzing, and validating grid data: the "IEEE39bus" grid model data has been imported into the CIM browser via an entry in the chat window. In a second prompt, the number of transformers contained therein is queried. The LLM provides a detailed and correct response to the request. Interactions are possible in English as well as in German.
Local AI chat model with reasoning and highest security requirements
PSI Software SE relies on state-of-the-art technology for the chat model it uses. The model has strong reasoning capabilities, enabling it to analyze complex technical contexts in a structured manner, derive conclusions transparently, and handle multi-level problems. In addition, the LLM supports in-application calculations by triggering them via the MCP, and is capable of analyzing data or performing automated model analyses.
The operation can be carried out entirely locally on an air-gapped GPU (Graphic Processing Unit) server. This physical separation from the network ensures maximum data security and meets the highest security requirements. At the same time, the dedicated GPU infrastructure ensures high performance, enabling even computationally intensive analysis and reasoning processes to be processed efficiently.
Operating the system via an air-gapped GPU server is a costly investment. Alternatively, the MCP can also be operated via a Google server located within the EU; this option is more cost-effective and faster, but as a cloud solution, it offers lower data security, which is an important factor to weigh up.
Your advantages
-
Increased efficiency
The MCP reduces the training time for new employees and ensures excellent performance, allowing even computationally intensive analysis and reasoning processes to be handled efficiently.
-
Excellent interoperability
The standardized protocol enables cross-software evaluation and process automation. The combination with additional AI connections open up further potential.
-
Basis for agentic systems
The MCP provides the founda-tion for AI agents that perform complex tasks autonomously, context-based, and efficiently—the MCP is an investment into the future!
Conclusion: The MCP will significantly boost user friendliness
The MCP makes working with complex software much more efficient. It enables cross-product communication, intelligent data evaluation, a better understanding of complex relationships. In addition, the MCP empowers AI agents to further automate and optimize workflows and business processes.
We are currently still in the early stages and are first testing the use of the MCP in the CIM browser as part of a pilot project. Our research department plans to apply these experiences to other products in the future and evaluate them there. It remains to be seen whether and to what extent the MCP and its potential can be extended to our entire product range in the long term.
