The duo of AI and energy

There is a significant need for power quality meters to measure power quality, especially at data centres, due to potential load spikes.

The relationship between AI and energy is complex, but there is a chance that energy tech companies can support AI development by using AI. In India, there is a significant need for power quality meters to measure power quality, especially at data centres, due to potential load spikes. Data usage is not as linear as a house, and AI models can cause data usage to be unpredictable. To prevent these spikes, power quality meters and other devices, such as STATCOMs and capacitor banks, are necessary. AI can predict demand spikes by measuring data at high frequencies, enabling early protection of devices to avoid such spikes. This creates a virtuous cycle between AI and energy companies, ensuring the quality of power and lower costs in data centres.

Kimble’s journey

We have completed around two crores of the 25 crores planned smart meters in India, with over 50 million installations to date, and currently hold a 25 percent market share in RDSS installations. We have a comprehensive data acquisition, decision-making, and communication stack for smart meters. It includes everything from data generation to decision-making, and is currently developing new products for the AI-enabled grid.

We are developing a scalable software platform that utilises AI/ML in specific areas and employs appropriate software stacks, including data pipelines and databases, to ensure optimal performance and efficiency.

We are developing a high-speed, customisable FRT (Fault Ride-Through) and addressing the issue of substation upgrades by creating a modular FRTO (Fast Reclosing Transient Overvoltage). This architecture separates different input/output modules into separate cards, allowing for easy hot swapping, making substation upgrades a breeze. Writing modular firmware can help FRTO function as both a load controller and a battery controller, enabling it to perform multiple applications. We also have non-intrusive load monitoring products. We use ML to detect individual devices from the main smart meter of a home to understand the disaggregation of loads and electrical faults.

Advertising

AI is essential to accelerate the energy transition. AI is not everything; it is a combination of mathematics and a black box, and it is solved when the human brain cannot comprehend the complete solution. Misusing AI can lead to incorrect solutions and problems, as it is not a classic solution to any problem. For example, in a load forecasting case, an AI model achieved a 90 per centper cent accuracy rate using a five-day moving average, but improved to over 95 per centper cent accuracy due to constant load. However, AI usage can be problematic and needs to be guarded against.

AI modelling should be done at scale, transparently, and differentiated between data, device, communication, computation, and visualisation layers. Interpretable algorithms should be used, and open-source models should be employed, avoiding reliance on black boxes. Proper data validation and provenance are crucial for scalability and versatility in various scenarios. It is essential to understand the inputs and outputs, and to utilise open-source models without relying on black boxes. This ensures the model is scalable and can be used in various scenarios.

AI is not real intelligence; we must understand this and utilise AI effectively to enhance our skills, rather than competing with or overusing it, and not try to compete or avoid its use.

To conclude, AI and energy must evolve together to foster reliability, efficiency, and innovation. Though AI can enhance forecasting, monitoring, and optimisation, it must be applied transparently, with human oversight and modular solutions.

***************************