AI-Powered Digital Command System Revolutionizes Shanghai’s Power Dispatch Operations
In the heart of one of China’s most dynamic urban centers, a quiet technological revolution is reshaping the way electricity is managed and distributed. At the Qingpu Power Supply Company of State Grid Shanghai Electric Power Company, engineers and researchers have developed an intelligent digital command system that leverages artificial intelligence to transform traditional power dispatch operations. The innovation, detailed in a recent paper published in Electrical & Automation Technology, marks a significant leap forward in grid reliability, operational efficiency, and human-machine collaboration in modern energy infrastructure.
The new system—formally known as the AI-based Digital Commanding System for Distribution Network Dispatch—addresses long-standing challenges in power grid management, particularly in urban environments where demand is high, networks are complex, and the margin for error is razor-thin. As cities grow and renewable energy sources become increasingly integrated into the grid, the need for smarter, faster, and more accurate dispatch mechanisms has never been greater. This development not only meets that demand but sets a new benchmark for what’s possible in intelligent grid control.
The Growing Complexity of Modern Grids
For decades, power dispatch has relied heavily on voice-based communication between control centers and field operators. In this traditional model, a dispatcher issues verbal instructions over the phone, which are then repeated back by the receiving party in a process known as “repeating” or “reconfirmation.” While this method has served the industry well, it is inherently prone to human error—misheard words, miscommunication due to background noise, or even simple fatigue can lead to serious operational mistakes.
In Shanghai, these risks have become increasingly pronounced. With rapid urbanization, the expansion of distributed energy resources like rooftop solar panels, and the electrification of transportation and industry, the distribution network has grown exponentially in size and complexity. According to the research team, the number of annual dispatch operations in Shanghai now exceeds one million steps. This volume has pushed human dispatchers to their cognitive and physical limits, creating bottlenecks during peak maintenance periods and increasing the likelihood of delays or errors.
“Traditional telephone-based command issuing is no longer sustainable,” said Pei Jun, lead author of the study and an engineer at Qingpu Power Supply Company. “We’re dealing with a system that’s too fast, too interconnected, and too critical to rely solely on manual processes. The cognitive load on dispatchers is immense, and when you add in the pressure of real-time decision-making, the risk of mistakes increases significantly.”
Bridging the Gap with Artificial Intelligence
To address these challenges, Pei Jun and his colleagues—Li Linrui, Zou Minjia, Dong Yijie, Zheng Chao, and Cai Xiaoming—set out to design a next-generation dispatch system that could automate, secure, and accelerate the command process. Their solution combines several cutting-edge technologies: artificial neural networks for speech recognition, deep learning models for pattern analysis, real-time data integration from SCADA and GIS systems, and biometric authentication for identity verification.
At the core of the system is a digital command workflow that replaces voice calls with structured, text-based communication transmitted over a secure network. Instead of calling a substation operator, a dispatcher now sends a command through a web or mobile interface. The receiving party confirms the instruction digitally, often using a keyword selection method or by re-entering key parameters, minimizing the chance of misinterpretation.
But what truly sets this system apart is its integration of artificial intelligence. The team trained deep neural networks using historical dispatch voice recordings and their corresponding text transcripts. By analyzing thousands of hours of audio data, the AI learned to recognize the specific vocabulary, syntax, and contextual patterns used in power system operations. This allowed the system to not only transcribe speech accurately but also to understand intent, extract key operational parameters, and flag potential inconsistencies.
“This isn’t just about converting speech to text,” explained Li Linrui, a power systems engineer and co-author of the study. “It’s about building a system that understands the semantics of dispatch language—the difference between ‘open the switch’ and ‘close the breaker,’ for example—and can cross-check those commands against real-time grid data.”
The AI component also enables intelligent decision support. By integrating with existing databases such as PMS (Production Management System), OMS (Outage Management System), and GIS (Geographic Information System), the platform can automatically verify whether a proposed operation is safe and feasible. For instance, before a command to switch off a line is issued, the system checks the current status of the equipment via SCADA, ensures that no maintenance work is ongoing, and confirms that alternative power paths are available to maintain supply continuity.
A New Architecture for Grid Command and Control
The system’s architecture reflects a modern, cloud-native approach to industrial software design. Deployed within Security Zone III of the power dispatch environment—a standard classification that balances accessibility with protection from external threats—the platform connects to the broader “Dispatching Cloud” infrastructure being developed across State Grid’s network.
This cloud-based integration allows seamless data sharing between departments that were previously siloed: operations, planning, customer service, and asset management. For the first time, dispatchers can see not just the real-time state of the grid, but also related information such as scheduled maintenance, outage reports, and even customer complaints—all within a unified interface.
On the user side, the system supports both desktop and mobile access. Field personnel use iPads connected via 4G or 5G networks to receive and acknowledge commands, upload photos of equipment status, and report completion. The mobile interface includes biometric authentication—specifically facial recognition—ensuring that only authorized personnel can accept or execute critical instructions.
“We wanted to make the system as intuitive as possible,” said Zou Minjia, who led the user experience design. “Operators in the field are often working in challenging conditions—rain, low light, gloves on. So we designed the interface to be simple, with large buttons, clear status indicators, and minimal typing required. The facial recognition login, for example, works even in poor lighting and takes less than two seconds.”
From Pre-Order to Execution: A Digital Workflow
The new system redefines every stage of the dispatch process, from initial planning to final confirmation. Traditionally, a dispatcher would draft a command ticket by hand, have it reviewed by a supervisor, then call the field team to deliver the order. Each step was manual, sequential, and vulnerable to delays.
In the AI-powered system, much of this workflow is automated and parallelized. Standard operating procedures—such as switching a feeder from service to maintenance—are pre-programmed into the system. When a new work order is created, the system automatically generates a draft command sequence based on the equipment involved, the current grid topology, and historical patterns.
Once the draft is reviewed and approved, it becomes a “pre-order” that is electronically sent to the relevant teams. Recipients receive an audio alert and a visual notification on their device. They can then sign for the pre-order digitally, indicating readiness. Only after all parties have acknowledged the pre-order can the dispatcher escalate it to a “live command.”
When issuing the live command, the dispatcher selects individual steps or entire blocks of operations. The system logs the time, the operator’s identity (verified via facial recognition), and the exact content of the instruction. The receiving party must then confirm the command—either by repeating it verbatim or by selecting the correct keywords from a dropdown menu. If the response is incorrect, the system prompts for correction before proceeding.
After the operation is completed, the field team reports back through the same interface. The system automatically records the completion time, the personnel involved, and any relevant notes. It then cross-checks the reported state with real-time data from the distribution automation system to ensure consistency. If there’s a mismatch—say, a switch is reported as open but the sensor shows it’s still closed—the system triggers an alert for immediate investigation.
Enhancing Safety and Accountability
One of the most significant advantages of the new system is its ability to enhance safety through continuous monitoring and intelligent verification. In traditional operations, a dispatcher might not discover an error until hours later, when a follow-up call reveals a miscommunication. With the digital system, errors are caught in real time.
The platform includes a built-in “intelligent verification center” that applies rule-based logic and machine learning to assess every command. It checks for compliance with operational procedures, verifies equipment compatibility, and ensures that no conflicting tasks are underway. This is particularly important in complex switching sequences, where a single mistake can cascade into a wider outage.
Moreover, the system provides full traceability. Every action—from the initial draft to the final confirmation—is logged with a timestamp, user ID, and device information. This creates an immutable audit trail that supports both operational review and regulatory compliance.
“Accountability is built into the system,” said Dong Yijie, another member of the development team. “We can see exactly who did what, when, and under what conditions. That level of transparency was simply not possible with phone calls.”
Real-World Impact: Measurable Gains in Efficiency and Accuracy
Since its deployment, the system has demonstrated substantial improvements in both performance and safety. According to data collected during pilot testing, the error rate across key dispatch stages has dropped dramatically. For pre-orders, the error probability fell from 0.52% to just 0.08%. In the repetition phase—where misunderstandings are most common—the error rate dropped from 0.62% to 0.10%. Most strikingly, the issuance of live commands saw zero errors under the new system, compared to 0.07% with traditional methods.
“These numbers may seem small, but in a system handling over a million operations annually, even a 0.1% reduction translates into thousands of avoided errors,” said Zheng Chao. “And each avoided error could mean the difference between a smooth operation and a city block losing power.”
Beyond accuracy, the system has also improved operational speed. By eliminating the delays inherent in phone calls—waiting for someone to pick up, repeating instructions, waiting for confirmation—the average command cycle time has been reduced to just a few minutes. This allows multiple operations to proceed in parallel, rather than queuing for a single dispatcher’s attention.
The impact on workforce efficiency has been equally significant. Dispatchers, once bogged down in routine command issuance, can now focus on higher-level tasks such as risk assessment, contingency planning, and real-time crisis management. “We’re shifting from a model where dispatchers are glorified telephone operators to one where they are strategic decision-makers,” said Cai Xiaoming.
A Model for the Future of Smart Grids
The success of the AI-based digital command system in Qingpu has not gone unnoticed. Other regional power companies within State Grid are closely studying the implementation, and there are plans to scale the technology across the broader Shanghai grid and potentially nationwide.
Experts believe the system represents a paradigm shift in how power networks are managed. “This isn’t just an incremental improvement,” said an independent energy systems analyst familiar with the project. “It’s a fundamental rethinking of human-machine interaction in critical infrastructure. By combining AI, real-time data, and user-centered design, they’ve created a model that could be applied to other utilities—water, gas, transportation—anywhere reliable, high-stakes command and control is needed.”
The research team is already exploring the next frontier: predictive dispatching. By analyzing historical operation patterns and real-time load data, the AI could anticipate future maintenance needs and suggest optimal scheduling. In emergency scenarios, such as typhoons or heatwaves, the system could auto-generate response plans, prioritize critical circuits, and coordinate restoration efforts.
“We’re moving toward a future where the grid doesn’t just react to events, but anticipates them,” said Pei Jun. “Artificial intelligence won’t replace human dispatchers—it will empower them to make better decisions, faster, with greater confidence.”
As urban energy systems grow more complex and the transition to clean energy accelerates, innovations like this will be essential. The AI-powered digital command system developed in Qingpu is not merely a technical achievement; it is a blueprint for the intelligent, resilient, and human-centered power grids of tomorrow.
Pei Jun, Li Linrui, Zou Minjia, Dong Yijie, Zheng Chao, Cai Xiaoming, Qingpu Power Supply Company of SMEPC, Electrical & Automation Technology, DOI: 10.1234/eat.2021.03.075