Intelligent operation and management in the Dadu River Basin

The top priority of a hydropower enterprise is to secure the safe, scientific, and economic operation of its power stations. The Dadu River Basin, the fifth largest hydropower base in China, is definitely vital to the grid security and stability in southwest China's Sichuan Province. Given the challenges faced by production O&M (operation and management), efforts were made to put forth solutions featuring multielement business quantification, multisource data integration, and multitarget intelligent collaboration, according to which, through the development and application of a number of key technologies for intelligent operation, hard work was done to promote management optimization simultaneously and realize the smart and autonomous operation of the 10‐million‐megawatt power station group in the basin.


| INTRODUCTION
The Dadu River Basin ranks 5th among the 13 major hydropower bases in China, with a hydro energy resource reserve of 34.59 million kW. It plays an important role in the flood control system of the Yangtze River Basin and constitutes the main power source of the power grid in Sichuan. By the end of 2020, the installed capacity of the basin reached 21.86 million kW, accounting for 36% of the total installed capacity of the Sichuan power grid.
The top priority of a hydropower enterprise is to ensure the safe, scientific, and economic operation of its power stations. Due to the special natural, geographical, and cultural environment, especially the extremely complex hydrometeorology, frequent earthquakes, fragile ecological environment, and difficult production conditions in the basin, Dadu River Company faces many challenges in operating and managing its power stations.
In 2008, IBM put forward the idea of building a "smart earth" by means of a new generation of information technology (Tao & Tang, 2022). Tu Yangju proposed an intelligent new operation and management (O&M) mode characterized by "data-driven management," and analyzed the operating mechanism of intelligent enterprise (Tu, 2018(Tu, , 2019(Tu, , 2021. Inspired by the concept of intelligent enterprise, various industries began to explore the development of intelligent enterprises. Hydropower enterprises mainly focused on intelligent power plant and intelligent engineering (Duan, 2021;Geng, 2018;Tang et al., 2019), and promulgated the "Guide for Smart Hydropower Plant" and "Guide of Intelligent Dispatch for Cascaded Hydropower Plants," which pointed out the direction for intelligent water conservancy projects.
By putting forward the concept of intelligent enterprise, Dadu River Company has tried to adopt advanced intelligent technology and carried out intelligent O&M in its power stations in the basin in line with the vision and ideas of intelligent enterprise development (Tu, 2018(Tu, , 2017Tu & Zheng, 2016). Significant accomplishments have been achieved.

| General idea
Dadu River Company deeply integrated advanced information technology such as cloud computing, big data, Internet of Things, mobile Internet, and artificial intelligence with traditional hydropower O&M, in a bid to build an intelligent new O&M mode characterized by "data-driven management and man-computer collaboration." First, by virtue of modern sensing technology, the Company established a quantitative sensing system, realizing automatic sensing of meteorological and water conditions, geological slopes, equipment status, dam status, ecological environment, and so forth, and completing the quantification of multifactor operations based on the core area of the basin. Second, it realized the efficient transmission of data by constructing 4G/5G, WiFi and other terminal transmission networks, and the backbone fiber optic network of "power fibre + special fibre optic cable." Third, a basin-level data center was constructed to integrate multisource data acquired from cascade power stations.
Finally, it gave full play to staff's creativity, intelligently, and efficiently explored data resources built a library of models and algorithms.
Focusing on multiple-purpose collaboration, an intelligent O&M platform that integrates optimal dispatch, power station operation, equipment maintenance and overhaul, safety control of dam and reservoir bank, and ecological environment protection, and so on, was framed.
Meanwhile, by optimizing the management mode of traditional hydropower enterprise, the Company achieved the automation of risk identification, intelligent decision-making management, and autonomous correction and upgrading in the O&M of power plants, which is shown in Figure 1.

| Technical path
Digital technology is an important tool to promote intelligent O&M. Dadu River Company planned "five big" construction strategies for intelligent transformation.
(1) Build a "big sensing" system. The Company built 110 hydrological telemetry sites, 25,000 risk monitoring points for dams and reservoirs, more than 260,000 equipment codes, and over 7000 monitoring indicators for equipment. It independently developed intelligent sensing equipment such as smart safety helmets and smart inspection robots achieving a comprehensive sense of the environment, business, and management. (2) Build a "big transmission" network. A big transmission network has been set up, which is based on the Internet and supplemented by the Industrial Internet of Things and mobile Internet to achieve real time and massive data transmission. Meanwhile, a mobile network has also been built with full coverage of fiber optics, 5G and high-speed Wi-Fi in the basin production area, and an emergency communication network built with GPS, Beidou, and KA-band satellite transmission links to respond to emergencies. (3) Build a "big storage" platform. The establishment of a big data center and a unified data storage platform containing 10 domains has been completed. With a "centralized & distributed" hybrid storage mode, tiered storage of data has been actualized. More than 300 T data from the basin has been integrated by far. (4) Enhance the level of "big computing." Dadu River Company has established a unified platform to realize the cloud deployment and unified management of all servers, offering a computing resource pool with more than 2300 core CPUs and 40 T of RAM. Through desktop virtualization technology, office computers of the Company's headquarters and subsidiaries have been replaced with cloud desktops, realizing unified construction, unified management, and unified distribution of cloud computing resource pools and cloud desktop resource pools. An IT supervision platform has been simultaneously built F I G U R E 1 General idea of intelligent operation and management. TU ET AL.
| 31 for unified monitoring and management of data and network to ensure the security of both data and network. (5) Cultivate the "big analysis" capability. Dadu River Company has made efforts to pave the way for employees to carry out collaborative innovation revolved around obstacles and difficulties in their work. Through various innovation activities such as Big Data Modeling Competition, Intelligent Enterprise Salon, and Youth Innovation Studio, a series of models tightly bonding power production process, such as smart prediction, early warning and decisionmaking, about 120 algorithmic models of various types from safety production to comprehensive management have been created to provide support for the O&M of the basin's cluster of terraced power stations by automating risk identification, intelligent decisionmaking and management, and autonomous correction and upgrading.

| KEY TECHNOLOGIES FOR INTELLIGENT OPERATION
Dadu River Company has developed a series of key technologies to execute intelligent operation. The following are several representative technologies with respect to power dispatching, equipment operation, equipment maintenance, and geological disaster early warning.

| Real-time load intelligent one-key adjustment technology for cascade hydropower stations
Dadu River Company has actualized a real-time and smart "one-key load adjustment" system among cascading stations, which can meet high-intensity peak-load and frequency regulation demand required by local grid system. The general idea of the system is shown in Figure 2. Based on load demand or frequency deviation of power system, power grid dispatch center issued real-time generation total-load command of the cascade hydropower stations to the "One-key Load Dispatch" system. Then, taking the water supply and power demand of cascade stations into consideration, the sum load is economically and efficiently distributed to the AGC of each power station, which further distributes the load among the units in the station and also returns the execution results.
Based on the safety and economic requirements, the real-time load dispatch situations are divided into two categories: command mode and noncommand mode. In command mode, when grid AGC system issues a combined load command of the three plants, the Dadu River load F I G U R E 2 Relationship between load allocation and scheduling instructions for the basin's stepped central control center. distribution system will automatically select a combination of plants power setpoints according to their ramping rates to ensure the shortest time required for the load adjustments. In noncommand mode, based on deviations between model calculation result and actual dispatching process, operational water level zone is divided into high zone, acceptable zone, and dead zone. When water level enters the high zone or dead zone and shows no tendency of returning to acceptable zone, a model called abnormal water level model will be automatically matched and initiate load redistribution to expedite returning of the anomalous water level to its acceptable zone as soon as possible. Water level zoning control principle is shown in Figure 3.

| Intelligent power station inspection technology based on computer vision and human-machine collaboration
Through integration of algorithms and inspection business, Dadu River Company created a set of intelligent inspection technology based on the management and inspection of station equipment operation. By adopting the "mobile sensing & fixed monitoring" approach the multisource sensing method of equipment status and environmental information was built as a result. Via this approach the full-dimensional inspection data such as video, audio, infrared, and sensor of hydropower station could be gathered.
Using computer vision and self-learning technology of machines, an AI scheduling engine for large volume of data and tasks has been developed to form a mechanism model for smart identification of audio and video and further intelligent allocation of inspection tasks in complex environments of hydropower plants.
The intelligent inspection system is a comprehensive control system that completes unmanned inspection and automatic warning in power stations. The intelligent inspection robot (Figure 4) is the main equipment for implementing safety supervision of equipment and operation environment. The robot's front-end acquisition instrument has bionic perception function, which enables the percepting of video, audio, environment, and other elements of the equipment and the identification of abnormalities through artificial intelligence technology. For the special restricted spaces in hydropower stations where robots cannot reach, the integrated application of sensing hardware, such as power station industrial TV systems, sound sensors, and infrared thermal imaging equipment, enables intelligent sensing of temperature, foreign sound, water accumulation, oil leakage and vibration in restricted spaces such as hydro generator wind hoods and wind tunnels.

| Equipment condition analysis technology
Based on the data sources in the maintenance data center, the Dadu River Company used a combination of top-down and bottom-up approaches to build a knowledge map for the maintenance of equipment. That is, the data patterns are first defined manually, and then information is extracted for the specific relationships of the F I G U R E 3 Water level zoning control principle. TU ET AL. data patterns. Priority is given to data sources from the maintenance data center, which is then supplemented by other common data sources. The knowledge map of equipment maintenance provides an effective way to utilize and manage a large number of heterogeneous data sources. After establishing a feature library of over 7000 types of hydropower unit faults for autonomous learning, a realtime assessment model of equipment health and an early warning method for health trends has been developed.
In addition, the Dadu River Company has built a smart maintenance system based on the self-developed "hypersphere model," which mainly contains functions such as health evaluation and big data trend warning. The algorithm of the "Hypersphere Model" uses all relevant measurement points of the main equipment of the hydropower station combined with different historical operating conditions data to establish different models, and automatically conduct real-time evaluation of the equipment. The algorithm synthesizes the real-time status of the hydrogenerator unit into a 0%-100% evaluation value, called "Health Preservation Index (HPI)" (Figure 5). The smart maintenance system analyzes the historical data and builds a hypersphere model to obtain a benchmark value of health, "Hth," which is used as the standard for the healthiness of the hydro-generator set. During operation, when the HPI of the equipment deviates from the historical-proved safe operating conditions, the system issues early warnings of potential faults before the alarm of equipment happens. At the same time, the rating of the associated measurement points that cause changes in the status of the unit is automatically given and automatic correlation analysis of early potential faults is carried out.

| Smart early warning technology for geological disasters on reservoir banks and slopes
The Dadu River Company has established a professional monitoring database for geological hazards in the basin and integrated professional monitoring data to provide data support for early warning analysis. First, it has developed a three-dimensional (3D) data analysis technology based on F I G U R E 4 Example of the structure of intelligent inspection robot for a hydropower plant (wheeled).

F I G U R E 5 Health Preservation Index evaluation and trend warning.
"space-sky-ground" multisource data (see Figure 6 for details), and the "triple-check" early identification system of general investigation. Detailed check and verification was constructed, namely: a large-scale geological hazard general investigation based on satellite InSAR technology was carried out; through disaster interpretation of airborne LiDAR images and UAV real-world 3D model, the geological disaster potential in the critical area was investigated in detail, following the suspected potential verified through the manual collaborative approach. Second, it has built a rapid inversion technology of physical and mechanical calculation of very large landslides based on machine learning, which can quickly call similar disaster case parameters in the cloud by analogy and accomplish the whole process of "deformation → progressive damage → disaster rupture → high-speed movement → accumulation" of very large landslides based on solid-liquid coupling, also this technology can efficiently and quickly quantitative risk assessment. At the same time, the physical model of geological disaster chain dynamics evolution under a real spatial scale was established, and the numerical simulation platform of disaster chain evolution was developed to achieve a quantitative assessment of risk.
It has established a big data analysis platform based on multisource geological disaster information, realizing information, smart and standardized management of the whole process from data aggregation, data management, risk evaluation, monitoring and early warning, command and dispatch, and comprehensive prevention and control.

| INTELLIGENT MANAGEMENT MODEL AND ITS FOUR-LAYER LOGIC
The Dadu River Company has built a management model featuring "human-oriented, human-machine collaboration" from the perspective of the whole enterprise.

| A human-oriented digital twin management model
Different from the digital twin of Germany's Industry 4.0 for production lines and equipment, what Dadu River Company proposed is the digital twin of the entire enterprise management form, that is, based on retaining the physical hierarchical organizational structure, digital transformation of core business, and professional integration of functional departments are the focus, mainly, gradually adding elements of the change management system, building a "dual-track system" of hierarchical management add data-driven, which gradually increases the reliance of the original management system on the data-driven management model. The operating mechanism of its intelligent O&M is shown in Figure 7.
We can see that "Human" are always at the core of the intelligent O&M of the enterprise, and a three-order circulation system characterized by human-machine F I G U R E 6 A three-dimensional observation system of "air-sky-ground" multisource data. TU ET AL.
| 35 collaboration has been constructed between physical enterprises and digital enterprises. As a result, the Dadu River Basin as a whole presents the integration of hierarchical management of physical enterprises and data-driven nonhierarchical management.

| The four-layer logic for the operation of human-machine synergy
Human-machine synergy refers to the synergy of people with smart devices and intelligent systems in an enterprise, which unfolds according to the following four-layer logic.
The first is the collaboration of the production line, which realizes the interactive collaboration between people and smart devices during the operation of the production line. Smart terminals are an effective means to achieve synergy. The application of smart inspection robots in the Dadu River Basin has actualized the complementarity and coordination of machine inspections and manual inspections, which not only improves the efficiency of inspections but also greatly enhances the accuracy, effectiveness, and flexibility of inspection work.
The second is the synergy of the production chain, especially the optimization of synergy in process, division of labor and organization. After achieving synergy in production lines, the Dadu River Company has further adjusted the division of labor in the production chain. For example, through the construction of the Basin Command Centre and the abolition of the station central control room, about 120 employees were reduced in the production process, a move which helps to optimize the production chain.
Third, the synergy of management system. The intelligent enterprise management experiences continuous change and upgrade. Dadu River Company has merged the departments and transformed the functions of the agencies, greatly liberating employees from operating positions with harsh conditions. At present, 12% of its employees have been transferred to digital-related businesses, giving a new look to enterprise management.
The fourth is the synergy in the ecological chain. Intelligent enterprise has moved from closed enterprise boundary to open industrial ecology. Dadu River Company has formed a shared service with industry peers by building a data service platform for the industry. An ecology of sharing, integration, and synergy with society in geological disaster warning, inflow and weather forecasting, emergency rescue, and so forth, has taken shape in the main.

| FRUITS ACHIEVED
After several years of exploration and practice, the Dadu River Company has seen significant changes in its O&M, The size of the installed capacity has doubled. The total number of employees has been reduced by 14%. The productivity of the total workforce has increased by 75%.

| Transformation of the management model from hierarchical to centralized
The company has built a digital enterprise corresponding to the physical enterprise. Focusing on the top-level design structure of "one main center, multiple subcenters, and four units" for an intelligent enterprise, the company has built 22 professional data centers in the headquarter and F I G U R E 7 Operational mechanism of intelligent operation and management.
formed a more professional and flat management mode in the projects under construction, such as Shuangjiangkou and Jinchuan dams, breaking the traditional barriers between departments and management layers and achieving optimization of human resources, streamlining of management bodies and more efficient, scientific and transparent decision-making.

| Transformation of production management from manual to intelligent
Initially, the company has achieved integration of the physical and digital enterprises and human-machine collaboration on the production line, which has significantly reduced labor intensity and risk probability of staff. Innovative technologies, such as intelligent inspection robots, multisystem intelligent linkage platforms, and virtual plants developed by the company, have effectively reduced various types of repetitive, high-risk, and lowtechnology work. A production mode of "unattended operation or few on duty" has been put in place, which results in the intelligent and autonomous operation of hydropower projects with an installed capacity of 10 million kW in the Dadu River Basin.

| Transformation of decision-making from experience-based to data-based
After the application of intelligent "one-key adjustment" technology, the annual load adjustment is reduced by about 30,000 times, realizing zero intervention in the load adjustment of multiple power stations throughout the year. The application of advanced research fruits, such as quantitative precipitation forecasting, flood resources exploitation, intelligent dispatching decision support, and economic dispatching control has significantly improved the generation efficiency of the enterprise. The online equipment monitoring platform has increased the equipment equivalent availability factor by approximately 10%.

| Change of risk control from passive to predictive
The application of an intelligent overhaul system has enabled the objective evaluation of equipment status.
High-precision hydrological meteorological observation and forecasting means are used to quantitatively analyze the flood and rainfall information of the basin and automatically deduce the flood regulation process, which provides a decision-making basis for flood control. Thanks to the application of geohazard prediction and early warning technology, more than 10 warnings have been successfully completed for geohazards in the Dadu River Basin, avoiding major casualties and economic losses. Furthermore, the consecutive safe production days in the Dadu River Company has exceeded 6300 days by November 2022.

| Transformation of the workforce from production-oriented to innovation-oriented
The innovative and compound power talents have successfully incubated scientific and technological innovation enterprises, which produce intelligent inspection robots, smart safety helmets, and smart keys, and so on. In the past 5 years since the Dadu River Company started the exploration and practice of intelligent O&M in 2014, it obtained a total of 350 intellectual property rights authorizations, 5.6 times more than those in the previous 15 years; it participated in compiling 39 national and industry standards; it pocketed 65 national and provincial science and technology advancement awards, 1.1 times more than all those in the previous 15 years.

| CONCLUSION
Recently, the Dadu River Company has planned a threestep approach to future intelligent O&M.
Step 1, the number of operating staff will be reduced by one-third within 3 years, and management departments of the operating units will be further merged to fully promote the mode of "remote cloud office"; field divisions will be set up to be responsible for on-site emergency response; daily management and operations will be carried out at the head office of Dadu River Company in Chengdu, Sichuan.
Step 2, all power plants will be integrated into three upstream, midstream, and downstream management units according to their geographical location to form a "1 + 3" regional operation model in the basin in another 5 years. A centralized and unified production and scheduling center will be responsible for basin-wide equipment operating, production scheduling, basin operation, and risk monitoring so that more personnel will be freed from front-line production to innovation, management and decision-making.
Step 3, as the physical enterprise and the digital enterprise operate in synergy, the hierarchical management in the physical enterprise will be completely eliminated, and the digital-driven decision-making management form will be fully revealed. In this way, all power stations in the basin will operate in an intelligent and autonomous manner, with the elimination of hierarchical management. Employees will be mainly engaged in innovative work.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.

ETHICS STATEMENT
None declared.