The basic idea of "twin" originated from the Apollo program in 1969. The spacecraft that stayed on the earth simulated the working state of the spacecraft launched into space, and then assisted the astronauts to complete the decision-making and reduce various operational results. the unknown.
Since 2002, the concept and definition of digital twins have been gradually raised and valued in different fields. The definition of digital twins will be different in different time stages and different application fields. The definition of each stage is as follows:
Phase 1: The concept of "Digital Twin" was first proposed
The digital model constructed in the virtual space is interactively mapped with the physical entity, faithfully describing the running trajectory of the physical entity in its entire life cycle.
Phase 2: The concept of "fuselage digital twin" is proposed for the first time
A twin is an integrated system of data, models, and analysis tools. The system can not only represent the aircraft airframe throughout its life cycle, but also make decisions on the entire fleet and individual airframes based on non-deterministic information, including current diagnosis and future prediction.
Phase 3: The "digital twin" is used independently for the first time
A system or aircraft simulation process that integrates multiple physical quantities, multiple scales, and multiple probabilities.
Phase 4: Standardization Organization Conducts Research
A digital twin is a digital representation of a real thing or process with a specific purpose, and the synchronization between the physical instance and the digital instance tends to be consistent through appropriate frequency synchronization.
Phase 5: Domestic Organizations Conduct Research
Digital twin is a comprehensive use of information technologies such as perception, computing, and modeling, through software definition, to describe, diagnose, predict, and make decisions on physical space, and then realize the interactive mapping between physical space and virtual space.
From the comparative analysis of these definitions, several key points of digital twins can be found:
-
A digital twin is a high-fidelity digital modeling of a physical entity
-
Physical entities and digital twins need to interact with each other, real-time mapping and control
-
The digital twin has intelligent capabilities and can diagnose, predict, and simulate physical spaces
Modeling: Digital Twin Key Points
From the definition and characteristics of digital twins, it can be found that modeling is the core key point of creating digital twins, and it is also the basis for upper-level operations of digital twins. Modeling includes not only three-dimensional modeling of the geometric structure and shape of the physical entity, but also full digital modeling of the physical entity's own operating mechanism, internal and external interfaces, software, and control algorithms.
Digital twin modeling has strong special features and is closely related to the fields of its twin application. Combined with the application scenario of using digital twin technology in the computer room, Haojing Technology needs to model the entire computer room structure and equipment status, abstractly summarizes that the corresponding twins need to have six types of data, and constructs the "six-sided twin" of the computer room scene body modeling” method.
Side 1: Data Information Model
Define the basic information of twins, this part is basically consistent with the traditional management method. For example: assign a Chinese name to the equipment, and record its manufacturer, model, price, purchase time, power, maintenance personnel, etc. in the database to facilitate daily management and maintenance.
Second side: Geometric location model
Digitally describe two pieces of information: ①Simulation depicts the appearance and size of the twin body. By simulating its geometric model, texture map, ambient lighting, and state characterization performance, the target and the real object are presented 1:1. ②Record its spatial coordinates, the outdoor equipment records the GIS coordinates (longitude, latitude), and the indoor equipment records its coordinates (x, y, z) in the relative space.
The third aspect, performance data model
Simulate the real-time running status of the object, and record the energy consumption, capacity, bandwidth, and delay of the equipment and network in real-time. And through the analysis of performance indicators, simulate the health of the twins, prevent performance deterioration, and eliminate hidden dangers before they happen.
The fourth aspect, alarm data model
When a failure occurs in reality, the twin body will send out an alarm. There are two situations: ①The alarm is internally generated, and when the performance deteriorates to the threshold (or meets the fifth rule), the alarm is automatically generated. For example: the sensor temperature exceeds the allowable value alarm. ②Passive alarm, after the twin body receives the equipment alarm information, it generates an alarm.
The fifth aspect, the operation rule model
Define the business logic inside the twin, which is different from the traditional data model. The specific performance of this part can be a piece of code, an algorithm, an expression or a process. For example: when the outdoor temperature is lower than the temperature in the computer room, execute the heat exchange strategy, open the fresh air in the computer room, and gradually increase or turn off the air conditioner. The logic of this execution process can be defined as a running rule model.
Sixth aspect, control feedback model
Operate the twin body, and after the command is issued, its information changes, and the simulation feedback is given. For example: Press the temperature down button of an air-conditioning twin, the system sends an instruction to the real air-conditioning system, and feedbacks its operating temperature after successful execution, and the temperature environment changes in the computer room are reflected in real-time on the temperature cloud map.
The twin computer room built based on the above modeling method can be applied to the planning scenario of the computer room space, the operation and maintenance and control scenario of the equipment in the computer room, the simulated cutover scenario of the equipment in the computer room, the failure/hidden danger analysis and prediction of the computer room, etc. Business scenarios .
The temperature control of the computer room on the basis of the twin computer room will change the traditional method of temperature control in the computer room, which can only be checked by the operation and maintenance personnel to check the value of the temperature sensor and monitor the alarm information of the sensor, and to adjust the state of the refrigeration system manually; Realize computer room temperature distribution simulation and simulation, advance prediction and multi-means automatic regulation, change passive regulation to active regulation, ensure that the temperature of the computer room is always maintained within a safe range, and further realize the PUE pressure drop of the computer room.
New solution to the old problem of temperature control in the computer room
In the traditional mode, there is no intuitive display of various equipment layouts and temperature distributions in the entire computer room space; the control method is relatively simple and relies on operation and maintenance experience; it is impossible to actively adjust the temperature in advance. Through the introduction of digital twin technology, combined with the six-sided model of the computer room for data fusion processing and analysis, the twin visualization of the computer room, the prediction and simulation of the three-dimensional temperature of the computer room, and the active temperature control of multi-factor fusion are realized.
At this stage, the goal of twin machine room temperature control is to realize the temperature trend prediction of the machine room, the optimal control of multi-parameters of refrigeration equipment/units, and the simulation of the effect of refrigeration equipment parameter adjustment.
Combined with the twin characteristics of various equipment in the computer room, it is also possible to further adjust the heat source equipment, shut down the equipment, reduce the heat generation, optimize the heat dissipation, etc., and realize the temperature control of the computer room from the source. Including but not limited to the following scenarios:
-
Rack U-bit capacity planning, simulation equipment rack position adjustment, optimization of heat dissipation and temperature propagation
-
Business analysis of high-heating equipment, simulation of service cutover, equipment cutover or network withdrawal
-
Power equipment load analysis to reduce no-load heat loss
-
Temperature control of lighting/monitoring equipment to realize automatic switching
1 Modeling of the geometric location of the computer room
In order to fully realize the entire business scenario process, starting from the geometric modeling of the park, to the building floors, computer rooms, and various equipment in the computer room (racks, power equipment, refrigeration equipment, IT equipment, sensors, monitoring equipment, etc.), and Geometric modeling of various pipelines ensures the integrity of business operations in twin scenarios.
Through the computer room space location designer, the modeling of the relative position of various equipment in the computer room can be flexibly realized, so as to realize the scene construction of twin computer rooms and prepare for various subsequent business applications.
2 Computer room performance and alarm modeling
For the performance and alarm data in the temperature control scene of the computer room, it is the real-time three-dimensional temperature cloud map of the computer room, and real-time calculation and analysis of the heat island area exceeding the limit from the cloud map.
When modeling the three-dimensional temperature cloud map of the computer room, it is necessary to obtain the real-time temperature data of the temperature sensor (the performance model of the sensor) and the temperature alarm data exceeding the limit (the alarm model of the sensor), and combine the spatial position of the sensor in the computer room, Realize the modeling of the three-dimensional temperature cloud map of the computer room, so as to realize the precise control of the temperature of each area in the computer room.
3 Modeling of computer room operation rules
In order to change the current passive control to active control, there are three main aspects related to the operating rules in the process of temperature control in the computer room:
Computer room temperature change prediction
Each twin body in the computer room records the index data of temperature, and uses the trend prediction algorithm to predict the temperature change of the computer room in advance through the accumulated historical data. The trend prediction algorithm can use the time series decomposition algorithm and the LSTM algorithm to train the model The sensor temperature is used as the main data, and the temperature and heat dissipation of IT equipment, seasonal room temperature, and holiday information are used as influencing parameters, and an applicable trend prediction model is comprehensively trained.
Optimal Path Refrigeration Equipment Search
After the location of the heat island area is found in the three-dimensional temperature cloud map of the computer room, the optimal path can be found through the space distance of the computer room, the design of the air duct path and other information combined with the spatial path search ability, and the refrigeration equipment near the heat island can be found, thus obtaining the most direct temperature control object . Corresponding search methods include: spatial distance range search, air duct vector path search, etc.
Optimal Energy Saving Parameter Search for Multiple Temperature Control Strategies
When data centers design their air conditioning systems, they often introduce a variety of cooling and energy saving modes, such as:
-
Dual-cooling source air-conditioning unit: According to different seasons and different indoor and outdoor temperature differences, choose to use air-cooled heat exchange, water-cooled heat exchange, or heat exchange through compressors
-
Fresh air system: When the outdoor temperature is lower than the indoor temperature in winter, the fresh air system is directly used to cool down, reducing the use of air conditioners and saving energy
The analysis found that when there is a heat island in the computer room, in addition to finding the optimal air conditioner, there are other control methods, such as: stop the air conditioner + start the fresh air; turn off the fresh air + start the air conditioner compressor; turn off the air conditioner compressor + start the water cooling exchange etc. That is, in addition to meeting the temperature requirements of the computer room, it is also necessary to find the control parameters with the lowest energy consumption.
Through the particle swarm optimization algorithm (Particle Swarm Optimization, PSO), search for the optimal solution in various cooling modes, quickly obtain and locate the state of the refrigeration equipment, give control adjustment parameters, and ensure energy efficiency under the premise of meeting the cooling requirements Lowest consumption.
4. Feedback modeling of computer room control
The control of the computer room is actually the control of the equipment in the computer room. For the control of the temperature control scene, it is the parameter adjustment of the terminal air conditioning equipment and refrigeration unit.
When defining the control model of air conditioning and refrigeration units, the corresponding instructions and interfaces are defined for the parameters that can be adjusted. Through the control parameters and optimal refrigeration equipment analyzed by the operation rule model of the computer room, the parameters that need to be adjusted are sent to the corresponding refrigeration equipment, so as to realize the temperature control of the computer room.
The key point of the feedback model of the computer room is the simulation feedback, that is, after the operation rule model of the computer room actively analyzes the control parameters, the AI algorithm model is established by establishing the relationship between various control parameters and temperature changes, and the control effect feedback and Simulation, which simulates the control effect after the parameters are delivered. It appears through the simulation of the three-dimensional cloud image of the temperature in the computer room.
Use the AI algorithm to predict temperature changes in advance, intelligently select the strategy with the least energy consumption under various temperature control parameter adjustment strategies, conduct simulations for the selected control parameters, predict and verify the control effect, and finally obtain the optimal control strategy for physical release Execute for air conditioner/BA system.
The six-sided model data of the computer room is integrated, and through the intensive learning and continuous optimization of the above process, the active and precise temperature control of the computer room is finally realized, which further reduces the power consumption PUE of the computer room.