Self-adaptive terminal air volume compensation control system and method for construction ventilation of underground powerhouse of pumped storage power station and application of self-adaptive terminal air volume compensation control system and method

Abstract

The invention discloses a self-adaptive terminal air volume compensation control system and method for construction ventilation of an underground powerhouse of a pumped storage power station and application, and belongs to the field of intelligent ventilation control. Comprising a ventilation parameter sensing module, a digital twin modeling module, an intelligent control decision module and an autonomous optimization module. Through a mechanism and data fusion digital twinning technology and in combination with an innovative control and optimization strategy, the response speed, the control precision and the energy efficiency of the system are remarkably improved, response to faults such as air leakage is converted into active early warning and accurate positioning from passive maintenance, and the fault diagnosis accuracy is improved. Safer, more reliable and more economical ventilation guarantee is provided for underground engineering construction. The problems of control response lag, air volume compensation misalignment, high energy consumption and the like caused by pipeline aging and sudden air leakage of a traditional ventilation system under long-distance and multi-branch complex working conditions are solved.

Description

Technical Field

[0001] This invention relates to the field of intelligent ventilation control in underground engineering, specifically to a dynamic air volume compensation control system, method, and application applicable to complex working conditions such as long-distance, multi-branch, and variable cross-section tunnel groups, and is particularly applicable to the construction ventilation process of underground powerhouses in pumped storage power stations. Background Technology

[0002] During the construction of underground powerhouses in pumped-storage hydroelectric power stations, a long-distance ventilation duct network is required to continuously supply fresh air to the terminal working face to ensure the safety of workers and the normal operation of construction equipment. Currently, ventilation in underground powerhouse construction mainly employs two control schemes: constant air volume (CAV) control systems and conventional PID feedback control systems. CAV systems set a fixed compensation coefficient based on historical leakage rates, but they cannot adapt to the non-linear increase in leakage rates caused by aging ducts, and their response to sudden leakage events is significantly delayed, making it impossible to judge the leakage status in real time, resulting in inaccurate airflow compensation at the working face. Conventional PID feedback control systems adjust the fan speed based on the terminal air pressure deviation, but they suffer from severe integral saturation, are prone to overshoot during adjustment, and lack the ability to sense the health status of the ducts, making it impossible to predict the risk of sudden leakage changes.

[0003] Under complex operating conditions of long-distance ventilation duct networks and variable cross-section tunnel groups, the existing system has two significant shortcomings: First, the existing air volume compensation mechanism relies on a single source, and the terminal pressure feedback does not integrate multi-dimensional information such as gas concentration distribution, resulting in a lag in the adjustment of sudden air leakage events. Second, the system lacks an energy efficiency optimization module, and the fans operate in non-economic operating conditions for extended periods.

[0004] The following problems exist in the existing ventilation control technology for large-scale underground engineering projects such as pumped storage power stations:

[0005] 1. Traditional control systems (such as CAV or conventional PID) cannot adapt to dynamic changes in operating conditions such as pipeline aging and sudden air leakage, resulting in inaccurate air volume compensation and low control precision at the terminal working face.

[0006] 2. The control system has a slow response and lacks the ability to perceive and predict the overall health status of the pipeline network, making it unable to anticipate and respond quickly to emergencies such as air leakage.

[0007] 3. The system lacks an energy efficiency optimization mechanism, and the wind turbines operate in non-economic operating conditions for a long time, resulting in serious energy waste and high operating costs.

[0008] Therefore, there is an urgent need to develop an intelligent ventilation system with dynamic airflow compensation and energy efficiency optimization to ensure construction safety in high-risk environments. Summary of the Invention

[0009] To address the aforementioned technical problems, this invention provides an adaptive terminal air volume compensation control system and method for ventilation during construction of underground powerhouses in pumped storage power stations. This system enables accurate prediction of air volume demand and significantly reduces the long-term operating costs of the ventilation system.

[0010] This invention is implemented as follows: an adaptive terminal airflow compensation control system for construction ventilation of underground powerhouse in pumped-storage power stations, comprising:

[0011] The ventilation parameter sensing module is used to collect the operating parameters of the ventilation duct network in real time through multi-source sensors deployed in the ventilation duct network. The operating parameters include at least temperature, humidity, wind speed and CO2 concentration.

[0012] The digital twin modeling module is used to receive the operating parameters and perform calculations based on a preset digital twin model. The digital twin model includes: a mechanism model unit, used to calculate the frictional resistance and local resistance of the ventilation duct network based on fluid dynamics principles and real-time wind speed data; a data model unit, including at least one prediction model, used to predict future air volume demand based on historical operating parameters; and at least one mapping model, used to learn and establish a nonlinear mapping relationship between air volume demand and fan control parameters.

[0013] The intelligent control decision module is used to generate wind turbine control commands based on the calculation results of the digital twin modeling module through a feedforward-feedback dual-loop control logic.

[0014] The autonomous optimization module is used to periodically employ a multi-objective optimization algorithm with system energy consumption and terminal air volume as at least two optimization objectives to solve for the optimal combination of operating points of each fan in the system, and use this combination to update the regulation benchmark of the intelligent control decision module.

[0015] Furthermore, the multi-source sensor deployment method in the ventilation parameter sensing module is as follows: a monitoring section is set every 100 meters along the axis of the main ventilation duct, and the monitoring section includes at least one wind speed sensor and one temperature and humidity sensor; at least two redundant wind speed sensors are set at the inlet of each branch pipe; and CO2 sensors are deployed at 50-meter intervals in the end working area according to the distribution density of construction machinery.

[0016] Furthermore, the digital twin modeling module also includes a leakage location unit, which is used to: abstract the ventilation duct network into a graph structure composed of nodes and edges; define the residual weight of the edges based on the difference between the theoretical air volume and the measured air volume of each duct segment; and use a graph theory partitioning algorithm to find the maximum residual cut set in order to locate the most likely leakage path.

[0017] Furthermore, the feedforward-feedback dual-loop control logic in the intelligent control decision module is specifically as follows:

[0018] The feedforward loop calculates the target speed of the fan in advance based on the future air volume demand predicted by the data model unit, in order to compensate for the system inertial delay.

[0019] The feedback loop obtains the air volume deviation by comparing the measured air volume at the end working face with the target air volume, and uses a nonlinear control algorithm to dynamically calculate the fan speed adjustment based on the air volume deviation and the deviation change rate.

[0020] Furthermore, the nonlinear control algorithm adopts the fuzzy PID algorithm. The control rule base of the fuzzy PID algorithm is optimized by combining offline simulation and online learning. The membership function parameters are automatically corrected every 24 hours based on the historical control effect to improve control robustness.

[0021] Furthermore, the autonomous optimization module supports the coordinated control of multiple different types of fans in the system. It solves the optimal combination of fans in the branch pipeline by enumeration method, and optimizes the speed of the main fan in the main ventilation duct by gradient descent method under this constraint.

[0022] An adaptive terminal airflow compensation control method using the above system includes the following steps:

[0023] S1: Data acquisition, through multi-source sensors deployed in the ventilation duct network, to collect the operating parameters of the ventilation duct network in real time;

[0024] S2: Modeling and calculation, inputting the operating parameters into the digital twin model, this step includes: (a) calculating the pipeline resistance based on the principle of fluid dynamics; (b) using a data-driven model to predict future air volume demand and establish a mapping relationship between air volume and fan control parameters;

[0025] S3: Decision generation: Based on the modeling and calculation results, wind turbine control commands are generated through feedforward-feedback dual-loop control logic.

[0026] S4: Optimization iteration, with system energy consumption and terminal air volume as optimization objectives, periodically perform multi-objective optimization, solve and update the control logic's regulation benchmark.

[0027] A computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform the above-described method.

[0028] A computer device includes a memory, a processor, and a program stored in the memory and executable thereon, the program being executed by the processor to implement the steps of the method described above.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] 1. Precise control and rapid response: Through a digital twin model that integrates mechanisms and data, accurate prediction of air volume demand is achieved; the use of feedforward-feedback dual-loop control to compensate for system delays in advance greatly improves the response speed and accuracy of the control system, ensuring that the air volume at the terminal working face is stable and meets the standards.

[0031] 2. Intelligent traceability and proactive operation and maintenance: The integrated air leakage location unit can reduce the location time of air leakage events from several hours to minutes and achieve meter-level accurate location, transforming passive emergency repairs into proactive operation and maintenance, significantly improving construction safety and system reliability.

[0032] 3. Autonomous optimization and high efficiency: Through periodic Pareto optimization of air volume and energy consumption, the system can autonomously find and operate at the global optimal energy efficiency point, achieving simultaneous improvement in control accuracy and overall energy efficiency, and significantly reducing the long-term operating cost of the ventilation system.

[0033] 4. Robust and adaptive evolution: Through online learning and correction of the fuzzy PID rule base and the closed-loop evolution of the entire control strategy, the system can continuously adapt to long-term effects such as pipeline aging and changes in operating conditions, reducing the need for manual intervention and improving the level of automation. Attached Figure Description

[0034] Figure 1 This is a diagram of the overall system architecture of the present invention;

[0035] Figure 2 This is a schematic diagram of the ventilation duct network and sensor layout of the present invention;

[0036] Figure 3 This is a schematic diagram of the digital twin modeling principle;

[0037] Figure 4 It is a flowchart for the coordinated optimization of control and energy efficiency. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0039] Example 1

[0040] This embodiment provides an adaptive terminal air volume compensation control system and method for construction ventilation of underground powerhouse of pumped storage power station.

[0041] Reference Figure 1 Logically, the system comprises: a ventilation parameter sensing module 100, a digital twin modeling module 200, an intelligent control decision-making module 300, and an autonomous optimization module 400. These four modules work together to form a closed-loop control process of "sensing-modeling-decision-optimization".

[0042] I. System Physical Deployment and Implementation of Ventilation Parameter Sensing Module 100

[0043] Reference Figure 2 The physical application scenario of this embodiment is a ventilation duct network in an underground factory. This network includes a main ventilation duct ( Figure 2 The main ventilation path extends from the ground-level fan room entrance to the underground main plant area. Three branch ventilation ducts extend from the main ventilation duct. Figure 2 The secondary ventilation paths lead to construction adits A, B, and C respectively. The final destination of the ventilation is the construction working face at the end of each adit.

[0044] The ventilation parameter sensing module 100 is implemented through multi-source sensors deployed on this physical pipe network. The specific deployment method is as follows:

[0045] 1. Wind speed and temperature / humidity monitoring: A monitoring section is installed every 100 meters along the axis of the main ventilation duct. Each monitoring section is equipped with a set of sensors, including one wind speed sensor and one temperature / humidity sensor. In this embodiment, the wind speed sensor is a German Testo 405i (range 0-20 m / s), and the temperature / humidity sensor is a Swiss Sensirion SHT35. At the inlet of each branch ventilation duct, two wind speed sensors are installed side-by-side to achieve data redundancy and cross-validation.

[0046] 2. CO2 Concentration Monitoring: In the terminal construction work area, based on the typical distribution of construction equipment and personnel, a CO2 sensor (e.g., ...) is deployed every 50 meters along the construction area. Figure 2 (The CO2 sensor is marked as shown in the image). This embodiment uses the British GSS sensor. (Measuring range 0-5000ppm).

[0047] The analog signals acquired by all sensors undergo preliminary processing by a field edge controller (e.g., Siemens SIMATIC S7-1500). This processing includes:

[0048] 1. A / D conversion: Converting analog signals into digital signals.

[0049] 2. Data filtering: A moving average filtering algorithm with a window length of 10 seconds is applied to the acquired high-frequency digital signal to filter out instantaneous disturbances and obtain stable operating parameter values.

[0050] 3. Data Upload: The filtered data, along with the timestamp, is sent to the digital twin modeling module 200 via industrial Ethernet. II. Implementation of the Digital Twin Modeling Module 200

[0051] Reference Figure 3 The digital twin modeling module 200 receives real-time data from the perception module 100, and its core is the fusion of the mechanism model unit and the data model unit.

[0052] 1. Calculation process of the mechanistic model unit:

[0053] This unit calculates the real-time resistance characteristics of the pipeline network based on fluid dynamics principles.

[0054] Step 1: Obtain geometric and physical parameters. Automatically extract the length L and diameter D of each ventilation duct section from the pre-built 3D BIM model of the factory building. For example, for a main ventilation duct section, L = 450m and D = 1.2m. The air density ρ is taken as 1.293 kg / m³. 3 .

[0055] Step 2: Calculate friction drag. Receive real-time wind speed data (v) and calculate friction drag using the Darcy-Wiesbach equation.

[0056]

[0057] In the formula:

[0058] ΔP: Friction loss along the route (Pa);

[0059] f: Darcy friction coefficient;

[0060] L: Pipe length (m), automatically extracted from the BIM model;

[0061] D: Pipe diameter (m), measured value 1.2m;

[0062] ρ: Air density (kg / m³) 3 ), take 1.293 kg / m 3 ;

[0063] υ: Air velocity inside the duct (m / s), from real-time data from sensors.

[0064] The friction coefficient f is solved iteratively using the Colbrook-White formula and dynamically corrected based on the pipeline operating time Δt (months) to reflect pipeline aging.

[0065]

[0066] In the formula:

[0067] ε: Pipe roughness (mm), 0.15mm for galvanized steel plate;

[0068] Re: Reynolds number (Where v is the wind speed inside the pipe, D is the pipe diameter, and ν is the kinematic viscosity of air, 1.5 × 10⁻⁶) -5 m 2 / s).

[0069] Finally, the local drag coefficient is corrected every thirty seconds based on wind speed sensor data.

[0070] ξ=ξ0·(1+0.02Δt)

[0071] In the formula:

[0072] ξ0: Initial local drag coefficient (design value);

[0073] Δt: Pipeline operating time (in months), reflecting the degree of aging.

[0074] Data Example: Assume a pipe with L = 100m and D = 1.2m. The sensor measures a real-time wind speed v = 15m / s and an air kinematic viscosity ν = 1.5 × 10⁻⁶ m / s. -5 m 2 / s. First, calculate the Reynolds number Re = vD / ν ≈ 1.2 × 10⁶. Assuming the pipe roughness ε = 0.15 mm, the Kohlbrook-White formula is solved iteratively to obtain f ≈ 0.0135. Then, the friction loss along the pipe section ΔP = 0.0135 * (100 / 1.2) * (1.293 * 15) 2 / 2)≈163.4Pa.

[0075] 2. Calculation process of data model unit:

[0076] This unit uses historical data for prediction and mapping.

[0077] Future air volume demand forecast: An ARIMA(p,d,q)=(3,1,2) time series forecasting model is used. This model is trained based on terminal CO2 concentration and air volume data over a past period (e.g., 24 hours). Its calculation logic is as follows: Based on historical data sequences, predict the terminal air volume demand Qt+5 for the next 5 minutes. The prediction equation is as follows:

[0078]

[0079] In the formula:

[0080] The autoregressive coefficients are obtained through training with historical data.

[0081] θ j Moving average coefficient;

[0082] ε: white noise sequence.

[0083] Airflow-Speed ​​Nonlinear Mapping: An LSTM neural network is employed. The network takes historical operating data (including wind speed, CO2 concentration, valve opening, and fan speed) as input over 60 time steps and outputs the target fan speed required to achieve a specific airflow demand. The network learns the nonlinear relationship under historical operating conditions through offline training. The LSTM network structure is as follows:

[0084] Input layer: Historical data for 60 time steps (wind speed, valve opening, fan speed);

[0085] Hidden layers: 2 layers, 128 neurons per layer, with ReLU activation function;

[0086] Output layer: 1 neuron (predicts wind turbine speed), activation function is Sigmoid;

[0087] Training data: operating data from the past 30 days, batch size 32, and 1000 iterations of intelligent control decision-making.

[0088] 3. Implementation of the air leakage positioning unit:

[0089] 1) Anomaly detection triggering mechanism

[0090] When the system detects an anomaly, it will trigger the air leakage location calculation process. The specific judgment conditions are as follows:

[0091] Airflow deviation threshold: |Q 实际 -Q 预测 |≥5%·Q 设计 ;

[0092] CO2 concentration gradient anomaly: CO2 concentration in the working area exceeds 500 ppm and the concentration difference between adjacent sensors is ≥100 ppm / m;

[0093] Pressure drop: The rate of pressure drop at the branch pipe inlet is ≥10 Pa / min.

[0094] 2) Pipeline segment diagnosis

[0095] Step 1: Construct a weighted graph. Abstract the ventilation duct network as a graph G = (V, E), where node V represents the sensor location and edge E represents the duct segment.

[0096] Step 2: Calculate residual weights. For each edge (pipe segment ij), define the weights for each edge e. ij Residual weights:

[0097]

[0098] Where, L ij This refers to the length of the pipe section.

[0099] Step 3: Find the maximum residual cut set. The Ford-Fulkerson algorithm is used to find a cut set in graph G that maximizes the sum of the weights of the cut edges. The physical pipe segment represented by this cut set is the most likely leakage area. The objective function is:

[0100]

[0101] The probability of air leakage in each pipe section was calculated using a Bayesian probability model.

[0102]

[0103] In the formula,

[0104] D: Observational data (wind speed, pressure, CO2, etc.);

[0105] Prior probability P (air leakage k): set according to the aging degree of the pipe section / usage time.

[0106] 3) Real-time correction and verification

[0107] After detecting an air leak, add a virtual air leak (area A) to the model. 漏 Recalculate the overall network airflow distribution until the error between the simulation results and the measured data is ≤2%. Apply a short-duration airflow pulse (±5% disturbance) to the inlet of the suspected leaking branch pipe, and analyze the response delay using transfer function analysis.

[0108]

[0109] The response amplitude of leaky branch pipes decays faster and the phase lag decreases. The leakage level classification and response strategy are as follows:

[0110]

[0111] In the implementation case, the leakage detection process for branch pipe B is as follows:

[0112] a) The system detected a 12% decrease in airflow at the working face and an increase in CO2 concentration to 800 ppm;

[0113] b) The digital twin model shows that the theoretical / measured airflow residual R in branch pipe B is 18%.

[0114] c) Pressure gradient analysis indicates that there is an equivalent leakage area A in the middle section of branch pipe B. 漏 =0.05m 2 ;

[0115] d) The temperature and humidity sensor detected a sudden drop in humidity of 5% RH at 150m in branch pipe B;

[0116] e) The graph theory algorithm located the air leakage point in the 140-160m range of branch pipe B (92% confidence level);

[0117] f) Maintenance personnel confirmed on-site that the duct weld in the area was cracked and carried out emergency repairs.

[0118] This method reduces the time for locating air leaks in traditional ventilation systems from several hours to within 10 minutes, with a positioning accuracy of up to 5% of the pipe section length. Maintenance personnel were able to quickly locate and repair cracked welds using this method.

[0119] III. Implementation of the Intelligent Control Decision Module 300

[0120] Reference Figure 4 Based on the calculation results of the digital twin module 200, this module generates the final wind turbine control command through feedforward-feedback dual-loop control logic.

[0121] Feedforward loop: Its purpose is to compensate for system inertial delay in advance. The calculation steps are as follows:

[0122] Receive the air volume demand Qt+5 predicted by the data model unit for the next 5 minutes, and then calculate it according to the formula.

[0123]

[0124] In the formula:

[0125] Kf: Feedforward gain coefficient, dynamically adjusted through the LSTM model;

[0126] τ d Pipeline delay time (measured value, average 45 seconds);

[0127] T s Control cycle (10 seconds).

[0128] The base target rotational speed is calculated. Kf is the feedforward gain coefficient, which is dynamically adjusted by the LSTM model.

[0129] Feedback loop: Its purpose is to eliminate the deviation between the current actual air volume and the target air volume. The calculation steps are as follows:

[0130] Step 1: Calculate the deviation. Calculate the current airflow deviation e = Q 设定 -Q 实际 and rate of change of deviation

[0131] Step 2: Fuzzification. Input e and ec into 7 membership functions (such as NB (negative large), NM (negative medium), NS (negative small), ZO (zero), PS (positive small), PM (positive medium), PB (positive large)) to obtain their membership degrees on each fuzzy subset.

[0132] Step 3: Fuzzy Inference. The output variable is the fan speed adjustment amount Δn. Using the same fuzzy partitioning, a nonlinear control surface is constructed through 49 fuzzy rules. When the air volume is severely insufficient (e = PB) and the deviation continues to increase (ec = PB), the "PB→PB" rule is triggered, and the maximum positive speed compensation is output. If the air volume is close to the set value (e = ZO) but there are small fluctuations (ec = NS), the "ZO→PS" rule is executed for fine-tuning, and Mamdani inference is performed to obtain the fuzzy speed adjustment amount Δn.

[0133] Step 4: Defuzzification. The precise speed adjustment Δn_feedback is calculated using the center of gravity method.

[0134] Step 5: Integral Saturation Prevention. In PID calculations, the sat(x) function is introduced to limit the accumulation of the integral term. When the deviation e exceeds a preset threshold, integral accumulation is temporarily stopped to ensure that the overshoot is always ≤3%. The anti-integral saturation mechanism uses conditional integration, freezing the integral term when the error exceeds the threshold.

[0135]

[0136] The final calculated target wind turbine speed is n_target = n_feedforward + Δn_feedback. This target speed will be encapsulated into a speed control command and sent to the wind turbine's frequency converter via the network for execution.

[0137] Level 3 response strategy:

[0138] Level 1 Response (Normal Operating Condition): Output fan speed command n target = your feedforward + Δ your feedback, and maintain the current value of the damper opening.

[0139] Secondary response (residual 5%~15%): triggers coordinated regulation of branch dampers.

[0140]

[0141] In the formula:

[0142] α j : Opening degree of the j-th branch damper; ΔQ j The airflow deviation in this branch.

[0143] Level 3 response (residual > 15%): Start the backup fan to run in parallel and send an alarm signal to the operation and maintenance terminal (including air leakage location information).

[0144] IV. Implementation of the Autonomous Optimization Module 400

[0145] This module, acting as an auxiliary to the control layer, periodically (e.g., every 10 minutes) performs global energy efficiency optimization. Its optimization process is as follows:

[0146] Define the optimization problem: the objective function is min P_total = ΣP_fan (minimize total power consumption), and the constraint is ΣQ_terminal ≥ Q_demand * 105% (satisfy total air volume demand).

[0147] Solution Method: A multi-objective optimization algorithm combining enumeration and gradient descent is employed. In the multi-fan scenario of this embodiment, the main fan speed is first fixed, and the optimal speed combination of each branch fan is solved using enumeration. Then, the operating conditions of the branch fans are used as constraints, and the gradient descent method is used to optimize the main fan speed. The objective function is as follows:

[0148]

[0149] Application of Results: The optimal operating point in the Pareto optimal solution set (e.g., main fan speed 2900 rpm, A / B / C branch fan speeds 2100 / 1800 / 1800 rpm respectively), and its corresponding target air volume and speed values ​​will be used as the update benchmark for the feedforward control loop in the intelligent control decision module 300, thereby realizing the closed-loop evolution of the system.

[0150] Data Case: In branch tunnel A, the main fan (ZTF-55kW) and branch tunnel fan (ZTP-22kW) are mixed and controlled. The engineering example configuration is as follows:

[0151]

[0152] In the actual project, a sudden air leak occurred in branch tunnel A, increasing the required total air volume to 32m³. 3 / s, Current operating status of main fan ZTF-55kW: 2700rpm / 28m 3 / s; Branch tunnel fan ZTP-22kW: A-1800rpm, B-1500rpm, C-1650rpm.

[0153] Pareto generates 6 nondominated solutions, with the optimal solution being:

[0154]

[0155] Total air volume: 30.2 + 3.1 + 2.8 × 2 = 38.9 m³ 3 / s, total power: 55.3kW.

[0156] Feedforward injection: nZTF=2900 was directly written into the feedforward control command, and the branch tunnel fan started in a stepped manner according to the optimized results, so that the actual air volume reached 36.7m³. 3 / s, with a total power of 57.1kW, which is 12% lower than manual control.

[0157] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented, in whole or in part, as a computer program product, the computer program product includes one or more computer instructions. When the computer program instructions are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape) or an optical medium.

[0158] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and these variations are still within the protection scope of the present invention.

Claims

1. An adaptive terminal airflow compensation control system for construction ventilation of underground powerhouse of pumped storage power station, characterized in that, include: The ventilation parameter sensing module is used to collect the operating parameters of the ventilation duct network in real time through multi-source sensors deployed in the ventilation duct network. The operating parameters include at least temperature, humidity, wind speed and CO2 concentration. The digital twin modeling module is used to receive the operating parameters and perform calculations based on a preset digital twin model. The digital twin model includes: a mechanism model unit, used to calculate the frictional resistance and local resistance of the ventilation duct network based on fluid dynamics principles and real-time wind speed data; a data model unit, including at least one prediction model, used to predict future air volume demand based on historical operating parameters; and at least one mapping model, used to learn and establish a nonlinear mapping relationship between air volume demand and fan control parameters. The intelligent control decision module is used to generate wind turbine control commands based on the calculation results of the digital twin modeling module through a feedforward-feedback dual-loop control logic. The autonomous optimization module is used to periodically employ a multi-objective optimization algorithm with system energy consumption and terminal air volume as at least two optimization objectives to solve for the optimal combination of operating points of each fan in the system, and use this combination to update the regulation benchmark of the intelligent control decision module.

2. The adaptive terminal airflow compensation control system for construction ventilation of underground powerhouse of pumped storage power station according to claim 1, characterized in that, The multi-source sensor deployment method in the ventilation parameter sensing module is as follows: a monitoring section is set every 100 meters along the axis of the main ventilation duct, and the monitoring section includes at least one wind speed sensor and one temperature and humidity sensor; at least two redundant wind speed sensors are set at the inlet of each branch pipe; and CO2 sensors are set at 50-meter intervals in the end working face area according to the distribution density of construction machinery.

3. The adaptive terminal airflow compensation control system for construction ventilation of underground powerhouse of pumped storage power station according to claim 1, characterized in that, The digital twin modeling module also includes a leakage location unit, which is used to: abstract the ventilation duct network into a graph structure composed of nodes and edges; define the residual weight of the edges based on the difference between the theoretical air volume and the measured air volume of each duct segment; and use a graph theory partitioning algorithm to find the maximum residual cut set in order to locate the most likely leakage path.

4. The adaptive terminal airflow compensation control system for construction ventilation of underground powerhouse of pumped storage power station according to claim 1, characterized in that... The feedforward-feedback dual-loop control logic in the intelligent control decision module is specifically as follows: The feedforward loop calculates the target speed of the fan in advance based on the future air volume demand predicted by the data model unit, in order to compensate for the system inertial delay. The feedback loop obtains the air volume deviation by comparing the measured air volume at the end working face with the target air volume, and uses a nonlinear control algorithm to dynamically calculate the fan speed adjustment based on the air volume deviation and the deviation change rate.

5. The adaptive terminal airflow compensation control system for construction ventilation of underground powerhouse of pumped storage power station according to claim 4, characterized in that... The nonlinear control algorithm adopts the fuzzy PID algorithm. The control rule base of the fuzzy PID algorithm is optimized by combining offline simulation and online learning. The membership function parameters are automatically corrected every 24 hours based on the historical control effect to improve the control robustness.

6. The adaptive terminal airflow compensation control system for construction ventilation of underground powerhouse of pumped storage power station according to claim 1, characterized in that, The autonomous optimization module supports the coordinated control of multiple different types of fans in the system. It solves the optimal combination of fans in the branch pipeline by enumeration method, and optimizes the speed of the main fan in the main ventilation duct by gradient descent method under this constraint.

7. An adaptive terminal airflow compensation control method using the system described in any one of claims 1-6, characterized in that, Includes the following steps: S1: Data acquisition, through multi-source sensors deployed in the ventilation duct network, to collect the operating parameters of the ventilation duct network in real time; S2: Modeling and calculation, inputting the operating parameters into the digital twin model, this step includes: (a) calculating the pipeline resistance based on fluid dynamics principles; (b) Use a data-driven model to predict future air volume demand and establish a mapping relationship between air volume and fan control parameters; S3: Decision generation: Based on the modeling and calculation results, wind turbine control commands are generated through feedforward-feedback dual-loop control logic. S4: Optimization iteration, with system energy consumption and terminal air volume as optimization objectives, periodically perform multi-objective optimization, solve and update the control logic's regulation benchmark.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform the method of claim 7.

9. A computer device, characterized in that, The computer device includes a memory, a processor, and a program stored in and executable on the memory, the program being executed by the processor to perform the steps of the method as described in claim 7.

Images