Air compression and supply coordinated ventilation method for small and large fans of ultra-long tunnel

Abstract

The invention discloses an ultra-long tunnel small and large fan air compression and supply coordinated ventilation method, and relates to the technical field of tunnel ventilation, and the method comprises the following steps: collecting tunnel environment parameters in real time, and generating a time-space associated environment parameter data set; a main and auxiliary fan performance attenuation model is combined to establish an air quantity coordination intelligent model, and air quantity dynamic distribution according to needs is achieved; frequency conversion air pressure adjustment, partition air volume supplement and air valve intelligent opening degree adjustment operation are executed; and a multi-parameter fusion deviation evaluation model is constructed, the spatial distribution deviation of the adjusted environmental parameters and target parameters is monitored in real time, fan control parameters are dynamically corrected based on a reinforcement learning algorithm, a self-adaptive ventilation adjustment closed loop is formed, and rapid air volume redistribution under the emergency working condition is carried out. According to the method, dynamic data processing, process adaptive air volume calculation, distance perception distribution, partition regulation and control, reinforcement learning optimization and a multi-dimensional emergency response technology are fused, and double improvement of dynamic cooperation of main and auxiliary fans of the ultra-long tunnel, ventilation energy efficiency and construction safety is achieved.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of tunnel ventilation, in particular to a small and large fan pressure ventilation and air supply coordination method for an ultra-long tunnel. BACKGROUND

[0002] With the extension of traffic infrastructure projects to complex geological conditions, the construction scale of ultra-long tunnels is expanding year by year. During the tunnel construction process, the ventilation system needs to solve the environmental control problems brought by high altitude, long distance and multi-process cross operation: on the one hand, the wind volume significantly decreases as the construction section increases with the increasing depth of excavation, and the traditional single fan air supply mode cannot meet the requirements of wind speed, harmful gas concentration and temperature control at the remote operation face; on the other hand, the different processes such as drilling and blasting, supporting and lining have great differences in the demand for ventilation volume, and the fixed air volume distribution strategy is easy to cause energy waste or insufficient ventilation.

[0003] The existing technology has three major pain points: first, the environmental parameter collection is greatly affected by the tunnel electromagnetic interference and dust, and the data noise causes the wind volume calculation deviation; second, the main and auxiliary fan coordination control lacks a dynamic adjustment mechanism, and the fan output cannot be corrected in real time after the performance decay, which is easy to cause local wind volume excess or deficiency; third, there is no quick response air volume redistribution scheme in emergency conditions (such as sudden increase of harmful gas and fan failure), which has safety hazards. Therefore, it is urgent to build a coordinated ventilation method that takes into account real-time, adaptability and safety. SUMMARY

[0004] To solve the above technical problems, a small and large fan pressure ventilation and air supply coordination method for an ultra-long tunnel is provided, which solves the above problems.

[0005] To achieve the above purposes, the technical scheme adopted by the application is as follows:

[0006] A small and large fan pressure ventilation and air supply coordination method for an ultra-long tunnel, comprising:

[0007] Real-time collection of tunnel environmental parameters, acquisition of wind speed, wind pressure, harmful gas concentration and temperature data of different sections through a distributed sensor array, noise suppression through an adaptive filtering algorithm, dynamic adjustment of the data standardization interval based on the location of the construction section, and generation of a spatiotemporal correlated environmental parameter data set;

[0008] Fusion of tunnel structure parameters, construction progress and real-time environmental data to calculate the total required air volume, introduction of a construction condition coefficient to dynamically correct the air volume distribution ratio, combination of a main and auxiliary fan performance decay model to establish an air volume coordination intelligent model, and realization of dynamic air volume distribution according to demand;

[0009] According to the wind volume coordination intelligent model output main fan frequency conversion curve, wind pressure dynamic adjustment parameters and auxiliary fan partition start-stop timing, angle self-adaptive adjustment scheme, through the edge calculation node real-time issue control instruction, execute variable frequency wind pressure regulation, partition wind volume supplement and wind valve intelligent opening degree regulation operation;

[0010] A multi-parameter fusion deviation evaluation model is constructed to monitor the spatial distribution deviation of the adjusted environmental parameters and target parameters in real time. Based on the reinforcement learning algorithm, the fan control parameters are dynamically corrected to form a self-adaptive ventilation regulation closed loop for rapid redistribution of air volume under emergency working conditions.

[0011] Preferably, the tunnel environment parameters are collected in real time, the wind speed, wind pressure, harmful gas concentration and temperature data of different sections are obtained through a distributed sensor array, and after noise suppression by an adaptive filtering algorithm, the data standardization interval is dynamically adjusted based on the location of the construction section to generate a space-time correlated environmental parameter data set, which specifically includes:

[0012] The distributed sensor array is arranged along the tunnel axis at intervals of 50-100 meters, and each monitoring node integrates a wind speed sensor, a wind pressure transmitter, a harmful gas detector and a temperature sensor;

[0013] The sensor collects raw data at a frequency of 1 per second, synchronously records the data collection time and corresponding tunnel mileage position information, and retrieves historical environmental parameter data from the tunnel management system;

[0014] The correlation analysis method is used to evaluate the correlation degree of each environmental parameter and the ventilation effect, remove redundant monitoring data, and retain key feature parameters;

[0015] An adaptive filtering algorithm based on wavelet transform is used to suppress noise in the raw data, and the standardization interval is dynamically adjusted according to the distance L of the construction section from the tunnel entrance. When L≤1000 meters, the basic interval is used, and when L>1000 meters, the upper limit of the wind speed interval is expanded to 1.8 m / s and the upper limit of the wind pressure interval is expanded to 2200 Pa to construct a space-time correlated environmental parameter data set.

[0016] Preferably, the adaptive filtering algorithm based on wavelet transform is used to suppress noise in the raw data, and the standardization interval is dynamically adjusted according to the distance L of the construction section from the tunnel entrance. When L≤1000 meters, the basic interval is used, and when L>1000 meters, the upper limit of the wind speed interval is expanded to 1.8 m / s and the upper limit of the wind pressure interval is expanded to 2200 Pa to construct a space-time correlated environmental parameter data set, which specifically includes:

[0017] The collected wind speed, wind pressure, harmful gas concentration and temperature data are mapped to the [0, 1] interval using a normalization method;

[0018] Based on the construction section position information, spatial tags are added to the data, and a space-time matrix is formed by sorting in time sequence;

[0019] The missing data is supplemented by the adjacent monitoring point interpolation algorithm to ensure the integrity of the data set, and the standardized processed environmental parameter data set is output.

[0020] Preferably, the total required air volume is calculated based on the tunnel structure parameters, construction progress and real-time environmental data, the construction condition coefficient is introduced to dynamically correct the air volume distribution ratio, the air volume coordination intelligent model is established combined with the performance attenuation model of the main and auxiliary air fans, and the air volume is dynamically distributed according to the demand, which specifically includes:

[0021] The tunnel structure parameters are obtained, including the cross-sectional area, the current construction section length, the tunnel slope, the construction process type and the real-time environmental data;

[0022] Based on the construction process type, the construction condition coefficient, the drilling and blasting process, the supporting process and the lining process are obtained;

[0023] The total required air volume is calculated by the total required air volume calculation formula;

[0024] The performance attenuation model of the main and auxiliary air fans is established, and the attenuation coefficient is calculated according to the cumulative running time of the air fan;

[0025] The actual output of the air fan is corrected, the training set, the validation set and the test set are divided according to the ratio of 7:2:1, and the air volume coordination intelligent model is trained.

[0026] Preferably, the performance attenuation model of the main and auxiliary air fans is established, and the attenuation coefficient is calculated according to the cumulative running time of the air fan, which specifically includes:

[0027] The initial air volume distribution ratio is set as the basic ratio, with the main air fan bearing 60% to 80% and the auxiliary air fan bearing 20% to 40%;

[0028] The distance correction factor is introduced, the real-time wind pressure loss data is fused, the air volume distribution ratio is optimized by the BP neural network model, the mean square error is taken as the loss function to adjust the model parameters, and the target air volume of the main air fan and the target air volume of the auxiliary air fan are output.

[0029] Preferably, the main air fan variable frequency curve, the wind pressure dynamic adjustment parameter and the auxiliary air fan partition start-stop timing, angle adaptive adjustment scheme are output according to the air volume coordination intelligent model, the control instructions are issued in real time through the edge computing node, and the variable frequency wind pressure adjustment, partition air volume supplement and air valve intelligent opening degree adjustment operation are executed.

[0030] The fan characteristic curve is queried according to the target air volume of the main air fan, the variable frequency curve of the wind pressure-frequency corresponding relationship is generated, and the frequency adjustment range is set to 30 to 50 Hz;

[0031] An auxiliary fan partition evaluation model is constructed, and based on the tunnel wind speed distribution thermodynamic map, the wind volume demand index of each partition is obtained through weighted calculation;

[0032] The optimal air supply angle of each auxiliary fan is obtained through computational fluid dynamics simulation, and the angle adjustment range is 0°-30°, and the adjustment gear is set every 5°;

[0033] After the control instruction is compressed and encoded by the edge computing node, it is transmitted to the fan control cabinet and air valve actuator through industrial Ethernet, and the instruction transmission delay is controlled within 500ms, and the air volume adjustment operation is executed.

[0034] Preferably, the auxiliary fan partition evaluation model is constructed based on the tunnel wind speed distribution thermodynamic map, and the wind volume demand index of each partition is obtained through weighted calculation, which specifically includes:

[0035] The ventilation area is divided into several independent partitions according to the tunnel mileage, and each partition is numbered, with a length of 100-200 meters;

[0036] Three evaluation indexes are set for each partition, including wind speed compliance rate, harmful gas concentration exceeding frequency and temperature deviation value, and the weights are 0.5, 0.3 and 0.2 respectively;

[0037] The standardization values of each index are obtained, including wind speed compliance rate standardization value, harmful gas concentration exceeding frequency standardization value and temperature deviation value standardization value;

[0038] The partition wind volume demand index is calculated through the weighted sum formula;

[0039] According to the index I, the demand level is divided, I≥0.8 is low demand area, 0.5≤I<0.8 is medium demand area, and I<0.5 is high demand area, and the auxiliary fan air volume is preferentially allocated in the high demand area.

[0040] Preferably, the multi-parameter fusion deviation evaluation model is constructed, the spatial distribution deviation of the adjusted environmental parameters and the target parameters is monitored in real time, the fan control parameters are dynamically corrected based on the reinforcement learning algorithm, the adaptive ventilation adjustment closed loop is formed, and the air volume is quickly redistributed under emergency working conditions, which specifically includes:

[0041] The multi-parameter weight coefficients are set, the wind speed weight is 0.4, the harmful gas concentration weight is 0.3, the temperature weight is 0.2, and the wind pressure weight is 0.1. The deviation evaluation model is constructed, and the comprehensive deviation index is output;

[0042] The deviation value of the actual parameters and the target parameters of each monitoring point is calculated, and the comprehensive deviation index is obtained by weighted sum through the above formula;

[0043] Based on the reinforcement learning algorithm, a parameter adjustment strategy library is constructed, and the comprehensive deviation index is taken as the state input, and the fan frequency adjustment amount, the auxiliary fan increase / decrease number and the air valve opening adjustment value are taken as the output.

[0044] After three consecutive adjustments, if the comprehensive deviation index is still > 0.3, the model optimization mechanism is started, the air volume coordination intelligent model parameters are retrained, and a self-adaptive ventilation adjustment closed loop is formed.

[0045] Preferably, the S403 specifically comprises:

[0046] A greedy strategy is adopted to select the adjustment action, and the exploration probability is set to 10%, that is, a historical optimal adjustment strategy is selected from the strategy library with a probability of 90%, and a new adjustment combination is randomly explored with a probability of 10%;

[0047] After each adjustment operation is executed, the change of the comprehensive deviation index before and after the adjustment is recorded, and the reward value is calculated;

[0048] The strategy library parameters are updated through the Q-learning algorithm, the state, action and reward data are stored in the experience replay pool, and samples are extracted from the experience replay pool at a proportion of 5% and supplemented to the model training set;

[0049] After 100 cumulative adjustment operations, the strategy library is retrained based on the newly supplemented sample data, the output precision of the adjustment parameters is optimized, and the dynamic iteration of the strategy library is realized.

[0050] Preferably, after three consecutive adjustments, if the comprehensive deviation index is still > 0.3, the model optimization mechanism is started, the air volume coordination intelligent model parameters are retrained, and a self-adaptive ventilation adjustment closed loop specifically comprises:

[0051] When it is detected that the harmful gas concentration exceeds ≥0.5% and the wind speed drops ≥50%, the emergency mode is triggered;

[0052] The main fan is immediately raised to the rated frequency operation, the wind pressure parameter is increased by 10%-20%, and the air volume of the construction surface is preferentially guaranteed;

[0053] The branch air valves within 50 meters upstream of the over-standard area are closed, all auxiliary fans are directed to the emergency area, and the air supply angle is adjusted to 20°-30°;

[0054] The air volume supplement rate of the emergency area is calculated once every 30 seconds, wherein the supplement rate calculation formula is:

[0055]

[0056] In the formula, V r is the air volume supplement rate, Q s is the emergency supplement air volume, Δt is the time interval, and S is the cross-sectional area of the emergency area.

[0057] Until the concentration of harmful gas is reduced to 0.3% or less and the wind speed is restored to 1.5 m / s or more, the normal ventilation parameters are restored by gradient, and the emergency treatment data are recorded for optimizing the model.

[0058] Compared with the prior art, the beneficial effects of the present application are:

[0059] 1. By arranging monitoring nodes at intervals of 50-100 meters through a distributed sensor array, combining a wavelet transform adaptive filtering algorithm to suppress noise, the data signal-to-noise ratio can be improved by more than 30%, the standardized interval is dynamically adjusted according to the distance of the construction section, and the missing data is supplemented by the interpolation algorithm of adjacent monitoring points to ensure the integrity of the spatiotemporal correlated data set, providing a reliable data basis for air volume calculation, combining tunnel structure parameters and construction working condition coefficients to accurately calculate the total required air volume, introducing a distance correction factor and a fan performance decay model to dynamically optimize the distribution ratio of primary and auxiliary fans, and the air volume utilization rate is increased by 20%-25%.

[0060] 2. The air volume coordination intelligent model is trained through a BP neural network model to realize "on-demand air supply", which reduces energy consumption by 15%-20% compared with the traditional fixed-frequency ventilation mode, divides the low, medium and high demand areas based on the air volume demand index, and the auxiliary fan supplements air volume according to the partition start-stop time sequence and angle orientation, combined with the sectional control strategy of air valves, so that the compliance rate of ventilation parameters in each partition is improved to more than 95%.

[0061] 3. The multi-parameter fusion deviation evaluation model monitors the environmental deviation in real time, dynamically corrects the fan parameters through a reinforcement learning algorithm, and the comprehensive deviation index can be stably controlled below 0.3 after continuous adjustment for 3 times. In the emergency mode, the frequency of the main fan is increased, the auxiliary fan is directionally supplied, the air valve is stopped, and the like, the air volume supplement rate is greatly improved, the processing time of harmful gas concentration exceeding the standard is shortened, the construction safety is significantly improved, the fan output is corrected in real time through the performance decay model to avoid long-term overload operation of the fan, prolong the service life of the equipment, the dynamic adjustment strategy reduces the frequency of starting and stopping the fan and the full-load running time, and the model parameters are optimized combined with the feedback of emergency data to reduce the operation and maintenance cost. BRIEF DESCRIPTION OF DRAWINGS

[0062] Figure 1 The step flow framework diagram of the present application;

[0063] Figure 2 The step flow framework diagram of S2 in the present application;

[0064] Figure 3 The step flow framework diagram of S3 in the present application. DETAILED DESCRIPTION

[0065] The following description is used to disclose the present application to enable a person skilled in the art to implement the present application. The preferred embodiments in the following description are only as examples, and other obvious modifications can be conceived by those skilled in the art.

[0066] Referring to Figure 1 As shown in the figure, a super-long tunnel small and large fan pressurized air supply and coordinated ventilation method comprises:

[0067] S1: Real-time acquisition of tunnel environment parameters, acquisition of wind speed, wind pressure, harmful gas concentration and temperature data of different sections through a distributed sensor array, noise suppression by using an adaptive filtering algorithm, dynamic adjustment of the data standardization interval based on the location of the construction section, generation of a spatiotemporally correlated environmental parameter dataset;

[0068] S2: Fusion of tunnel structure parameters, construction progress and real-time environmental data to calculate the total required air volume, introduction of a construction condition coefficient to dynamically correct the air volume distribution ratio, combination of a main and auxiliary fan performance degradation model to establish an air volume coordination intelligent model, and realization of dynamic distribution of air volume on demand;

[0069] S3: According to the air volume coordination intelligent model output, the main fan variable frequency curve, the wind pressure dynamic adjustment parameter and the auxiliary fan partition start-stop timing, angle adaptive adjustment scheme, through the edge computing node real-time issuing control instruction, executes the variable frequency wind pressure regulation, the partition air volume supplement and the wind valve intelligent opening degree regulation operation;

[0070] S4: Construction of a multi-parameter fusion deviation evaluation model, real-time monitoring of the spatial distribution deviation of the adjusted environmental parameters and the target parameters, dynamic correction of the fan control parameters based on the reinforcement learning algorithm, formation of an adaptive ventilation regulation closed loop, and rapid redistribution of air volume under emergency conditions.

[0071] S1 specifically includes:

[0072] S101: Distribute the distributed sensor array along the tunnel axis at an interval of 50-100 meters, and integrate a wind speed sensor, a wind pressure transmitter, a harmful gas detector and a temperature sensor in each monitoring node; S102: Collect raw data through the sensor at a frequency of 1 per second, synchronously record the data collection time and the corresponding tunnel mileage position information, and call the historical environmental parameter data from the tunnel management system; S103: Use correlation analysis method to evaluate the correlation degree of each environmental parameter with the ventilation effect, remove redundant monitoring data, and retain key characteristic parameters; S104: Use the adaptive filtering algorithm based on wavelet transform to suppress noise of the raw data, dynamically adjust the standardization interval according to the distance L of the construction section from the tunnel entrance, and extend the upper limit of the wind speed interval to 1.8 m / s and the upper limit of the wind pressure interval to 2200 Pa when L>1000 meters, to construct a spatiotemporally correlated environmental parameter dataset;

[0073] Based on the distance of the construction section, the data standardization interval is dynamically adjusted to solve the standardization deviation problem caused by the distribution difference of environmental parameters in long-distance tunnels, and the data signal-to-noise ratio is improved by combining wavelet transform filtering.

[0074] S104 specifically includes:

[0075] The collected wind speed, wind pressure, harmful gas concentration and temperature data are mapped to the [0, 1] interval using the normalization method, wherein the wind speed normalization formula is:

[0076]

[0077] In the formula, v norm is the normalized wind speed value, v is the original wind speed collection value, v min is the minimum wind speed v max is the maximum wind speed;

[0078] Based on the position information of the construction section, spatial labels are added to the data, and a space-time matrix is formed by sorting the time sequence;

[0079] The missing data is supplemented by using the interpolation algorithm of adjacent monitoring points to ensure the integrity of the data set, and the environmental parameter data set after normalization processing is output, wherein the interpolation algorithm formula is:

[0080]

[0081] In the formula, X m is the parameter interpolation result of the missing position, X i is the parameter value of the left adjacent monitoring point, X j is the parameter value of the right adjacent monitoring point, L i is the left monitoring point mileage position, L j is the right monitoring point mileage position, and L m is the mileage of the missing data position;

[0082] The space-time matrix is constructed by the spatial label and the time sequence, the missing sensor data is solved by the interpolation algorithm of adjacent monitoring points, and the space-time correlation and integrity of the data set are ensured.

[0083] Referring to Figure 2 S2 specifically includes:

[0084] S201: Obtain tunnel structure parameters, including cross-sectional area, current construction section length, tunnel slope, construction process type and real-time environmental data; S202: Obtain construction condition coefficients, drilling and blasting process, support process and lining process based on the construction process type; S203: Calculate the total required air volume by the total required air volume calculation formula, wherein the total required air volume calculation formula is: Q=K×K CIn the formula, Q is the total required air volume, K is the safety factor, K C is the construction condition coefficient, V is the minimum wind speed standard, and S is the tunnel cross-sectional area; S204: Establish a main and auxiliary fan performance attenuation model, and calculate the attenuation coefficient according to the cumulative running time of the fan, wherein the attenuation coefficient calculation formula is:

[0085] λ = 1-0.0001 × t

[0086] In the formula, λ is the performance attenuation coefficient, and the value range is 0.8-1.0, and t is the cumulative running time of the fan;

[0087] Correct the actual output of the fan, and divide the training set, the verification set and the test set according to the proportion of 7:2:1 to train the air volume coordination intelligent model;

[0088] Introduce the construction process differentiation condition coefficient, and dynamically correct the output combined with the fan performance attenuation model, to solve the problem that the traditional air volume calculation does not consider equipment aging and process difference.

[0089] S204 specifically includes:

[0090] The initial air volume distribution ratio is set to a basic ratio of 60% to 80% for the main fan and 20% to 40% for the auxiliary fan;

[0091] Introduce a distance correction factor, and the distance correction factor calculation formula is:

[0092]

[0093] In the formula, f(L) is the distance correction factor, L is the distance of the construction section from the tunnel entrance, and when L>2000m, the air volume proportion of the auxiliary fan is increased by 5% to 10%;

[0094] Fuse real-time air pressure loss data, optimize the air volume distribution ratio through a BP neural network model, use mean square error as a loss function to adjust model parameters, and output the target air volume of the main fan and the target air volume of the auxiliary fan to meet:

[0095]

[0096] In the formula, Q1 is the target air volume of the main fan, Q2 is the target air volume of the auxiliary fan, Q is the total required air volume, and λ is the performance attenuation coefficient;

[0097] Propose a distance correction factor to dynamically adjust the distribution ratio of the main and auxiliary fans, and combine BP neural network optimization to solve the problem of coordinated control of air volume attenuation and local air pressure loss in long-distance tunnels.

[0098] Refer to Figure 3As shown, S3 specifically includes: S301: querying the fan characteristic curve according to the main fan target air volume Q1, generating a variable frequency curve of the air pressure-frequency correspondence, and setting the frequency adjustment range to 30-50 Hz; S302: constructing an auxiliary fan partition evaluation model, based on the tunnel air speed distribution thermodynamic map, and obtaining the air volume demand index of each partition through weighted calculation; S303: obtaining the optimal air supply angle of each auxiliary fan through computational fluid dynamics simulation, the angle adjustment range is 0°-30°, and the adjustment gear is set at an interval of 5°; S304: the edge computing node compresses and encodes the control instruction, and then transmits it to the fan control cabinet and air valve actuator through industrial Ethernet, the instruction transmission delay is controlled within 500 ms, and the air volume adjustment operation is performed;

[0099] The fan air supply angle is optimized by fluid dynamics simulation, the control instruction is issued with low delay by edge computing, and the local air volume regulation precision and response speed are improved.

[0100] S302 specifically includes: dividing the ventilation area into several independent partitions according to the tunnel mileage, the length of each partition is 100-200 meters, and the partitions are numbered; setting three evaluation indexes for each partition, including the air speed compliance rate, the harmful gas concentration exceeding frequency and the temperature deviation value, and the weights are 0.5, 0.3 and 0.2 respectively;

[0101] The standardization values of each index are obtained, including the air speed compliance rate standardization value, the harmful gas concentration exceeding frequency standardization value and the temperature deviation value standardization value;

[0102] The partition air volume demand index is calculated by a weighted sum formula:

[0103] I=0.5×a+0.3×b+0.2×c

[0104] In the formula, I is the partition air volume demand index, a is the air speed compliance rate standardization value, b is the harmful gas concentration control standardization value, and c is the temperature control standardization value;

[0105] According to the index I, the demand level is divided, I≥0.8 is a low demand area, 0.5≤I<0.8 is a medium demand area, and I<0.5 is a high demand area, and the auxiliary fan air volume is preferentially allocated to the high demand area;

[0106] A multi-index weighted partition air volume demand evaluation model is established to realize differentiated hierarchical regulation and control of the tunnel ventilation area, and solve the problem of energy waste caused by traditional uniform air supply.

[0107] S4 specifically includes:

[0108] S401: Set the multi-parameter weight coefficient, the wind speed weight is 0.4, the harmful gas concentration weight is 0.3, the temperature weight is 0.2, and the wind pressure weight is 0.1, construct a deviation evaluation model, and output a comprehensive deviation index, wherein the deviation evaluation model is constructed according to the following formula:

[0109] DI = 0.4 * |v-v0| + 0.3 * |c-c0| + 0.2 * |t-t0| + 0.1 * |p-p0|

[0110] In the formula, DI is the comprehensive deviation index, v is the actual wind speed, v0 is the target wind speed, c is the actual harmful gas concentration, c0 is the target concentration, t is the actual temperature, t0 is the target temperature, p is the actual wind pressure, and p0 is the target wind pressure;

[0111] S402: Calculate the deviation value of each monitoring point actual parameter and target parameter, and obtain the comprehensive deviation index DI by weighted summation according to the above formula;

[0112] S403: Based on the reinforcement learning algorithm, a parameter adjustment strategy library is constructed, the comprehensive deviation index DI is taken as the state input, and the wind fan frequency adjustment amount, the auxiliary wind fan increase / decrease quantity and the wind valve opening degree adjustment value are outputted;

[0113] S404: After continuous adjustment for three times, if DI is still greater than 0.3, the model optimization mechanism is started, the wind volume coordination intelligent model parameters are retrained, and a self-adaptive ventilation adjustment closed loop is formed;

[0114] A multi-parameter fusion deviation evaluation model is constructed, the reinforcement learning is combined to realize dynamic correction of the fan control parameters, and a self-adaptive closed loop of "monitoring-evaluation-adjustment-optimization" is formed.

[0115] S403 specifically includes:

[0116] The greedy strategy is adopted to select the adjustment action, the exploration probability is set to 10%, that is, the historical optimal adjustment strategy is selected from the strategy library with a probability of 90%, and a new adjustment combination is randomly explored with a probability of 10%;

[0117] After each adjustment operation is completed, the change of the comprehensive deviation index before and after adjustment is recorded, and the reward value is calculated, wherein the reward value is calculated according to the following formula:

[0118] R = DI b -DI a

[0119] In the formula, R is the reward value, DI b is the comprehensive deviation index before adjustment, and DI a is the comprehensive deviation index after adjustment;

[0120] The strategy library parameters are updated through the Q-learning algorithm, and the state, action and reward data are stored to the experience replay pool, and samples are extracted from the experience replay pool at a proportion of 5% to supplement the model training set;

[0121] After each cumulative adjustment operation of 100 times, the strategy library is retrained based on the newly supplemented sample data, the output precision of the adjustment parameters is optimized, and dynamic iteration of the strategy library is realized;

[0122] The greedy strategy of reinforcement learning and the Q-learning algorithm are introduced into ventilation adjustment, and the self-learning ability of the fan control strategy is improved through experience replay and strategy iteration optimization.

[0123] S404 specifically comprises:

[0124] When it is detected that the harmful gas concentration exceeds 0.5% and the wind speed drops by more than 50%, the emergency mode is triggered; the main fan is immediately raised to the rated frequency operation, the wind pressure parameter is increased by 10%-20%, the air volume of the construction surface is preferentially guaranteed; the branch air valves within 50 meters upstream of the over-standard area are closed, all auxiliary fans are directed to the emergency area, and the air supply angle is adjusted to 20°-30°; the air volume supplement rate of the emergency area is calculated every 30 seconds, wherein the supplement rate calculation formula is:

[0125]

[0126] In the formula, V r is the air volume supplement rate, Q s is the emergency air volume supplement, and Δt is the time interval, and S is the cross-sectional area of the emergency area;

[0127] Until the harmful gas concentration is reduced to below 0.3% and the wind speed is restored to above 1.5 m / s, the normal ventilation parameters are restored according to the gradient, and the emergency treatment data are recorded for optimization of the model;

[0128] A multi-dimensional emergency response mechanism is designed, the main and auxiliary fans are linked, the air valve is directionally controlled, and the supplement rate is monitored in real time, so that the air volume is quickly redistributed and safely recovered under emergency conditions.

[0129] In summary, the advantages of the present application are:

[0130] The super-long tunnel ventilation collaborative system of "data collection-intelligent distribution-precise regulation-closed loop optimization" is created, the reinforcement learning algorithm is deeply integrated with the collaborative control of the main and auxiliary fans, the static regulation limitations of the traditional ventilation mode are broken through, the full-dynamic self-adaptive closed loop from environmental perception to parameter adjustment is realized, the dynamic standardization interval adjustment method based on the distance of the construction section is proposed, the wavelet transform filtering and the interpolation algorithm of adjacent monitoring points are combined, the problems of noise interference and missing of environmental data of long-distance tunnels are solved, and the spatiotemporal correlation and integrity of the data set are ensured;

[0131] The construction condition coefficient and the fan performance attenuation model are introduced, the total air volume calculation result is dynamically corrected, the equipment aging degree and process difference are first taken into account in the air volume decision, the accuracy and equipment adaptability of the air volume calculation are improved, the distance correction factor is used to dynamically adjust the distribution ratio of the main and auxiliary fans, the BP neural network optimization algorithm is combined to solve the coordinated control problem of long distance tunnel air volume attenuation and local wind pressure loss, and realize the dynamic ratio optimization of 60% to 80% main fan and 20% to 40% auxiliary fan;

[0132] A multi-index weighted partition air volume demand evaluation model is constructed, the auxiliary fan air supply angle is optimized through fluid mechanics simulation, the low delay instruction transmission of the edge computing node is combined, the differentiated hierarchical regulation of the tunnel ventilation area is realized, the greedy strategy and Q-learning algorithm are introduced into the ventilation regulation, the comprehensive deviation index is taken as the state input, the experience replay pool and the strategy iteration mechanism are used, so that the fan control strategy has self-learning ability, and after continuous regulation, the DI can be stably controlled below 0.3.

[0133] A double emergency triggering mechanism of harmful gas exceeding standard and wind speed sudden drop is established, through the linkage operation of main fan frequency raising, auxiliary fan directional air supply, wind valve regional shutdown and the like, in combination with wind volume supplement rate real-time monitoring, the wind volume rapid redistribution and safety recovery under emergency condition are realized, through dynamic wind volume distribution, partition on-demand air supply and equipment attenuation correction, the energy consumption is reduced by 15% to 20% compared with the traditional fixed frequency ventilation mode, at the same time, the harmful gas exceeding standard processing time is shortened to within 5 minutes, the ventilation energy efficiency and construction safety are doubled.

[0134] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above examples, and the above examples and descriptions in the specification are only the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection claimed by the present application is defined by the appended claims and their equivalents.

Claims

1. A method for coordinated ventilation of a super-long tunnel by a small and large fan, characterized in that, Comprise: S1: real-time acquisition of tunnel environment parameters, through a distributed sensor array to obtain different section wind speed, wind pressure, harmful gas concentration and temperature data, using adaptive filtering algorithm for noise suppression, based on the position of the construction section dynamic adjustment data standardization interval, generation of space-time associated environmental parameter data set; S2: fusion tunnel structure parameters, construction progress and real-time environmental data to calculate the total required air volume, introduce construction condition coefficient dynamic correction air distribution ratio, combined with the performance attenuation model of main and auxiliary fan to establish air coordination intelligent model, realize air dynamic distribution according to demand; S3: according to the output of air coordination intelligent model main fan frequency conversion curve, wind pressure dynamic adjustment parameters and auxiliary fan partition start-stop timing, angle adaptive adjustment scheme, through the edge computing node real-time issue control instruction, execute variable frequency pressure regulation, partition air supplement and intelligent opening degree regulation operation of air valve; S4: construction of multi-parameter fusion deviation evaluation model, real-time monitoring of the spatial distribution deviation of adjusted environmental parameters and target parameters, based on reinforcement learning algorithm to dynamically correct fan control parameters, form adaptive ventilation regulation closed loop, carry out air rapid redistribution under emergency condition.

2. The method according to claim 1, wherein, The S1 specifically comprises: S101: along the tunnel axis, 50-100 meters interval distribution of distributed sensor array, each monitoring node integrates wind speed sensor, wind pressure transmitter, harmful gas detector and temperature sensor; S102: through the sensor to collect the original data with a frequency of 1 times per second, synchronous recording data acquisition time and corresponding tunnel mileage position information, call historical environmental parameter data from tunnel management system; S103: correlation analysis method is used to evaluate the correlation degree of each environmental parameter and ventilation effect, remove redundant monitoring data, and retain key feature parameters; S104: using adaptive filtering algorithm based on wavelet transform to suppress noise of original data, dynamically adjusting the standardization interval according to the distance L of construction section from tunnel entrance, L≤1000 meters, using basic interval, L>1000 meters, expanding the upper limit of wind speed interval to 1.8 m / s, and the upper limit of wind pressure interval to 2200 Pa, to construct the space-time associated environmental parameter data set.

3. The method according to claim 2, wherein, The S104 specifically comprises: The collected wind speed, wind pressure, harmful gas concentration and temperature data are mapped to the [0,1] interval by normalization method, wherein the wind speed normalization formula is: wherein v norm is the normalized wind speed value, v is the original wind speed value, v min is the minimum wind speed value max is the maximum wind speed value; Based on the position information of construction section, add spatial label to the data, and form space-time matrix according to time sequence; The missing data is supplemented by using the interpolation algorithm of adjacent monitoring points to ensure the integrity of the data set, and the normalized environmental parameter data set is output, wherein the interpolation algorithm formula is: where X m is the parameter interpolation result at the missing position, X i is the parameter value of the left adjacent monitoring point, X j is the parameter value of the right adjacent monitoring point, L i is the left monitoring point mileage position, L j is the right monitoring point mileage position, L m is the mileage at the missing data position.

4. The method according to claim 1, wherein, The S2 specifically comprises: S201: obtain tunnel structure parameters, including cross-sectional area, current construction section length, tunnel slope, construction process type and real-time environmental data; S202: based on construction process type to obtain construction condition coefficient, drilling and blasting process, support process and lining process; S203: calculate the total required air volume by total required air volume calculation formula, wherein the total required air volume calculation formula Q = K x K C V x S In the formula, Q is the total air requirement, K is the safety factor, K C is the construction condition coefficient, V is the minimum wind speed standard, and S is the tunnel cross-sectional area. S204: Establish a main and auxiliary fan performance attenuation model, and calculate an attenuation coefficient according to a cumulative running time of the fan, wherein an attenuation coefficient calculation formula is: λ = 1 - 0.0001 x t In the formula, λ is a performance attenuation coefficient, and a value range is 0.8-1.0, and t is a cumulative running time of the fan; Correct the actual output of the fan, and divide the training set, the verification set and the test set according to a 7:2:1 ratio to train the air volume coordination intelligent model.

5. The method according to claim 4, wherein, The S204 specifically includes: An initial air volume allocation ratio is set as a basic ratio according to 60% to 80% of the main fan and 20% to 40% of the auxiliary fan; A distance correction factor is introduced, and a distance correction factor calculation formula is: In the formula, f(L) is a distance correction factor, L is a distance of a construction section from a tunnel entrance, and the auxiliary fan air volume ratio is increased by 5% to 10% when L>2000 meters; Fusion real-time wind pressure loss data, the air volume allocation ratio is optimized through a BP neural network model, a mean square error is taken as a loss function to adjust model parameters, and main fan target air volume and auxiliary fan target air volume are output to meet: In the formula, Q1 is main fan target air volume, Q2 is auxiliary fan target air volume, Q is total required air volume, and λ is a performance attenuation coefficient.

6. The method according to claim 1, wherein, The S3 specifically includes: S301: According to the main fan target air volume Q1, the fan characteristic curve is queried, the variable frequency curve of the wind pressure-frequency corresponding relationship is generated, and the frequency regulation range is set to 30-50 Hz; S302: A partition evaluation model of the auxiliary fan is constructed, the tunnel air velocity distribution thermal diagram is based, and the air volume demand index of each partition is obtained through weighted calculation; S303: The optimal air supply angle of each auxiliary fan is obtained through computational fluid dynamics simulation, the angle regulation range is 0°-30°, and the regulation gear is set every 5°; S304: The edge computing node compresses and encodes the control instruction, and then sends it to the fan control cabinet and the air valve actuator through the industrial Ethernet, the instruction transmission delay is controlled within 500 ms, and the air volume regulation operation is performed.

7. The method according to claim 6, wherein, The S302 specifically includes: The ventilation area is divided into a plurality of independent partitions according to the tunnel mileage, and each partition has a length of 100-200 meters and is numbered; Three evaluation indexes are set for each partition, including a wind speed compliance rate, a harmful gas concentration exceeding frequency and a temperature deviation value, and the weights are 0.5, 0.3 and 0.2 respectively; Standardized values of the indexes are obtained, including a wind speed compliance rate standardized value, a harmful gas concentration exceeding frequency standardized value and a temperature deviation value standardized value; The partition air volume demand index is calculated through a weighted summation formula: I = 0.5 x a + 0.3 x b + 0.2 x c In the formula, I is a partition air volume demand index, a is a wind speed compliance rate standardized value, b is a harmful gas concentration control standardized value, and c is a temperature control standardized value; According to the index I, the demand level is divided, I≥0.8 is a low demand area, 0.5≤I<0.8 is a medium demand area, and I<0.5 is a high demand area, and the auxiliary fan air volume is preferentially allocated to the high demand area.

8. The method according to claim 1, wherein, The S4 specifically includes: S401: Set the multi-parameter weight coefficient, wind speed weight 0.4, harmful gas concentration weight 0.3, temperature weight 0.2, and wind pressure weight 0.1, construct a deviation evaluation model, and output a comprehensive deviation index, wherein the calculation formula of the deviation evaluation model is: DI=0.4×|v-v0|+0.3×|c-c0|+0.2×|t-t0|+0.1×|p-p0| Wherein, DI is the comprehensive deviation index, v is the actual wind speed, v0 is the target wind speed, c is the actual harmful gas concentration, c0 is the target concentration, t is the actual temperature, t0 is the target temperature, p is the actual wind pressure, and p0 is the target wind pressure; S402: Calculate the deviation value of each monitoring point actual parameter and target parameter, and obtain the comprehensive deviation index DI by weighted summation according to the above formula; S403: Based on the reinforcement learning algorithm, a parameter adjustment strategy library is constructed, the comprehensive deviation index DI is taken as the state input, and the wind fan frequency adjustment amount, the auxiliary fan increase / decrease quantity and the wind valve opening degree adjustment value are outputted; S404: After continuous adjustment for three times, if DI is still greater than 0.3, the model optimization mechanism is started, the wind volume coordination intelligent model parameters are retrained, and a self-adaptive ventilation adjustment closed loop is formed.

9. The method according to claim 8, wherein, The S403 specifically includes: A greedy strategy is used to select the adjustment action, and the exploration probability is set to 10%, that is, a historical optimal adjustment strategy is selected from the strategy library with a probability of 90%, and a new adjustment combination is randomly explored with a probability of 10%; After each adjustment operation is completed, the change of the comprehensive deviation index before and after adjustment is recorded, and the reward value is calculated, wherein the reward value calculation formula is: R = DI b - DI a where R is a reward value, DI b is the integrated deviation index before adjustment, DI a is the integrated deviation index after adjustment; The strategy library parameters are updated through the Q-learning algorithm, and the state, action and reward data are stored in the experience replay pool, and samples are extracted from the experience replay pool at a proportion of 5% and supplemented to the model training set; After each cumulative adjustment operation is completed, the strategy library is retrained based on the newly supplemented sample data, the output precision of the adjustment parameters is optimized, and the dynamic iteration of the strategy library is realized.

10. The method of claim 8, wherein the method is a method of super-long tunnel small and large fan pressure ventilation and air supply coordination ventilation, characterized in that, The S404 specifically includes: When it is detected that the harmful gas concentration exceeds ≥0.5% and the wind speed drops by ≥50%, the emergency mode is triggered; The main fan is immediately raised to the rated frequency operation, the wind pressure parameter is increased by 10%-20%, and the wind volume on the construction surface is preferentially guaranteed; The branch air valves within 50 meters upstream of the over-standard area are closed, all auxiliary fans are directed to the emergency area, and the air supply angle is adjusted to 20°-30°; The wind volume supplement rate of the emergency area is calculated every 30 seconds, wherein the supplement rate calculation formula is: In the formula, V r is the air volume replenishment rate, Q s is the emergency air volume replenishment, Δt is the time interval, and S is the cross-sectional area of the emergency area. Until the harmful gas concentration is reduced to below 0.3% and the wind speed is restored to above 1.5m / s, the normal ventilation parameters are restored according to the gradient, and the emergency treatment data is recorded for model optimization.

Images