Intelligent layered control method for low-temperature sludge drying machine

Abstract

The invention discloses an intelligent hierarchical control method for a sludge low-temperature drying machine, which is used for associating operation data of the same batch of sludge through timestamps, solving the problem of data mismatch caused by feeding and discharging time delay in continuous drying, and adjusting the transition operation speed of each layer of chain network through a transition coefficient. And finally, the optimization model controls the optimal operation speed of each layer of chain network by taking whether the water content of the discharged sludge reaches the standard or not and the lowest unit energy consumption as constraint conditions, so that the water content of the sludge at a discharge port and the unit energy consumption are optimal, and the drying effect and the equipment operation economy are both considered.

Description

Technical Field

[0001] This invention relates to the field of sludge drying equipment technology, and in particular to an intelligent layered control method for a low-temperature sludge dryer. Background Technology

[0002] The low-temperature sludge dryer is a sludge treatment device that uses a heat pump as its core heat source. It employs a multi-layered chain mesh and multiple air vents within the silo. Dehumidifying air enters from the bottom of the silo and then passes through each layer of chain mesh and sludge layer from bottom to top to heat and dry the sludge. Heating air enters laterally from the side of the silo through the middle of the partition, mixed with the dehumidifying air coming from below, and continues upward through the remaining chain mesh and sludge layer. Finally, it passes through a filter and enters the heat source center. The air is then distributed: one portion is dehumidified by the evaporator and then heated by the condenser before being sent back to the bottom of the silo by the dehumidifying fan; the other portion is directly heated by the condenser and then sent into the silo through the middle layer of chain mesh. This cycle repeats continuously, drying the sludge at a relatively low temperature of 40-80℃, ultimately achieving the desired moisture content at the equipment outlet.

[0003] The moisture content of the influent and effluent sludge is a core indicator for low-temperature sludge dryers. Traditional low-temperature sludge dryers rely on manual sampling to test the moisture content, which has a significant lag, leading to untimely adjustments to drying parameters (such as residence time) and frequent fluctuations in moisture content exceeding the standard. With continuous technological advancements, modern low-temperature sludge dryers typically have a microwave / infrared online moisture content monitor installed at both the inlet and outlet of the drying chamber. This monitor calculates the change in sludge moisture content between the inlet and outlet to control the speed of the conveyor belt, ensuring the sludge moisture content at the outlet meets the standard. To further refine the control of the conveyor belt speed, many low-temperature sludge dryers install online moisture content monitors above each conveyor belt layer, such as the intelligent control system for adjusting the effluent moisture content of sludge drying equipment disclosed in Chinese Utility Model Patent CN208964777U.

[0004] However, all the above solutions regulate sludge moisture content by controlling the speed of the conveyor belt. If the equipment can process a unit weight of sludge earlier than expected, it will obviously help reduce energy consumption. But more often than not, the sludge moisture content fails to reach the expected level, causing the sludge to remain in the conveyor belt for too long, which inevitably increases the equipment's energy consumption. In addition, the price of online moisture content monitoring instruments ranges from several thousand to tens of thousands of yuan. For a large-scale sludge low-temperature dryer, it would be detrimental to cost control to equip each layer of the conveyor belt with at least one online moisture content monitoring instrument. Summary of the Invention

[0005] To address the aforementioned problems, this invention proposes an intelligent stratified control method for a low-temperature sludge dryer, which mainly solves the problems in the background technology.

[0006] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0007] A method for intelligent stratified control of a sludge low-temperature dryer includes the following steps:

[0008] The system's operational data is acquired. The operational data is linked to the full-process parameters of the same batch of sludge using timestamps, including the moisture content of the feed sludge, the real-time temperature of each layer of the chain mesh, the real-time humidity of each layer of the chain mesh, the real-time operating speed of each layer of the chain mesh, the moisture content of the discharge sludge, and the unit energy consumption.

[0009] Using the moisture content of the discharged sludge from the same batch of sludge as an index, a first correction coefficient is generated based on the corresponding moisture content of the fed sludge, and a second correction coefficient is generated based on the real-time temperature of each layer of the chain network. The product of the first correction coefficient and the second correction coefficient is used as a transition coefficient, and the operating speed of each layer of the chain network is recalculated as the transition operating speed of each layer of the chain network, thereby dynamically controlling the motor speed of each layer of the chain network.

[0010] An optimization model is established, and the operating data and target moisture content are input into the optimization model for training. The optimal operating speed of each layer of the chain network is output, and the motor speed of each layer of the chain network is controlled in conjunction with the target moisture content and the unit energy consumption is minimized, with the constraints being the moisture content of the discharged sludge approaching the target moisture content and the minimum unit energy consumption.

[0011] In some implementations, the calculation process for the transition speed includes:

[0012] The initial velocity of each layer of the chain is preset, wherein the initial velocity of each layer of the chain decreases sequentially from top to bottom;

[0013] Multiple reference moisture content ranges are set, and the moisture content of the feed sludge is matched with the corresponding reference moisture content ranges. Then, a first correction coefficient is generated according to preset conditions.

[0014] Multiple reference temperature ranges are set, the average real-time temperature of each layer of the chain network is calculated, the average temperature is matched with the corresponding reference temperature range, and a second correction coefficient is generated according to preset conditions.

[0015] The product of the first correction coefficient and the second correction coefficient is used as a transition coefficient to recalculate the operating speed of each layer of the chain network, which serves as the transition operating speed of each layer of the chain network, and dynamically controls the motor speed of each layer of the chain network.

[0016] In some implementations, after the optimal operating speed takes effect, the moisture content of the discharged sludge is detected. If the moisture content of the discharged sludge is greater than the target moisture content, a blower adjustment procedure is initiated. The blower adjustment procedure includes:

[0017] Determine whether the humidity of the bottom layer of the dehumidification network is greater than the first preset relative humidity and whether the temperature of the bottom layer of the dehumidification network is less than the first preset temperature. If so, increase the air volume of the dehumidification fan according to the preset ratio.

[0018] Determine whether the humidity of the middle layer of the chain network is greater than the second preset relative humidity and whether the temperature of the middle layer of the chain network is less than the second preset temperature. If so, increase the output temperature of the heating fan by the preset range.

[0019] Once the air intake volume of the dehumidifying fan and / or the air intake temperature of the heating fan are controlled and meet the standards, the current real-time operating data of the system is re-detected and input into the optimization model to obtain the new optimal operating speed.

[0020] In some implementations, the optimization model is a random forest regression model, and the loss function of the random forest regression model is defined as:

[0021] ;

[0022] In the formula, W1 is the moisture content of the discharged sludge, W_set is the target moisture content, λ is the energy consumption weighting coefficient, EER is the unit energy consumption, and the value of λ ranges from 0.1 to 0.3. When it is necessary to prioritize ensuring that the moisture content of the discharged sludge meets the standard, λ is set to 0.1-0.2, and when it is necessary to prioritize reducing energy consumption, λ is set to 0.2-0.3.

[0023] In some implementations, the training process of the optimization model further includes:

[0024] Every 500 new sets of running data are accumulated, the optimized model is automatically updated through incremental training. The incremental training process only performs feature learning on the new data and does not cover the historical training parameters.

[0025] Furthermore, during incremental training, high-quality samples in the operational data that meet preset conditions are assigned a sample weight of 1.5 times. These high-quality samples are those in which the moisture content of the discharged sludge falls within the preset target moisture content fluctuation range and the unit energy consumption is less than the historical average energy consumption.

[0026] In some implementations, abnormal sample filtering is also included: the running data is preprocessed before the optimization model is trained to remove abnormal values ​​such as temperature and humidity jumps, abnormal values ​​of chain network speed, and abnormal values ​​of sludge batch identification caused by sensor failure. When the proportion of abnormal samples exceeds a preset threshold, the sensor self-test program is triggered.

[0027] In some implementations, the method also includes monitoring operational data: during the training of the optimization model, the influence weight of each operational data on the optimal operating speed is calculated in real time. If the importance percentage of a certain layer feature decreases by more than 10% for three consecutive training sessions, an alarm is triggered to manually check for air duct blockage or sensor drift.

[0028] In some implementations, it also includes:

[0029] The fluctuation characteristics of the moisture content of the feed sludge are detected, and the sludge type is identified based on the fluctuation characteristics;

[0030] Based on the sludge type, a matching subset of data is retrieved from historical operating data to train the optimization model. During the training process, the corresponding energy consumption weight coefficient is set according to the sludge type.

[0031] The beneficial effects of this invention are as follows: by associating the operating data of the same batch of sludge with timestamps, the data mismatch problem caused by the time delay of feeding and discharging in continuous drying is solved; by adjusting the transition operating speed of each layer of the chain network through the transition coefficient, dynamic chain network speed adaptation before optimization model training is achieved; finally, the optimization model controls the optimal operating speed of each layer of the chain network with the standard of sludge moisture content and the minimum unit energy consumption as constraints, so that the sludge moisture content and unit energy consumption at the discharge port are optimal, taking into account both drying effect and equipment operation economy. Attached Figure Description

[0032] Figure 1 This is a schematic flowchart of the intelligent stratified control method for a low-temperature sludge dryer disclosed in an embodiment of the present invention. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of this invention clearer and more explicit, the content of this invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to this invention are shown in the accompanying drawings, not all of them.

[0034] Example 1

[0035] This embodiment proposes an intelligent layered control method for a sludge low-temperature dryer, which is applied to a sludge low-temperature dryer where dehumidifying air enters from the bottom of the silo and heating air enters from the middle of the silo. The following description takes an example with upper, middle and lower three-layer chain mesh inside the silo, but it does not limit this solution to dryers with three-layer chain mesh. In this solution, the drying machine includes only two moisture meters: the first moisture meter is located at the feed inlet of the silo and is used to detect the moisture content W0 (in %) of the feed sludge; the second moisture meter is located at the discharge outlet of the silo and is used to detect the moisture content W1 (in %) of the discharge sludge; and three temperature and humidity sensor groups are included: an upper layer temperature and humidity sensor group, a middle layer temperature and humidity sensor group, and a lower layer temperature and humidity sensor group. The upper layer temperature and humidity sensor group is located in the area corresponding to the upper layer chain mesh and is used to detect the upper layer ambient humidity H1 (in %RH) and the upper layer ambient temperature T1 (in °C). The middle layer temperature and humidity sensor group is located in the area corresponding to the middle layer chain mesh and is used to detect the middle layer ambient humidity H2 (in %RH) and the middle layer ambient temperature T2 (in °C). The lower layer temperature and humidity sensor group is located in the area corresponding to the lower layer chain mesh and is used to detect the lower layer ambient humidity H3 (in %RH) and the lower layer ambient temperature T3 (in °C).

[0036] This solution uses timestamps to link the operational data of the same batch of sludge, solving the data mismatch problem caused by the time delay between feeding and discharging in continuous drying. It also adjusts the transition speed of each layer of the chain network through a transition coefficient, achieving dynamic chain network speed adaptation before the optimization model is trained. Finally, the optimization model uses whether the moisture content of the discharged sludge meets the standard and the minimum unit energy consumption as constraints to control the optimal operating speed of each layer of the chain network, so that the sludge moisture content and unit energy consumption at the discharge port are optimal, balancing the drying effect and the economic efficiency of equipment operation.

[0037] like Figure 1 As shown, it includes the following steps:

[0038] Step 1: Acquire system operational data. This data is timestamped and linked to full-process parameters of the same batch of sludge, including the feed sludge moisture content, real-time temperature of each layer of the conveyor belt, real-time humidity of each layer, real-time operating speed of each layer, discharge sludge moisture content, and unit energy consumption. In Step 1, the operational data is timestamped to ensure temporal matching. For example, in a continuously operating dryer, the "same batch of sludge" refers to the same portion of sludge entering the feed inlet, flowing sequentially through the upper, middle, and lower layers of the conveyor belt, and finally discharging from the discharge outlet within the same time period (not a discrete batch, but a "time slice" of sludge in a continuous flow). The time period can be manually defined. For example, at a certain time t0, 100 kg of sludge enters the upper chain mesh from the feed inlet. After the residence time t1 in the upper chain mesh, the residence time t2 in the middle chain mesh, and the residence time t3 in the lower chain mesh, it is discharged from the discharge outlet at time t0+t1+t2+t3. This 100 kg of sludge constitutes "the same batch".

[0039] Step 2: Using the moisture content of the discharged sludge from the same batch of sludge as an index, a first correction coefficient is generated based on the corresponding moisture content of the fed sludge. A second correction coefficient is generated based on the real-time temperature of each layer of the chain mesh. The product of the first and second correction coefficients is used as a transition coefficient. The operating speed of each layer of the chain mesh is recalculated and used as the transition operating speed of each layer of the chain mesh to dynamically control the motor speed of each layer of the chain mesh.

[0040] Step 2 primarily considers the drying division of labor within the three-layer chain network. For the upper chain network, as the sludge first enters the dryer, the main focus is on removing free water (as moisture evaporates easily), requiring rapid transport to prevent accumulation (fastest speed). Simultaneously, the upper environment is used for initial dehumidification. For the middle chain network, the sludge enters a transitional stage, with reduced free water and increased bound water content. A longer residence time is needed (medium speed) to ensure sufficient contact with the mixed airflow from the middle heating air and the bottom dehumidifying air, enhancing moisture migration. For the lower chain network, the sludge enters the deep drying stage, requiring the removal of stubborn bound water. The longest residence time is needed (slowest speed) to ensure final drying is completed in the high-temperature, low-humidity environment at the bottom.

[0041] In one example, the calculation process for the transition speed includes:

[0042] Step 201: Preset the initial speed of each layer of the chain network, wherein the initial speed of each layer of the chain network decreases sequentially from top to bottom; specifically, assume that the initial speed is V1:V2:V3=1.5:1.2:1.0 (base speed V1=1.5m / min, V2=1.2m / min, V3=1.0m / min), and satisfy the following: the speed of the upper layer chain network V1> the speed of the middle layer chain network V2> the speed of the lower layer chain network V3.

[0043] Step 202: Set multiple benchmark moisture content ranges. After matching the moisture content of the feed sludge with the corresponding benchmark moisture content ranges, generate a first correction coefficient according to preset conditions. Specifically, if the moisture content of the feed sludge W0 > 80% (benchmark moisture content), the first correction coefficient K1 is set to 0.8-0.9 (the first correction coefficient K1 decreases by 0.05 for every 5 percentage point increase in the moisture content of the feed sludge W0); if W0 < 70%, the first correction coefficient K1 is set to 1.1-1.2 (the first correction coefficient K1 increases by 0.05 for every 5 percentage point decrease in the moisture content of the feed sludge W0); if 70% ≤ W0 ≤ 80%, the first correction coefficient K1 is set to 1.0. The first correction coefficient K1 is generated by matching the moisture content of the feed sludge W0 with the benchmark ranges. The core principle is to adjust the total retention time according to the total amount of water that needs to be removed from the sludge. For example, when W0 is higher than the baseline range (e.g., >80%), the proportion of free water in the sludge is high, and the total amount of water to be removed is large. If the system operates at the initial speed, insufficient residence time will lead to inadequate drying. Therefore, K1 < 1 (e.g., 0.8-0.9), and the total residence time is extended by reducing the speed to ensure sufficient evaporation of free water. When W0 is lower than the baseline range (e.g., <70%), the proportion of bound water in the sludge is already high, and the total amount of water to be removed is small. If the system operates at the initial speed, excessive residence time will lead to over-drying (energy waste). Therefore, K1 > 1 (e.g., 1.1-1.2), and the total residence time is shortened by increasing the speed to avoid ineffective energy consumption. When W0 is within the baseline range (70%-80%), the total amount of water matches the initial speed, and K1 = 1.0, requiring no adjustment. This correction directly links the chain speed to the amount of water that needs to be removed from the sludge, avoiding deviations in drying effect caused by fluctuations in feed characteristics.

[0044] Step 203: Set multiple reference temperature ranges, calculate the average real-time temperature of each layer of the mesh, match the average temperature with the corresponding reference temperature range, and generate a second correction coefficient according to preset conditions. Specifically, take the average real-time temperature of the three layers of the mesh as T_avg = (T1 + T2 + T3) / 3. If T_avg > 55℃ (reference average temperature), the second correction coefficient K2 is set to 1.05-1.1 (K2 increases by 0.05 for every 5℃ increase in the average temperature T_avg); if T_avg < 45℃, the second correction coefficient K2 is set to 0.9-0.95 (K2 decreases by 0.05 for every 5℃ decrease in the average temperature T_avg); if 45℃ ≤ T_avg ≤ 55℃, the second correction coefficient K2 is set to 1.0. The second correction coefficient K2 is generated by matching the average temperature T_avg of the three layers of the mesh with the reference range. The core is to correct the total residence time according to the actual drying efficiency of the environment. For example, when T_avg is higher than the baseline range (e.g., >55℃), the ambient temperature is high, the air saturation water vapor content is high (strong water-carrying capacity), and the drying efficiency is improved. Even if the speed is adjusted to K1, the actual drying efficiency is still higher than required, easily leading to over-drying. Therefore, K2 > 1 (e.g., 1.05-1.1), by increasing the chain speed to further shorten the residence time, utilizing the high drying efficiency to quickly remove moisture and reduce energy consumption. When T_avg is lower than the baseline range (e.g., <45℃), the ambient temperature is low, the air's water-carrying capacity is weak, and the drying efficiency decreases. Therefore, even if the speed is adjusted to K1, the actual drying efficiency is still lower than required, easily leading to substandard drying. Therefore, K2 < 1 (e.g., 0.9-0.95), by reducing the speed to further extend the residence time, compensating for the insufficient drying efficiency. When T_avg is within the baseline range (45℃-55℃), the environmental drying capacity matches the initial speed, K2 = 1.0, and no correction is needed. This modification dynamically links the drying speed to the drying capacity provided by the environment within the storage chamber, avoiding energy waste or drying defects caused by temperature and humidity fluctuations.

[0045] Step 204: The product of the first correction coefficient and the second correction coefficient is used as the transition coefficient. The operating speed of each layer of the chain network is recalculated and used as the transition operating speed of each layer of the chain network to dynamically control the motor speed of each layer of the chain network. Specifically, the transition coefficient K = K1 × K2. The initial speeds V1, V2, and V3 are multiplied by K respectively to obtain the transition operating speed V. 1过 = V1×K、V 2过 =V2×K、V 3过= V3×K (unit: m / min). In this scheme, K1 and K2 achieve bidirectional constraints through a multiplicative relationship. Specifically, when W0 is high but T_avg is also high, the transition coefficient K will not be too small, ensuring both the drying effect (K1 dominates at this time) and avoiding excessive retention due to high temperature (K2 correction); when W0 is low but T_avg is also low, the transition coefficient K will not be too large, avoiding both over-drying (K1 dominates at this time) and ensuring drying efficiency (K2 correction). At the same time, the speeds of the three layers are adjusted according to the same transition coefficient K, maintaining the initial speed ratio (upper layer > middle layer > lower layer), ensuring that the drying division of labor of each layer remains unchanged, so that the chain network speed in the transition stage can adapt to real-time working conditions without disrupting the basic drying logic of the equipment, ultimately achieving the dual goals of achieving the required sludge moisture content and controllable energy consumption during the model training gap period.

[0046] Step 3: Establish an optimization model. Input the operating data and target moisture content into the optimization model for training. Use the conditions that the moisture content of the discharged sludge approaches the target moisture content and the unit energy consumption is minimized, and output the optimal operating speed V of each layer of the chain network. 1优 V 2优 V 3优 It also controls the motor speed of each layer of the chain network.

[0047] In this scheme, the optimization model needs to accurately fit the nonlinear relationships of the system's operating data (such as the dynamic correlation between feed moisture content and the velocity of each layer, and the impact of temperature and humidity coupling on drying efficiency), thereby finding the optimal parameter matching relationship. This scheme uses a random forest regression model as an example, whose core advantages are: strong anti-interference capability against common sensor noise in industrial scenarios (such as temperature and humidity fluctuations of ±1%), no need for complex normalization processing of the original data, and the ability to quantify the influence weight of each layer's parameters on velocity adjustment through feature importance, facilitating engineering debugging.

[0048] From a model structure perspective, the random forest regression model consists of an ensemble framework of 80 independent CART regression decision trees (the number of trees is optimized based on a sample size of 500-5000 sets; too few trees can easily lead to overfitting, while too many trees increase computational latency). It simultaneously predicts V through a multi-output regression mechanism. 1优 V 2优 V 3优Three target values. The input features for each decision tree include: feed sludge moisture content W0 (%), upper layer network temperature and humidity (H1, % RH; T1, ℃), middle layer network temperature and humidity (H2, % RH; T2, ℃), and lower layer network temperature and humidity (H3, % RH; T3, ℃), for a total of 7 dimensions; the output is the network velocity of the corresponding layer (unit: m / min). Since the predictions of the three layer velocities by each decision tree are independent (i.e., a tree's prediction of V...),... 1优 The prediction is not affected by its influence on V 2优 V 3优 The impact of the prediction results), and finally V obtained through mean aggregation. 1优 V 2优 V 3优 Differential adjustments (such as V) can be achieved 1优 Increase by 5%, V 2优 Unchanged, V 3优 (Reduced by 3%) to meet the needs of tiered control.

[0049] Specifically, the optimization model is a random forest regression model, and the loss function of the random forest regression model is defined as:

[0050] ;

[0051] In the formula, W1 is the moisture content of the discharged sludge, W_set is the target moisture content, λ is the energy consumption weighting coefficient, EER is the unit energy consumption, and the value of λ ranges from 0.1 to 0.3. When prioritizing ensuring the moisture content of the discharged sludge meets the standard, λ is set to 0.1-0.2; when prioritizing reducing energy consumption, λ is set to 0.2-0.3. The dynamic adjustment logic of λ is as follows: when the system detects a fluctuation range of >5% in the moisture content of the fed sludge (e.g., a sudden increase from 75% to 82%), λ is automatically set to 0.1-0.2 to prioritize ensuring drying meets the standard and avoid exceeding the discharge moisture content limit; when the feed moisture content is stable (fluctuation range ≤3%) and the unit energy consumption for three consecutive batches is higher than the historical average, λ is automatically set to 0.2-0.3 to prioritize reducing energy consumption and reduce ineffective retention through speed optimization.

[0052] The training process for optimizing the model also includes:

[0053] Step 301: Every 500 sets of new operational data are accumulated, the model is automatically updated and optimized through incremental training. The incremental training process only performs feature learning on the new data and does not overwrite historical training parameters. Incremental training employs a hybrid mechanism of freezing historical trees and adding new decision trees. Specifically, it retains 60 trained historical trees to ensure the stability of core patterns, and uses only 20 newly added, untrained decision trees to learn recent changes in operating conditions, such as changes in drying efficiency caused by seasonal temperature and humidity fluctuations. These 20 new decision trees, together with the historical trees, constitute the updated forest, maintaining a total of 80 trees. This process does not overwrite historical training parameters, preventing the model from forgetting long-term patterns, and keeps the training time per session within 10 minutes to avoid affecting continuous equipment operation.

[0054] Step 302, and during incremental training, high-quality samples in the running data that meet preset conditions are assigned a sample weight of 1.5 times. High-quality samples are those whose effluent sludge moisture content falls within a preset target moisture content fluctuation range, such as... Furthermore, the unit energy consumption is less than the historical average energy consumption, ensuring that the update direction of the optimization model does not deviate from the constraints.

[0055] Step 303 also includes anomaly sample filtering: Before optimizing model training, the running data is preprocessed to remove the following outliers: ① Abnormal values ​​of temperature and humidity fluctuations caused by sensor malfunctions, such as temperature fluctuations >10℃ within 5 seconds; ② Abnormal values ​​of chain speed, such as actual speed deviating from model output by >30%, possibly due to motor malfunction; ③ Abnormal values ​​of sludge batch identification, such as the time-stamped interval between feed and discharge exceeding the normal residence time ±20%. When the proportion of abnormal samples exceeds a preset threshold, such as 10%, a sensor self-check program is triggered to prevent erroneous data from contaminating the model.

[0056] Step 304 also includes monitoring operational data: During the optimization model training process, the influence weight of each operational data on the optimal operating speed is calculated in real time. For example, by quantifying the decrease in Gini impurity, if the importance percentage of a certain layer feature decreases by more than 10% for three consecutive training sessions (e.g., the importance at the lower layer temperature T3 decreases from 30% to 18%), it indicates that there may be air duct blockage or sensor drift, and an alarm is triggered to prompt manual investigation to ensure that the input of the optimization model is consistent with the actual drying environment.

[0057] Step 305 also includes:

[0058] The system detects fluctuations in the moisture content of the feed sludge and identifies the sludge type based on these fluctuations. Specifically, the sludge type can be automatically identified by the fluctuation characteristics of W0 detected by the first moisture meter. For example, if the maximum daily fluctuation of W0 is ≤5 percentage points, it is identified as municipal sludge; if the maximum daily fluctuation of W0 is >8 percentage points, it is identified as industrial sludge; if the fluctuation is >5% and ≤8%, it is trained according to the municipal sludge logic (λ=0.18) by default, or manual confirmation of the sludge type is triggered, such as by using the sludge source ledger for assistance.

[0059] Based on the sludge type, a matching subset of historical operational data is used to train the optimization model. During training, corresponding energy consumption weight coefficients are set according to the sludge type. Specifically, after identifying the sludge type, the corresponding data subset (municipal sludge subset / industrial sludge subset) is used to train the optimization model. By controlling the value of the energy consumption weight coefficient, the optimal operating speed is indirectly controlled in the final output. The training objective for the municipal sludge subset focuses on moisture content stability (λ=0.15), while the training objective for the industrial sludge subset focuses on energy consumption optimization (λ=0.25). It should be noted that λ=0.15 for municipal sludge and λ=0.25 for industrial sludge are the basic objective weights, and the dynamic adjustment logic is a fine-tuning within the basic weights. That is, when dynamic adjustment is triggered for municipal sludge, λ will preferentially approach 0.15 in the 0.1-0.2 range, and for industrial sludge, it will preferentially approach 0.25 in the 0.2-0.3 range, ensuring that long-term type objectives are consistent with short-term operating conditions.

[0060] Step 4: After the optimal operating speed takes effect, the moisture content of the discharged sludge is detected. If the moisture content of the discharged sludge is greater than the target moisture content, the blower adjustment procedure is initiated. This step primarily addresses the issue of excessive moisture content after the optimal speed takes effect. It prioritizes adjusting the blowers based on stratified temperature and humidity, rather than directly reducing the speed. This avoids a surge in energy consumption caused by simply reducing the speed. Furthermore, after the adjustment reaches the target, the speed is re-entered into the model to update, forming a closed-loop correction, shortening the recovery time after the moisture content exceeds the target, and reducing the occurrence of high energy consumption. It should be noted that in this solution, step 401 (adjusting the lower-layer dehumidifying blower) must be executed first. After the lower-layer temperature and humidity reach the target, step 402 (adjusting the middle-layer heating blower) is then executed. If the middle-layer temperature and humidity have already reached the target after the lower-layer adjustment, step 402 is skipped. In addition, the above-mentioned "compliance" conditions are set by engineers according to actual needs, such as the lower layer temperature and humidity returning to the first preset range (H3≤60% RH and T3≥50℃), and / or the middle layer temperature and humidity returning to the second preset range (H2≤55% RH and T2≥55℃); or, the moisture content of the discharged sludge in two consecutive tests decreases by ≥1% compared to before adjustment, ensuring a positive drying trend.

[0061] In one example, the above-mentioned fan regulation procedure includes:

[0062] Step 401: Determine whether the humidity of the bottommost mesh is greater than the first preset relative humidity and whether the temperature of the bottommost mesh is less than the first preset temperature. If so, increase the airflow of the dehumidifying fan according to the preset ratio. In this scheme, the bottommost mesh is the key area for deep drying of sludge and is the inlet of dehumidifying air. For example, if the humidity of the lower layer is high (H3 > 60% RH) and the temperature is low (T3 < 50℃), it indicates that the dehumidification capacity of the bottommost mesh is insufficient, the air is close to saturation, and it cannot effectively absorb the moisture in the sludge, resulting in a decrease in moisture removal efficiency. Therefore, increasing the airflow of the dehumidifying fan can accelerate the discharge speed of moisture at the bottom, reduce the residence time of moisture in the lower layer, and introduce more low-humidity, high-temperature air that has been dehumidified by the evaporator and heated by the condenser, directly reducing the humidity of the lower layer, increasing the temperature, and enhancing the water-carrying capacity of the air.

[0063] Step 402: Determine whether the humidity of the middle layer of the sludge chain is greater than the second preset relative humidity and whether the temperature of the middle layer of the sludge chain is less than the second preset temperature. If so, increase the output temperature of the heating fan by the preset range. In this scheme, the middle layer of the sludge chain is the mixing area of ​​heating air and dehumidifying air. The temperature and humidity of the mixed air directly affect the water migration of the middle and upper layers of sludge (the removal of bound water from the middle layer of sludge depends on the thermal driving force of the mixed air). If the humidity of the middle layer is high (H2 > 55% RH) and the temperature is low (T2 < 55℃), it indicates that the heat input of the heating air is insufficient, and the temperature of the mixed air is insufficient to break the hydrogen bonds of the bound water in the sludge, making it difficult for water to migrate from the interior of the sludge to the surface. Therefore, increasing the output temperature of the heating fan can enhance the dehumidification capacity of the mixed air. On the one hand, the high temperature accelerates the desorption of bound water, and on the other hand, the high-temperature air can hold more water vapor (reducing relative humidity), indirectly enhancing the dehumidification capacity of the middle layer.

[0064] Step 403: Once the air intake volume of the dehumidifying fan and / or the air intake temperature of the heating fan are controlled and meet the standards, re-detect the current real-time operating data of the system and input it into the optimization model to obtain a new optimal operating speed.

[0065] Example 2

[0066] This embodiment proposes another intelligent layered control method for a low-temperature sludge dryer, applied to a sludge dryer with an upper and lower two-layer chain mesh structure. In this scheme, the dryer includes only two moisture meters: a first moisture meter located at the feed inlet of the silo to detect the moisture content W0 (%) of the fed sludge; and a second moisture meter located at the discharge outlet of the silo to detect the moisture content W1 (%) of the discharged sludge. Two temperature and humidity sensor groups are also included: an upper-layer temperature and humidity sensor group and a lower-layer temperature and humidity sensor group. The upper-layer temperature and humidity sensor group is located in the area corresponding to the upper chain mesh to detect the upper ambient humidity H1 (% RH) and upper ambient temperature T1 (°C), while the lower-layer temperature and humidity sensor group is located in the area corresponding to the lower chain mesh to detect the lower ambient humidity H3 (% RH) and lower ambient temperature T3 (°C).

[0067] The main difference between this embodiment and Embodiment 1 is that this embodiment uses a two-layer chain mesh sludge dryer for illustration. Specifically, the number of sensor groups has been reduced, but the data processing process still applies steps 1-3 as described in Embodiment 1.

[0068] In addition, the data processing has been modified accordingly. For example, in step 1, at a certain time t0, 100kg of sludge enters the upper conveyor belt from the feed inlet. After a residence time t1 in the upper conveyor belt and a residence time t3 in the lower conveyor belt, it is discharged from the discharge outlet at time t0+t1+t3. This 100kg of sludge constitutes the "same batch". In this embodiment, the time data t2, which originally represented the middle conveyor belt, can be deleted.

[0069] Furthermore, the calculation process for the transition speed in step 2 is as follows:

[0070] Step 201: Assume the initial velocity is V1:V3=1.5:1.0 (baseline velocity V1=1.5m / min, V3=1.0m / min), and that the velocity of the upper chain network V1 is greater than the velocity of the lower chain network V3.

[0071] Step 202 does not involve data from the middle layer network; please refer to the example described in Implementation Example 1.

[0072] Step 203: Set multiple reference temperature ranges, calculate the average real-time temperature of each layer of the chain network, match the average temperature with the corresponding reference temperature range, and generate a second correction coefficient according to preset conditions; specifically, take the average real-time temperature of two layers of the chain network. The second correction coefficient K2 is set according to the logic described in Example 1 based on T_avg.

[0073] Similarly, the subsequent steps simply involve deleting the data about the middle-layer blockchain in the original embodiment one; examples will not be provided here.

[0074] The above embodiments are merely illustrative of the technical concept and features of the present invention, and are intended to enable those skilled in the art to understand the content of the present invention and implement it accordingly. They should not be construed as limiting the scope of protection of the present invention. All equivalent changes or modifications made based on the essence of the content of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for intelligent stratified control of a low-temperature sludge dryer, characterized in that, Includes the following steps: The system's operational data is acquired. The operational data is linked to the full-process parameters of the same batch of sludge using timestamps, including the moisture content of the feed sludge, the real-time temperature of each layer of the chain mesh, the real-time humidity of each layer of the chain mesh, the real-time operating speed of each layer of the chain mesh, the moisture content of the discharge sludge, and the unit energy consumption. Using the moisture content of the discharged sludge from the same batch of sludge as an index, a first correction coefficient is generated based on the corresponding moisture content of the fed sludge, and a second correction coefficient is generated based on the real-time temperature of each layer of the chain network. The product of the first correction coefficient and the second correction coefficient is used as a transition coefficient, and the operating speed of each layer of the chain network is recalculated as the transition operating speed of each layer of the chain network, thereby dynamically controlling the motor speed of each layer of the chain network. An optimization model is established, and the operating data and target moisture content are input into the optimization model for training. The optimal operating speed of each layer of the chain network is output, and the motor speed of each layer of the chain network is controlled in conjunction with the target moisture content and the unit energy consumption is minimized, with the constraints being the moisture content of the discharged sludge approaching the target moisture content and the minimum unit energy consumption.

2. The intelligent stratified control method for a low-temperature sludge dryer as described in claim 1, characterized in that, The calculation process for the transition speed includes: The initial velocity of each layer of the chain is preset, wherein the initial velocity of each layer of the chain decreases sequentially from top to bottom; Multiple reference moisture content ranges are set, and the moisture content of the feed sludge is matched with the corresponding reference moisture content ranges. Then, a first correction coefficient is generated according to preset conditions. Multiple reference temperature ranges are set, the average real-time temperature of each layer of the chain network is calculated, the average temperature is matched with the corresponding reference temperature range, and a second correction coefficient is generated according to preset conditions. The product of the first correction coefficient and the second correction coefficient is used as a transition coefficient to recalculate the operating speed of each layer of the chain network, which serves as the transition operating speed of each layer of the chain network, and dynamically controls the motor speed of each layer of the chain network.

3. The intelligent stratified control method for a low-temperature sludge dryer as described in claim 1, characterized in that, Once the optimal operating speed takes effect, the moisture content of the discharged sludge is detected. If the moisture content of the discharged sludge is greater than the target moisture content, the blower adjustment procedure is initiated. The blower adjustment procedure includes: Determine whether the humidity of the bottom layer of the dehumidification network is greater than the first preset relative humidity and whether the temperature of the bottom layer of the dehumidification network is less than the first preset temperature. If so, increase the air volume of the dehumidification fan according to the preset ratio. Determine whether the humidity of the middle layer of the chain network is greater than the second preset relative humidity and whether the temperature of the middle layer of the chain network is less than the second preset temperature. If so, increase the output temperature of the heating fan by the preset range. Once the air intake volume of the dehumidifying fan and / or the air intake temperature of the heating fan are controlled and meet the standards, the current real-time operating data of the system is re-detected and input into the optimization model to obtain the new optimal operating speed.

4. The intelligent stratified control method for a low-temperature sludge dryer as described in claim 1, characterized in that, The optimization model is a random forest regression model, and the loss function of the random forest regression model is defined as: ； In the formula, W1 is the moisture content of the discharged sludge, W_set is the target moisture content, λ is the energy consumption weighting coefficient, EER is the unit energy consumption, and the value of λ ranges from 0.1 to 0.

3. When it is necessary to prioritize ensuring that the moisture content of the discharged sludge meets the standard, λ is set to 0.1-0.2, and when it is necessary to prioritize reducing energy consumption, λ is set to 0.2-0.

3.

5. The intelligent stratified control method for a low-temperature sludge dryer as described in claim 4, characterized in that, The training process of the optimized model also includes: Every 500 new sets of running data are accumulated, the optimized model is automatically updated through incremental training. The incremental training process only performs feature learning on the new data and does not cover historical training parameters. Furthermore, during incremental training, high-quality samples in the operational data that meet preset conditions are assigned a sample weight of 1.5 times. These high-quality samples are those in which the moisture content of the discharged sludge falls within the preset target moisture content fluctuation range and the unit energy consumption is less than the historical average energy consumption.

6. The intelligent stratified control method for a low-temperature sludge dryer as described in claim 4, characterized in that, It also includes abnormal sample filtering: before the optimization model is trained, the running data is preprocessed to remove abnormal values ​​such as temperature and humidity jumps, abnormal values ​​of chain network speed, and abnormal values ​​of sludge batch identification caused by sensor failure. When the proportion of abnormal samples exceeds a preset threshold, the sensor self-test program is triggered.

7. The intelligent stratified control method for a low-temperature sludge dryer as described in claim 4, characterized in that, It also includes operational data monitoring: during the training of the optimization model, the influence weight of each operational data on the optimal operating speed is calculated in real time. When the importance ratio of a certain layer feature decreases by more than 10% for three consecutive training sessions, an alarm is triggered to manually check for air duct blockage or sensor drift.

8. The intelligent stratified control method for a low-temperature sludge dryer as described in claim 4, characterized in that, Also includes: The fluctuation characteristics of the moisture content of the feed sludge are detected, and the sludge type is identified based on the fluctuation characteristics; Based on the sludge type, a matching subset of data is retrieved from historical operating data to train the optimization model. During the training process, the corresponding energy consumption weight coefficient is set according to the sludge type.

Images