Energy-saving electric porcelain insulator drying temperature and humidity coordinated control method

Abstract

This invention provides a method for coordinated temperature and humidity control during the drying of porcelain insulators for energy conservation. The method includes: reading the three-dimensional structural data, wall thickness distribution data, initial moisture content data, and loading location data of the porcelain insulator to be dried; establishing a structural region model; generating a set of structurally weak zones and an index table of structurally weak zones; collecting temperature response data, humidity response data, mass change data, dehumidification load data, and cumulative energy consumption data during the initial drying stage; generating a drying state feature set; establishing a drying front migration model based on the drying state feature set; predicting the current position, migration direction, and expected traversal area of ​​the front; generating front collision risk results; determining stage boundary data and generating a constraint set based on the front collision risk results; and selecting the optimal candidate parameter sequence using a wolf pack algorithm and a reinforcement learning parameter evaluator to generate a coordinated temperature and humidity control parameter sequence.

Description

Technical Field

[0001] This invention relates to the field of computer data processing technology, and in particular to a method for coordinated control of temperature and humidity for drying porcelain insulators in an energy-saving manner. Background Technology

[0002] After molding, porcelain insulators require drying. Porcelain insulators have a non-uniform structure, including sheds, porcelain heads, and thickness transition sections. During the drying process, if the temperature rises too quickly, the humidity is released improperly, or the moisture removal intensity changes too much, localized abrupt changes in moisture migration can easily occur at the root of the sheds, in the abrupt thickness transition area, and in the porcelain head transition area, leading to defects such as hidden cracks, deformation, and strength dispersion. If the overall heating rate is reduced or the humidity holding time is extended to avoid the above defects, the drying cycle and unit energy consumption will increase.

[0003] Current drying processes for porcelain insulators mostly rely on product specifications, experience curves, or feedback from the drying room's temperature and humidity to set temperature and humidity control parameters. They typically focus on the drying chamber's ambient temperature, relative humidity, exhaust ventilation, and total drying time, rarely considering the three-dimensional structural differences of the porcelain insulators, the location of structurally weak areas, initial moisture content, mass changes, humidity drop, dehumidification load, cumulative energy consumption, and historical drying defect data. Especially for the base of the skirts, areas of abrupt thickness changes, and the transition zone of the porcelain head, existing methods struggle to estimate the location, migration speed, and expected area of ​​the drying front within the blank based on initial drying response data. They also find it difficult to determine whether the drying front will pass through structurally weak areas under high-risk conditions.

[0004] Therefore, this invention proposes a method for coordinated control of temperature and humidity for drying porcelain insulators in an energy-saving manner. The information disclosed in the background section is only for enhancing understanding of the background of this disclosure and may therefore contain prior art information that is not common knowledge to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by providing a method for coordinated control of temperature and humidity for drying porcelain insulators in an energy-saving manner, thereby solving the technical problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A method for coordinated control of drying temperature and humidity in electrical porcelain insulators for energy conservation includes the following steps: S1. Read the three-dimensional structural data, wall thickness distribution data, skirt spacing data, initial moisture content data and loading position data of the porcelain insulator to be dried, establish a structural region model, identify candidate weak areas, and generate a set of structural weak areas and a structural weak area index table by combining historical drying defect data. S2. Collect temperature response data, humidity response data, mass change data, dehumidification load data and cumulative energy consumption data of the initial drying stage, generate the initial drying stage response data sequence, associate it with the structural weak area index table, and calculate the drying state feature set. S3. Based on the structural region model, initial moisture content data, and dry state feature set, establish a dry front migration model, determine the current position, migration speed, migration direction, and expected traversal area of ​​the front, and match it with the set of structurally weak areas to generate front collision risk results; S4. Based on the results of the frontier collision risk, divide the safe migration stage, the risk approach stage, and the weak zone passage stage. Combine the unit energy consumption dehydration contribution value and historical drying defect data to determine the stage boundary data and generate a parameter generation constraint set. S5. Generate a candidate parameter sequence group using the parameter generation constraint set as the search boundary. Select the optimal candidate parameter sequence through the wolf pack algorithm and the reinforcement learning parameter evaluator to generate the temperature and humidity coordinated control parameter sequence.

[0007] S1 specifically includes: reading the three-dimensional structural data, wall thickness distribution data, skirt spacing data, porcelain head transition zone data, initial moisture content data, and loading position data of the porcelain insulator to be dried; dividing the structural regions according to axial height, radial thickness, and skirt layer and configuring region identifiers to generate a structural region model; calculating the wall thickness variation range, thickness transition range, skirt root bending degree, and porcelain head connection transition degree based on the structural region model; marking structural regions exceeding the corresponding thresholds as candidate weak areas and generating a candidate weak area set; matching the candidate weak area set with historical drying defect data; using a weighted calculation method, assigning risk weights based on defect occurrence frequency, defect severity, and defect location matching degree; and generating a structural weak area set and a structural weak area index table.

[0008] S2 specifically includes: after the porcelain insulator to be dried enters the initial drying stage, collecting temperature response data, humidity response data, mass change data, dehumidification load data, and cumulative energy consumption data, and generating an initial drying stage response data sequence according to a unified sampling time identifier; temporally and spatially correlating the initial drying stage response data sequence with the structural weak zone index table, determining the temperature and humidity response values ​​corresponding to each structural weak zone based on spatial distance weighting, and determining the heating sequence, dehumidification path, and moisture accumulation location corresponding to each structural weak zone according to the loading location data, generating structural weak zone response correlation data; calculating the mass change rate, humidity drop rate, temperature lag, dehumidification load change rate, and unit energy consumption dehydration contribution value based on the structural weak zone response correlation data, generating a drying state feature set.

[0009] S3 specifically includes: establishing a drying front migration model based on the structural region model, initial moisture content data, and dry state feature set; calculating the front state value of each structural region based on the state data with weights to determine the current position of the front and the local moisture content change trend; predicting the migration speed, migration direction, and expected traversal area of ​​the drying front based on the current position of the front, the rate of mass change, and the rate of humidity decline under continuous sampling time markers, and generating a drying front migration prediction result; matching the drying front migration prediction result with the set of structural weak areas, and generating a front collision risk result including risk area identifier, risk occurrence stage, risk intensity calculated with weights, and risk cause type when the expected traversal area coincides with the structural weak area and the migration speed exceeds the safe migration threshold.

[0010] S4 specifically includes: based on the risk area identification, risk intensity and risk cause type in the front collision risk results, and combined with the front approach distance between the current front position and the weak area position, the drying process is divided into a safe migration stage, a risk approach stage and a weak area passage stage to obtain the drying stage division results; Based on the results of the drying stage division, the unit energy consumption dehydration contribution value, and historical drying defect data, the temperature rise boundary, humidity maintenance boundary, moisture removal intensity boundary, and energy consumption constraint boundary of each stage are determined, and the stage boundary data are obtained. The stage boundary data is combined with risk area identifiers, risk intensity, and risk cause types to generate a parameter-generated constraint set, which includes temperature target range, humidity target range, temperature rise slope range, humidity holding duration range, dehumidification window range, dehumidification intensity range, energy consumption upper limit range, and constraint priority.

[0011] S5 specifically includes: using the parameter generation constraint set as the search boundary, forming a candidate parameter sequence by combining the target temperature value, target humidity value, temperature rise slope, humidity holding time, dehumidification window, dehumidification intensity, and drying endpoint determination parameters, and using them as candidate individuals in the wolf pack algorithm to obtain a candidate parameter sequence group; The candidate parameter sequence group is input into the drying front migration model to calculate the reduction in front collision risk, the reduction in unit energy consumption, the reduction in historical drying defect risk, and the stability at the drying endpoint. The reinforcement learning parameter evaluator takes the drying state feature set as the state input, the candidate parameter sequence as the action input, and incorporates the penalty calculation for the constraint violation of the parameter generation constraint set by the candidate parameter sequence to output a comprehensive evaluation result. Based on the comprehensive evaluation results, the candidate parameter sequence group is iteratively updated, and the optimal candidate parameter sequence that meets the requirements of frontier collision risk threshold, energy consumption constraint boundary and drying endpoint determination is selected to generate temperature and humidity coordinated control parameter sequence.

[0012] The beneficial effects of this invention are as follows: This invention establishes a structural region model and generates a set of structurally weak areas, enabling the root of the umbrella skirt, the area of ​​abrupt changes in thickness, and the transition area of ​​the porcelain head to be digitized and identified. This avoids setting drying parameters solely based on the average state of the entire porcelain insulator, and improves the structural adaptability of the temperature and humidity coordinated control parameters.

[0013] This invention generates a set of drying state features by collecting temperature response data, humidity response data, mass change data, dehumidification load data, and cumulative energy consumption data. This allows the drying parameter generation process to simultaneously reflect the rate of water loss, humidity drop, temperature lag, and the contribution of dehydration per unit of energy consumption, avoiding parameter deviations caused by relying solely on empirical curves.

[0014] This invention establishes a drying front migration model to predict the current position, migration speed, migration direction, and expected traversal area of ​​the drying front. It then matches this model with a set of structurally weak areas to generate front collision risk results. This allows for early identification of migration mutation risks when the drying front passes through structurally weak areas, reducing the probability of latent cracks, deformation, and strength anomalies.

[0015] This invention divides the drying process into a safe migration stage, a risk approach stage, and a weak zone passage stage based on the front-end collision risk results, and generates heating boundary, humidity maintenance boundary, dehumidification intensity boundary, and energy consumption constraint boundary respectively. This gives the parameter generation of different drying stages clear constraints and avoids energy waste caused by conservative drying throughout the entire process.

[0016] This invention uses the parameter generation constraint set as the search boundary and combines the wolf pack algorithm and the reinforcement learning parameter evaluator to screen the optimal candidate parameter sequence. It can generate a temperature and humidity coordinated control parameter sequence under the conditions of meeting the front collision risk threshold, energy consumption constraint boundary and drying endpoint determination requirements, so as to achieve the synergy of defect suppression and energy saving optimization. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the energy-saving temperature and humidity coordinated control method for drying porcelain insulators according to the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example: Figure 1 As shown, this embodiment provides a method for coordinated control of temperature and humidity for drying porcelain insulators in an energy-saving manner, including the following steps: S1. Read the three-dimensional structural data, wall thickness distribution data, skirt spacing data, initial moisture content data and loading position data of the porcelain insulator to be dried, establish a structural region model, identify candidate weak areas, and generate a set of structural weak areas and a structural weak area index table by combining historical drying defect data. S2. Collect temperature response data, humidity response data, mass change data, dehumidification load data and cumulative energy consumption data of the initial drying stage, generate the initial drying stage response data sequence, associate it with the structural weak area index table, and calculate the drying state feature set. S3. Based on the structural region model, initial moisture content data, and dry state feature set, establish a dry front migration model, determine the current position, migration speed, migration direction, and expected traversal area of ​​the front, and match it with the set of structurally weak areas to generate front collision risk results; S4. Based on the results of the frontier collision risk, divide the safe migration stage, the risk approach stage, and the weak zone passage stage. Combine the unit energy consumption dehydration contribution value and historical drying defect data to determine the stage boundary data and generate a parameter generation constraint set. S5. Generate a candidate parameter sequence group using the parameter generation constraint set as the search boundary. Select the optimal candidate parameter sequence through the wolf pack algorithm and the reinforcement learning parameter evaluator to generate the temperature and humidity coordinated control parameter sequence.

[0020] This embodiment can also be referred to as a method for generating temperature and humidity coordinated control parameters for drying porcelain insulators in an energy-saving manner.

[0021] S1 specifically includes the following sub-steps: S110. Read the three-dimensional structural data, wall thickness distribution data, shed spacing data, porcelain head transition zone data, initial moisture content data, and loading position data of the porcelain insulator to be dried. Establish a structural region model according to the axial height, radial thickness, and shed layer of the porcelain insulator, and assign a unique region identifier to each structural region to obtain the structural region model of the porcelain insulator to be dried.

[0022] Specifically, the 3D structural data originates from the product design model, 3D scanned point cloud, or molding die parameters; the wall thickness distribution data originates from the design wall thickness table, ultrasonic thickness measurement records, or cross-sectional calculation results of the 3D structural data; the skirt spacing data and ceramic head transition zone data originate from the product structural drawings or 3D structural data analysis results; the initial moisture content data originates from post-molding weighing inspection, sampling drying inspection, or online moisture detection records; and the loading position data originates from the kiln car loading work order, loading position code, or drying chamber entrance scanning record. The system converts the 3D structural data into structural contour data in a unified coordinate system, establishing axial height coordinates with the central axis, radial thickness coordinates with the wall thickness direction, and establishing skirt layer numbering according to the top-to-bottom arrangement of the skirts.

[0023] Subsequently, based on the geometric boundaries of the porcelain head section, skirt section, skirt root, thickness transition section, and bottom connection section, the system divides the porcelain insulator to be dried into multiple structural regions, and assigns a region identifier to each region, consisting of a product number, axial section number, radial layer number, and skirt level number. These region identifiers are used to subsequently incorporate initial drying response data, historical drying defect data, and drying front migration prediction results. For example, the outer layer of the skirt root of the second skirt section can be configured as an independent region identifier, allowing subsequent crack locations and humidity response locations to be mapped to the same structural region model.

[0024] S120. Based on the structural region model, calculate the wall thickness variation range, thickness transition range, skirt root bending degree, and ceramic head connection transition degree of each structural region. Mark structural regions whose wall thickness variation range exceeds a preset wall thickness difference threshold, thickness transition range exceeds a preset transition threshold, skirt root bending degree exceeds a preset bending threshold, or ceramic head connection transition degree exceeds a preset connection threshold as candidate weak regions, and generate a set of candidate weak regions.

[0025] Specifically, the system reads multiple wall thickness sampling points in each structural region, takes the difference between the maximum and minimum wall thickness as the wall thickness variation range, takes the average wall thickness difference between the current structural region and the adjacent structural regions as the thickness transition range, takes the angle between adjacent tangential directions of the center line of the umbrella skirt root section as the bending degree of the umbrella skirt root, and takes the cross-sectional curvature change or wall thickness change rate of the porcelain head connection area as the porcelain head connection transition degree.

[0026] The variation in wall thickness is calculated using the following formula:

[0027] in, This represents the variation in wall thickness of the i-th structural region. This represents the maximum wall thickness within the i-th structural region. This represents the minimum wall thickness within the i-th structural region, where i represents the structural region number.

[0028] The thickness transition range is calculated using the following formula:

[0029] in, This represents the thickness transition amplitude between the i-th structural region and its adjacent structural regions. This represents the average wall thickness of the i-th structural region. This represents the average wall thickness of the preceding structural region adjacent to the i-th structural region.

[0030] The preset wall thickness difference threshold, preset transition threshold, preset bending threshold, and preset connection threshold are derived from the statistical results of historical qualified batches of products of the same specification, the allowable structural deviations in the process documents, the statistical values ​​of structural parameters corresponding to historical cracked areas, or the enterprise's quality control standards. When no existing threshold exists, the upper quantile value of the corresponding index of historical crack-free batches is used as the initial threshold.

[0031] The system further calculates the vulnerability risk value:

[0032] in, This represents the vulnerability risk value of the i-th structural region. This represents the normalized variation in wall thickness. This indicates the normalized thickness transition range. This indicates the degree of bending at the base of the umbrella skirt after normalization. This indicates the normalized transition degree of the ceramic head connection, where a, b, c, and d represent the weight coefficients of the corresponding data items.

[0033] In a preferred embodiment, the weighting coefficients range as follows: a = 0.25-0.35, b = 0.15-0.25, c = 0.25-0.35, and d = 0.15-0.25, satisfying a+b+c+d=1. More specifically, in this embodiment, a = 0.3, b = 0.2, c = 0.3, and d = 0.2.

[0034] The weighting coefficients can also be dynamically set by expert scoring based on different specifications of porcelain insulators using the analytic hierarchy process (AHP) to ensure that the contribution of sudden changes in wall thickness and structural bending to the risk of weakness conforms to the actual physical characteristics.

[0035] All data items involved in the weighted calculation are normalized to 0 to 1 according to the value range of historical qualified batches and historical drying defect batches of the same model of product. For example, if the maximum wall thickness of a certain structural region is 32mm and the minimum wall thickness is 24mm, with a wall thickness variation range of 8mm; if the preset wall thickness difference threshold is 6mm, then this structural region is written into the candidate weak area set.

[0036] S130. Match the candidate weak area set with historical drying defect data, assign risk weights to the candidate weak areas based on historical crack locations, deformation locations, and strength anomaly locations, obtain a set of structural weak areas including the umbrella skirt root, thickness abrupt change zone, and ceramic head transition zone, and generate a structural weak area index table.

[0037] Specifically, historical drying defect data includes at least the product model, batch number, defect type, defect location, defect severity, defect discovery process, and corresponding drying batch. The data comes from post-drying appearance inspection records, trimming records, pre-glazing inspection records, post-firing crack records, mechanical strength test records, and the quality traceability system.

[0038] The system converts the locations of historical defects into region identifiers consistent with the structural region model and determines whether the region identifier falls into the candidate weak region set. When the location of a historical crack, deformation, or strength anomaly is consistent with the region identifier of a candidate weak region, or is located in an adjacent structural region of the candidate weak region, the system confirms that the two match and determines the risk weight based on the frequency of defect occurrence, the severity of defect, and the degree of matching of defect location.

[0039] Risk weights are calculated using the following formula:

[0040] in, This represents the risk weight of the i-th candidate weak region. This indicates the frequency of historical defects after normalization. This indicates the severity of historical defects after normalization. This indicates the degree of matching of historical defect locations after normalization. , , These represent the weight coefficients of the corresponding data items.

[0041] As a preferred embodiment, the range of values ​​for the risk weight coefficient is: It is 0.4-0.5. It is 0.3-0.4. It is between 0.1 and 0.2, and satisfies + + =1. In this embodiment, the specific value is: =0.4, =0.4, =0.2. The frequency of historical defects. It occupies the highest weight to highlight the high warning significance of historically frequent defect areas in current process generation.

[0042] The system identifies candidate weak areas whose risk weights reach a preset weight threshold as structural weak areas. Candidate weak areas that do not reach the preset weight threshold but whose risk value exceeds a preset high-risk threshold are also included in the structural weak area set. The structural weak area index table includes at least the area identifier, area name, axial height range, radial thickness range, skirt layer, wall thickness variation range, thickness transition range, skirt root bending degree, ceramic head connection transition degree, historical defect type, risk weight, and subsequent mapping marker.

[0043] The set of structurally weak areas and the index table of structurally weak areas serve as the spatial mapping basis for generating structurally weak area response correlation data in S220, and as the matching basis for judging the leading-edge collision risk in S330. For example, if the candidate weak area at the root of the second umbrella skirt shows axial fine cracks in all three batches of the same model of product, and the crack locations are all mapped to the area identifier of the candidate weak area, then the system includes it in the set of structurally weak areas and retains the corresponding risk weight in the index table of structurally weak areas.

[0044] S2 specifically includes the following sub-steps: S210. After the porcelain insulators to be dried enter the initial drying stage, collect the temperature response data, humidity response data, mass change data, dehumidification load data and cumulative energy consumption data of the corresponding batch, and establish a unified time identifier for the above data according to the sampling time to obtain the initial drying stage response data sequence.

[0045] Specifically, the initial drying stage refers to the period from when the porcelain insulator to be dried enters the drying chamber until the surface moisture begins to be released stably. It can be set to the first 30 to 120 minutes after the start of drying according to the process documents, or it can be determined according to the stage before the mass change rate first reaches the preset water loss rate threshold.

[0046] Temperature response data comes from drying chamber temperature sensors, kiln car layer temperature sensors, or infrared thermometers on the surface of the billet; humidity response data comes from inlet air humidity sensors, return air humidity sensors, and exhaust port humidity sensors; mass change data comes from kiln car weighing devices, sampled billet weighing records, or online weight acquisition devices; exhaust load data comes from exhaust fan operating power, exhaust valve opening, exhaust air temperature, and exhaust air humidity; cumulative energy consumption data comes from electricity meters, gas meters, steam flow meters, or energy management systems. Cumulative energy consumption data is used to calculate the unit energy consumption dehydration contribution value in S230.

[0047] The system uses 1 min, 3 min, or 5 min as a unified sampling time identifier. It averages high-frequency temperature and humidity response data within the same sampling time window, and performs staged interpolation or linear completion on low-frequency quality change data according to adjacent sampling time identifiers. It also synchronously writes dehumidification load data and cumulative energy consumption data into the same record. For data from offline sensors, missing timestamps, or data exceeding the device's measurement range, the system removes the abnormal data, completes it with adjacent valid data, and retains the abnormality marker to prevent abnormal data from directly entering subsequent structurally weak area response-related data.

[0048] For example, when the temperature and humidity sensor collects data once every 1 minute and the kiln car weighing device outputs mass data once every 5 minutes, the system uses 5 minutes as a unified sampling time marker, takes the average value of the temperature response data and humidity response data within 5 minutes, and merges it with the mass change data, dehumidification load data and cumulative energy consumption data at that time point into a single drying initial stage response record.

[0049] S220. The initial stage of drying response data sequence is correlated with the index table of structural weak areas in time and space. Based on the loading location data, the heating sequence, dehumidification path and moisture accumulation location corresponding to different structural weak areas are determined, and structural weak area response correlation data is generated. The structural weak area response correlation data is used to characterize the temperature, humidity and dehumidification effects on each structural weak area in the initial stage of drying.

[0050] Specifically, the system determines the row number, column number, layer height, orientation angle, and spatial distance relative to the air inlet, air outlet, and exhaust outlet of the porcelain insulator to be dried in the kiln car based on the loading position data. Then, based on the area identifier, axial height range, radial thickness range, and skirt level in the structural weak area index table, the system determines the actual mapping position of each structural weak area in the drying chamber space.

[0051] For each structurally weak area, the system selects the nearest temperature and humidity sampling points, or performs a weighted calculation based on the spatial distance between multiple sampling points and the structurally weak area, to obtain the corresponding temperature and humidity response values ​​for that structurally weak area at each sampling time marker:

[0052] in, This represents the temperature or humidity response value of the i-th structurally weak area at the t-th sampling time marker. This represents the temperature or humidity sample value of the j-th sampling point at the t-th sampling time. This represents the spatial distance between the i-th structurally weak area and the j-th sampling point. This represents a correction constant used to avoid a denominator of 0, and n represents the number of sampling points involved in the calculation.

[0053] The system determines the heating sequence based on the distance from the structurally weak area to the air inlet, the airflow path, and the order of temperature rise; it determines the dehumidification path based on the connection path from the structurally weak area to the dehumidification outlet, the dehumidification valve opening, and the change in exhaust humidity; when the humidity response value corresponding to a structurally weak area is higher than the average humidity of the same floor within multiple consecutive sampling time intervals, the structurally weak area is marked as a moisture accumulation location.

[0054] For example, the weak structural area at the root of the second umbrella skirt is located on the second layer of the kiln car and close to the return air side. The system calculates the humidity response value based on the distance between it and the three humidity sampling points. If the humidity response value is higher than the average humidity of the same layer for three consecutive sampling time markers, a moisture accumulation marker is written into the response correlation data of the weak structural area.

[0055] S230. Calculate the rate of mass change, rate of humidity decrease, degree of temperature lag, degree of change of dehumidification load, and unit energy consumption contribution value of dehydration based on the response correlation data of the structurally weak area, and generate a dry state feature set to characterize the state of moisture migration; the dry state feature set serves as the input data for the subsequent establishment of the dry front migration model.

[0056] Specifically, the rate of mass change refers to the amount of mass reduction of the porcelain insulator or sample blank to be dried per unit time, used to characterize the rate of water loss; the rate of humidity reduction refers to the decrease in return air humidity, exhaust port humidity, or humidity response value corresponding to the structurally weak area after dehumidification per unit time, used to characterize the speed at which moisture is carried away; the degree of temperature hysteresis refers to the difference in response value of the structurally weak area relative to the target temperature of the drying chamber or the average temperature of the same floor; the degree of change in dehumidification load refers to the change in power of the dehumidification fan, opening degree of the dehumidification valve, or exhaust moisture load between adjacent sampling time markers; the unit energy consumption dehydration contribution value refers to the amount of mass reduction corresponding to unit energy consumption, used to determine whether the current energy is effectively used for dehydration.

[0057] The rate of mass change is calculated using the following formula:

[0058] in, This represents the rate of quality change at the t-th sampling time. This indicates the quality data at the previous sampling time. This indicates the quality data at the current sampling time. This indicates the time interval between two adjacent sampling time markers.

[0059] The rate of humidity decline is calculated using the following formula:

[0060] in, This represents the humidity decay rate at the t-th sampling time. This indicates the humidity response data at the previous sampling time. This indicates the humidity response data at the current sampling time.

[0061] The energy consumption contribution of dehydration per unit is calculated using the following formula:

[0062] in, This represents the unit energy consumption dehydration contribution value under the t-th sampling time marker. This represents the cumulative energy consumption data under the current sampling time identifier. This indicates the cumulative energy consumption data under the previous sampling time marker.

[0063] The dry state feature set includes at least the region identifier, sampling time identifier, mass change rate, humidity decline rate, temperature lag degree, dehumidification load change degree, unit energy consumption dehydration contribution value, moisture accumulation marker, and subsequent model input marker.

[0064] For example, if the mass of the sampled blank decreases from 12500g to 12480g between two adjacent 5-minute sampling time markers, and the cumulative energy consumption increases from 30kWh to 34kWh, then the mass change is 20g, and the unit energy consumption dehydration contribution is 5g / kWh. If the humidity drop rate of the weak structural area at the root of the second umbrella skirt is lower than the average value of the same layer during the same time period, then this record is written into the drying state feature set and used as input data for S310 to establish the drying front migration model.

[0065] S3 specifically includes the following sub-steps: S310. Based on the structural region model, initial moisture content data, and drying state feature set, establish a drying front migration model. The drying front migration model is used to describe the migration state of the moisture evaporation front inside the porcelain insulator to be dried between different structural regions, and outputs the current position of the front and the local moisture content change trend of each structural region.

[0066] Specifically, the drying front refers to the moving boundary inside the porcelain insulator to be dried, where it transitions from a high moisture content state to a low moisture content state. It is not a directly measurable physical boundary, but rather a moisture migration state boundary inferred by the structural region model, initial moisture content data, mass change rate, humidity decline rate, temperature hysteresis, unit energy consumption dehydration contribution value, and moisture accumulation markers.

[0067] The system uses the region identifiers in the structural region model as the calculation unit, calls the dry state feature set generated by S230, and combines the axial height range, radial thickness range and umbrella skirt level in the structural weak area index table to calculate the leading edge state value of each structural region under different sampling time identifiers.

[0068] The leading-edge state value is used to characterize the degree of transition of the structural region from an initial water-bearing state to a stable water-loss state, and is calculated according to the following formula:

[0069] in, This represents the leading-edge state value of the i-th structural region at the t-th sampling time. This represents the normalized initial moisture content data. This represents the normalized rate of change of mass. This represents the normalized rate of humidity decrease. This indicates the normalized temperature lag. This represents the moisture accumulation marker in the i-th structural region at the t-th sampling time. , , , , These represent the weight coefficients of the corresponding data items.

[0070] The weight coefficients for each data item have the following ranges: It is 0.1-0.2. It is 0.2-0.3. It is 0.2-0.3. It is 0.1-0.2. The value is between 0.1 and 0.2. In this embodiment, the specific value is: =0.15, =0.25, =0.25, =0.15, =0.20. This combination makes the frontier state value more sensitive to the dynamic migration response of moisture by increasing the weights of water loss rate (Vm) and humidity decline (Vh).

[0071] The system divides each structural region into high moisture content region, transitional region, and low moisture content region based on the frontier state value. When a change occurs between adjacent structural regions from high moisture content region to transitional region or low moisture content region, the position between the adjacent structural regions is determined as the current position of the frontier, and the corresponding region identifier and local moisture content change trend are recorded.

[0072] For example, if the humidity drop rate in the outer layer region at the root of the second umbrella skirt is lower than the average value of the same layer, and the temperature lag is higher than that of the adjacent outer layer region, the system will identify it as a region where moisture release is blocked; if the adjacent outer layer region has entered a low moisture content state, the boundary between the two regions will be determined as the current position of the leading edge under the sampling time marker.

[0073] S320. Using the drying front migration model, based on the current front location, local humidity change trend, mass change rate, and humidity decline rate, predict the migration speed, migration direction, and expected area of ​​the drying front in the subsequent drying stage, and generate the drying front migration prediction results.

[0074] Specifically, the system reads the current position of the leading edge under multiple consecutive sampling time markers and projects the current position of the leading edge onto the axial height direction, radial thickness direction, and umbrella skirt layer direction of the structural region model; if the current position of the leading edge moves from the outer layer to the middle or inner layer along the radial thickness direction, the migration direction is recorded as radial inward movement; if the current position of the leading edge moves from the outer edge of the umbrella skirt to the root of the umbrella skirt along the axial height direction, the migration direction is recorded as axial root approach; if the current position of the leading edge moves towards the ceramic head connection area, the migration direction is recorded as ceramic head transition approach.

[0075] The migration velocity of the drying front is calculated according to the following formula:

[0076] in, This represents the migration velocity of the drying front of the i-th structural region at the t-th sampling time. This indicates the current position of the leading edge of the i-th structural region at the t-th sampling time marker. This indicates the current position of the leading edge of the i-th structural region at the previous sampling time. This indicates the time interval between two adjacent sampling time markers.

[0077] Based on the adjacency relationships in the structural region model, the system expands from the current leading edge position to subsequent adjacent structural regions according to the migration direction to determine the expected traversal area. When the migration direction points to the root of the umbrella skirt, the abrupt change in thickness, or the transition zone of the ceramic head, the corresponding region identifier is written into the expected traversal area. If the migration speed increases under the continuous sampling time marker, while the humidity decline rate decreases or the temperature lag increases, the system marks the expected traversal area as a migration abrupt change area of ​​interest, and records the migration speed, migration direction, expected traversal area, and migration abrupt change area of ​​interest marker in the migration prediction results of the drying front.

[0078] For example, if the current leading edge position moves from the outer layer region of the second umbrella skirt to the middle layer region of the root of the second umbrella skirt, and the distance of the leading edge position change increases within two consecutive sampling time markers, then the system will write the middle layer region of the root of the second umbrella skirt into the expected passing area and record the migration direction as transitioning from the outer layer to the root.

[0079] S330. Match the predicted migration results of the drying front with the set of structurally weak areas. When the predicted area coincides with the root of the umbrella skirt, the abrupt change in thickness, or the transition zone of the ceramic head, and the migration speed exceeds the safe migration threshold of the corresponding structurally weak area, generate the front collision risk result. The front collision risk result includes the risk area identification, the risk occurrence stage, the risk intensity, and the risk cause type, and serves as the basis for generating the parameter boundary in the future.

[0080] Specifically, the system compares the region identifiers in the expected path area with the region identifiers in the set of structurally weak areas one by one. When the two match, or when the expected path area is within the range of adjacent areas recorded in the structurally weak area index table, it is confirmed that the drying front will pass through the corresponding structurally weak area. The safe migration threshold is derived from the statistical value of the drying front migration speed of historical crack-free batches of the same specification, the critical value of the migration speed corresponding to historical cracked batches, and the enterprise's drying process verification data or trial production records. When no existing threshold exists, the upper quantile value of the migration speed of the same type of structurally weak area in historical crack-free batches is used as the initial safe migration threshold, and it is subsequently corrected based on historical drying defect data.

[0081] The intensity of the leading-edge collision risk is calculated using the following formula:

[0082] in, This represents the frontal collision risk intensity of the i-th structurally weak region at the t-th sampling time marker. This represents the safe migration threshold corresponding to the i-th structurally weak region. This represents the vulnerability risk value of the i-th structural region. This represents the risk weight of the i-th candidate weak region. This represents the moisture accumulation marker in the i-th structurally weak area at the t-th sampling time. , , , These represent the weight coefficients of the corresponding data items.

[0083] The weight coefficients for each data item have the following ranges: It is 0.3-0.4. It is 0.2-0.3. It is 0.2-0.3. The value is between 0.1 and 0.2. In this embodiment, the specific value is: =0.35, =0.25, =0.25, =0.15.

[0084] Safe migration threshold The specific speed is set to 1.5 mm / min in the axial height direction, 1.2 mm / min in the radial thickness direction, and 1.8 mm / min in the umbrella skirt layer direction.

[0085] The system classifies the frontal collision risk results into low risk, medium risk, and high risk based on the risk intensity, and determines the risk cause type based on the trigger item. When the migration speed exceeds the safe migration threshold, the risk cause type is recorded as migration speed too fast. When the humidity drop rate is lower than the average value of the same layer or there is a moisture accumulation marker, the risk cause type is recorded as continuous moisture accumulation. When the temperature lag exceeds the preset lag threshold, the risk cause type is recorded as temperature response lag.

[0086] The results of the front-end collision risk include at least the risk area identification, the risk occurrence stage, the risk intensity, the risk cause type, the corresponding safe migration threshold, the corresponding migration speed, and the subsequent constraint generation mark, and are output to S410 to divide the safe migration stage, the risk approach stage, and the weak zone passage stage.

[0087] For example, the root region of the second umbrella skirt belongs to the set of structurally weak areas. The migration prediction results of the drying front show that this region will be passed through in the next stage, and the migration speed is 2.4 mm / min, which is higher than the safe migration threshold of 1.6 mm / min. At the same time, there are moisture accumulation markers in this region. Then the system generates the front collision risk result and records the risk cause type as too fast migration speed and continuous moisture accumulation.

[0088] S4 specifically includes the following sub-steps: S410. Based on the results of the frontal collision risk assessment, the drying process is divided into a safe migration stage, a risk approach stage, and a weak zone passage stage, thus obtaining the drying stage division results.

[0089] Specifically, the system reads the risk area identifier, risk occurrence stage, risk intensity, risk cause type, corresponding safe migration threshold, corresponding migration speed and subsequent constraint generation mark output by S330, and uses the current position of the front edge, the expected traversed area and the set of structural weak areas in the structural region model as the basis for division.

[0090] The safe migration stage refers to the stage where the area the drying front is expected to pass through does not fall into the set of structurally weak areas, or the collision risk intensity of the front is lower than the preset low-risk threshold; the risk approach stage refers to the stage where the migration prediction results of the drying front show that the area it is expected to pass through will enter the structurally weak area within a subsequent preset time window, and the collision risk intensity of the front reaches the preset medium-risk threshold; the weak area passing stage refers to the stage where the current position of the front is already within the area marked by the corresponding area of ​​the structurally weak area, or the area it is expected to pass through coincides with the set of structurally weak areas and the migration speed exceeds the corresponding safe migration threshold.

[0091] The system calculates the approach distance between the current position of the leading edge and the location of the structural weak zone:

[0092] in, This represents the approach distance of the leading edge of the i-th structural region at the t-th sampling time marker. This indicates the current position of the leading edge of the i-th structural region at the t-th sampling time marker. This indicates the location of the weak area corresponding to the i-th structural weak area.

[0093] When the approach distance to the leading edge is less than the preset approach distance threshold and the collision risk intensity of the leading edge reaches the preset medium risk threshold, the system divides the corresponding time window into the risk approach stage; when the current edge enters the structural weak zone at the current position and the migration speed exceeds the safe migration threshold, the system divides the corresponding time window into the weak zone passage stage.

[0094] For example, if the root of the second umbrella skirt is a structurally weak area, and the current position of the drying front is 8mm away from the boundary of this structurally weak area, the preset approach distance threshold is 10mm, and the risk intensity reaches the medium risk level, then the system will divide the next drying time window into the risk approach stage and output the stage identifier to S420.

[0095] S420. Combining the results of the drying stage division, the unit energy consumption dehydration contribution value, and historical drying defect data, the temperature rise boundary, humidity maintenance boundary, moisture removal intensity boundary, and energy consumption constraint boundary corresponding to each drying stage are determined to obtain the stage boundary data.

[0096] Specifically, the temperature rise boundary includes the range of the target temperature value and the range of the temperature rise slope allowed in the current drying stage; the humidity maintenance boundary includes the range of the target humidity value and the range of humidity maintenance duration; the dehumidification intensity boundary includes the range of dehumidification valve opening degree, the range of dehumidification duration, and the range of dehumidification frequency; and the energy consumption constraint boundary includes the upper limit of the cumulative energy consumption allowed in the current drying stage and the lower limit of the unit energy consumption dehydration contribution value.

[0097] The basic temperature target value, basic temperature rise slope, basic humidity target value, basic dehumidification valve opening degree, and basic energy consumption upper limit are derived from the historical qualified drying process of the same model product, enterprise process documents, trial production verification records, or historical crack-free batch data; historical drying defect data are used to correct the boundary of the corresponding structural weak area. When the historical crack frequency or risk weight of the same structural weak area increases, the system reduces the upper limit of the temperature rise slope in that stage, increases the lower limit of humidity holding time, and limits the sudden change in dehumidification intensity.

[0098] During the safe migration phase, if the unit energy consumption dehydration contribution value is higher than the preset effective dehydration threshold, the system retains the space for energy-saving acceleration; during the risk approach phase, the system reduces the upper limit of the temperature rise slope according to the risk intensity; during the weak zone passage phase, the system increases the humidity maintenance constraint and reduces the change range of the dehumidification valve opening.

[0099] The upper limit of the heating slope is corrected according to the following formula:

[0100] in, This represents the upper limit of the heating slope of the i-th structural weak zone at the t-th sampling time, corresponding to the corresponding stage. This indicates the lower limit of the preset heating slope. Indicates the slope of the basic temperature rise. This indicates the normalized frontier collision risk intensity. This represents the correction factor for the risk intensity relative to the upper limit of the temperature rise slope.

[0101] Basic heating slope The preset heating rate is 0.8 K / min, with a lower limit for the preset heating slope. The risk intensity is 0.1 K / min, which is the correction factor for the upper limit of the heating slope. The value is 0.6.

[0102] The energy consumption constraint boundary is corrected according to the following formula:

[0103] in, This represents the energy consumption constraint boundary under the t-th sampling time identifier. This indicates the upper limit of basic energy consumption for the corresponding drying stage. This represents the unit energy consumption dehydration contribution value under the t-th sampling time marker. This indicates the lower limit of the preset unit energy consumption dehydration contribution value. This represents the correction coefficient for the unit energy consumption dehydration contribution value to the energy consumption constraint boundary.

[0104] Basic energy consumption limit The preset energy consumption dehydration contribution value is 50kWh, with a lower limit. The value is 2.0 g / kWh, and the correction coefficient μ for the unit energy consumption dehydration contribution value to the energy consumption constraint boundary is 1.5.

[0105] For example, if the basic temperature rise slope of a certain product of the same model is 0.8K / min, and the risk intensity corresponding to the root of the second umbrella skirt is relatively high, the system will correct the upper limit of the temperature rise slope of this stage to 0.5K / min and increase the lower limit of humidity holding time from 10min to 18min. The results will be written into the stage boundary data.

[0106] S430. Combine the stage boundary data, risk area identification, risk intensity, and risk cause type to generate a parameter generation constraint set. The parameter generation constraint set includes the allowable temperature target range, humidity target range, temperature rise slope range, humidity holding time range, dehumidification window range, and energy consumption upper limit range for different drying stages, and serves as the constraint condition for the subsequent generation of temperature and humidity coordinated control parameter sequence.

[0107] Specifically, the system establishes constraint records according to stage identifiers. Each constraint record includes at least the stage identifier, risk area identifier, risk occurrence stage, risk intensity, risk cause type, target temperature range, target humidity range, temperature rise slope range, humidity holding duration range, dehumidification window range, dehumidification intensity range, upper limit of energy consumption range, lower limit of dehydration contribution value per unit of energy consumption, constraint priority, and candidate parameter search flag. When there is a conflict between defect suppression constraints and energy-saving constraints, the system prioritizes satisfying the leading-edge collision risk threshold, historical drying defect risk requirements, and the safety boundary of the weak zone passage stage, and then selects parameters with lower energy consumption within the parameter range that meets the above requirements.

[0108] Constraint priority is determined by the following formula:

[0109] in, This indicates the constraint priority of the i-th structural weak zone at the t-th sampling time, corresponding to the stage. This indicates the normalized frontier collision risk intensity. This represents the risk weight of the i-th candidate weak region. This represents the normalized unit energy consumption contribution to dehydration. , , These represent the priority coefficients of the corresponding data items.

[0110] The value range of the priority coefficient is: It is 0.4-0.5. It is 0.3-0.4. The value is between 0.1 and 0.2. In this embodiment, the specific value is: =0.45, =0.35, =0.20. This is determined by assigning a forward collision risk intensity... Assign the highest priority coefficient to ensure that the system performs energy consumption optimization calculations under safe conditions.

[0111] The higher the constraint priority, the earlier the constraint will be satisfied when S510 uses the wolf pack algorithm to generate the candidate parameter sequence.

[0112] For example, if the root of the second umbrella skirt is in the weak zone passage stage, and the risk causes are too fast migration speed and continuous moisture accumulation, then the system writes the temperature rise slope range, humidity holding time range, dehumidification valve opening change range and energy consumption upper limit range into the parameter generation constraint set, and sets the constraint priority of this stage to high; the parameter generation constraint set then serves as the search boundary of the candidate parameter sequence group in S510.

[0113] S5 specifically includes the following sub-steps: S510. Using the parameter generation constraint set as the search boundary, the phased temperature target value, humidity target value, temperature rise slope, humidity holding time, dehumidification window, dehumidification intensity, and drying endpoint determination parameters are combined into a candidate parameter sequence. Each candidate parameter sequence is used as a candidate individual in the wolf pack algorithm to obtain a candidate parameter sequence group. The wolf pack algorithm is used to search for candidate parameter sequences that take into account both defect suppression and energy saving objectives within the range limited by the parameter generation constraint set.

[0114] Specifically, the candidate parameter sequence refers to the combination of temperature and humidity coordinated parameters generated for the safe migration stage, the risk approach stage, and the vulnerable zone passage stage, respectively. Its values ​​are derived from the parameter generation constraint set generated in S430 and must not exceed the target temperature range, target humidity range, temperature rise slope range, humidity holding time range, dehumidification window range, dehumidification intensity range, and energy consumption upper limit range. Dehumidification intensity refers to the parameterized expression of the dehumidification valve opening, dehumidification duration, or dehumidification frequency generated by the parameter generation constraint set, which differs from the dehumidification load data collected in S210.

[0115] The candidate parameter sequence is represented by the following formula:

[0116] in, Let q be the sequence of candidate parameters. Indicates the target temperature value. Indicates the target humidity value. Indicates the slope of the temperature rise. Indicates the duration of humidity retention. Indicates the dehumidification window. Indicates the dehumidification intensity. q represents the drying endpoint determination parameter, and q represents the candidate parameter sequence number.

[0117] In the wolf pack algorithm, candidate individuals are sequences of candidate parameters that can be evaluated and iteratively adjusted within the parameter generation constraint set; the alpha wolf parameter sequence is the candidate parameter sequence with the highest current comprehensive evaluation result and that does not violate constraints; the scout wolf parameter sequence is a candidate parameter sequence that is locally searched near the alpha wolf parameter sequence. The system first generates candidate parameters for the weak zone traversal stage according to constraint priority, and then generates candidate parameters for the risk approach stage and the safe migration stage.

[0118] For example, if the parameter generation constraint set of the weak area at the root of the second umbrella skirt limits the temperature rise slope to 0.3K / min to 0.5K / min, the humidity holding time to 15min to 25min, and the change in the opening of the dehumidification valve to no more than 10%, then the temperature rise slope, humidity holding time, and dehumidification intensity in the corresponding candidate parameter sequence must all be within the above range, and parameter values ​​exceeding the temperature rise slope range must not be generated.

[0119] S520. Input the candidate parameter sequence group into the drying front migration model, calculate the reduction in front collision risk, reduction in unit energy consumption, and drying endpoint stability corresponding to each candidate parameter sequence, and evaluate each candidate parameter sequence through a reinforcement learning parameter evaluator. The reinforcement learning parameter evaluator takes the drying state feature set as the state input, the candidate parameter sequence as the action input, and the reduction in front collision risk, reduction in unit energy consumption, and reduction in historical drying defect risk as evaluation feedback, and outputs the comprehensive evaluation result of the candidate parameter sequence.

[0120] Specifically, the reinforcement learning parameter evaluator adopts a deep network structure, consisting of one input layer, three hidden layers, and one output layer; The number of neurons in the input layer is consistent with the dimension of the state input. The hidden layers all use the ReLU activation function. The number of neurons in the three hidden layers are 128, 64 and 32 respectively. The output layer outputs the Q value or policy probability value corresponding to each action.

[0121] Its state inputs include the dry state feature set generated by S230, the front collision risk result generated by S330, and the parameter generation constraint set generated by S430, and its action inputs are the candidate parameter sequence generated by S510.

[0122] The reduction in frontal collision risk refers to the difference in frontal collision risk intensity before and after adopting the candidate parameter sequence; the reduction in unit energy consumption refers to the decrease in the expected unit energy consumption corresponding to the candidate parameter sequence relative to the basic drying process; the reduction in historical drying defect risk refers to the contribution of the candidate parameter sequence to the reduction of the risk weight corresponding to the historical high-risk structural weak area; the stability of the drying endpoint refers to the degree to which the rate of mass change, the rate of humidity decline, and the endpoint moisture content simultaneously meet the requirements for determining the drying endpoint under the candidate parameter sequence.

[0123] The comprehensive evaluation result is calculated according to the following formula:

[0124] in, This represents the comprehensive evaluation result of the q-th candidate parameter sequence. This represents the reduction in frontier collision risk corresponding to the q-th candidate parameter sequence. This indicates the reduction in energy consumption per unit. This indicates the reduction in the risk of historical drying defects. Indicates the stability at the drying endpoint. Indicates the amount of constraint violation. , , , , These represent the weight coefficients of the corresponding evaluation items.

[0125] The range of values ​​for the weighting coefficients of the comprehensive evaluation results is as follows: It is 0.3-0.4. It is 0.2-0.3. It is 0.2-0.3. It is 0.1-0.2. The value is between 0.5 and 0.8. In this embodiment, the specific value is: =0.30, =0.25, =0.20, =0.15, =0.60.

[0126] Constraint Violation Quantity The calculation method is as follows: when any parameter exceeds the range defined by the parameter generation constraint set, the absolute value of the excess is multiplied by a penalty coefficient of 100 and then accumulated; if it does not exceed the constraint set, then... Set it to 0. This is achieved by setting a higher penalty coefficient. This forces reinforcement learning models to avoid unsafe and high-energy-consuming boundaries during training.

[0127] If the candidate parameter sequence does not exceed the parameter generation constraint set, the constraint violation amount is 0; if the heating slope, humidity holding time or dehumidification intensity exceeds the constraint range, the constraint violation amount is increased according to the extent of the excess.

[0128] For example, if a candidate parameter sequence sets the heating slope of the second umbrella skirt root passage stage to 0.45K / min and the humidity holding time to 20min, after calculation by the drying front migration model, the front collision risk intensity decreases and the parameter generation constraint set is not violated, then the reinforcement learning parameter evaluator outputs a comprehensive evaluation result that meets the preset high evaluation level.

[0129] S530. Based on the comprehensive evaluation results, the wolf pack algorithm is driven to iteratively update the candidate parameter sequence group, and the optimal candidate parameter sequence that meets the requirements of front collision risk threshold, energy consumption constraint boundary and drying endpoint determination is selected to generate temperature and humidity coordinated control parameter sequence. The temperature and humidity coordinated control parameter sequence is used to reduce the migration mutation risk when the drying front passes through the structurally weak area, and reduce the drying energy consumption under the condition of meeting the defect suppression requirements.

[0130] Specifically, the system sorts the candidate parameter sequences according to the comprehensive evaluation results, and determines the candidate parameter sequence with the highest comprehensive evaluation result and zero constraint violation as the current alpha parameter sequence. The remaining candidate parameter sequences are then iteratively updated based on the current alpha parameter sequence and constraint priority. If the reduction in frontal collision risk of a candidate parameter sequence is insufficient, the system prioritizes reducing the temperature rise slope of the corresponding stage, increasing the humidity holding time, or reducing the abrupt change in dehumidification intensity. If the frontal collision risk meets the threshold but the reduction in unit energy consumption is insufficient, the system shortens the low-contribution humidity holding time, adjusts the dehumidification window, or reduces the ineffective dehumidification intensity within the allowable range of the parameter generation constraint set.

[0131] The optimal candidate parameter sequence is determined by the following formula:

[0132] in, Represents the optimal candidate parameter sequence. This represents the set of candidate parameter sequences that meet the requirements of the leading-edge collision risk threshold, energy consumption constraint boundary, and drying endpoint determination. This represents the comprehensive evaluation result of the q-th candidate parameter sequence.

[0133] The system stops iterating when the number of iterations reaches a preset number, the improvement in the comprehensive evaluation result of multiple consecutive iterations is less than a preset improvement threshold, or the candidate parameter sequence has simultaneously met the frontier collision risk threshold, energy consumption constraint boundary, and drying endpoint determination requirements. The final output temperature and humidity coordinated control parameter sequence includes at least the stage identifier, temperature target value, humidity target value, temperature rise slope, humidity holding duration, dehumidification window, dehumidification intensity, drying endpoint determination parameters, and parameter execution order.

[0134] For example, if the overall evaluation result improves by less than 1% after five consecutive iterations, and the current candidate parameter sequence has reduced the collision risk intensity at the leading edge of the second umbrella skirt root to below the risk threshold, while the unit energy consumption meets the energy consumption constraint boundary, then the system stops iterating, and this candidate parameter sequence is determined as the temperature and humidity coordinated control parameter sequence. The temperature and humidity coordinated control parameter sequence can be used as the execution parameters for the current drying batch, or it can be fed back into the parameter generation process of the next batch along with subsequent historical drying defect data of this batch, for continuously correcting the risk weight of the structural weak area, the safe migration threshold, and the parameter generation constraint set.

[0135] All the above formulas are performed using dimensionless numerical calculations; the relevant formulas are based on empirical models that approximate the real situation, obtained through extensive data collection and software simulation fitting. The preset parameters and thresholds involved in the formulas can be conventionally set and adjusted by those skilled in the art according to the physical constraints of the actual application scenario.

[0136] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0137] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0138] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An energy-saving electric porcelain insulator drying temperature and humidity coordinated control method, characterized in that, Includes the following steps: S1. Read the three-dimensional structural data, wall thickness distribution data, skirt spacing data, initial moisture content data and loading position data of the porcelain insulator to be dried, establish a structural region model, identify candidate weak areas, and generate a set of structural weak areas and a structural weak area index table by combining historical drying defect data. S2. Collect temperature response data, humidity response data, mass change data, dehumidification load data and cumulative energy consumption data of the initial drying stage, generate the initial drying stage response data sequence, associate it with the structural weak area index table, and calculate the drying state feature set. S3. Based on the structural region model, initial moisture content data, and dry state feature set, establish a dry front migration model, determine the current position, migration speed, migration direction, and expected traversal area of ​​the front, and match it with the set of structurally weak areas to generate front collision risk results; S4. Based on the results of the frontier collision risk, divide the safe migration stage, the risk approach stage, and the weak zone passage stage. Combine the unit energy consumption dehydration contribution value and historical drying defect data to determine the stage boundary data and generate a parameter generation constraint set.

2. The energy-saving oriented drying temperature and humidity co-control method for the electric porcelain insulator according to claim 1, characterized in that, Also includes: S5. Generate a candidate parameter sequence group using the parameter generation constraint set as the search boundary. Select the optimal candidate parameter sequence through the wolf pack algorithm and the reinforcement learning parameter evaluator to generate the temperature and humidity coordinated control parameter sequence.

3. The energy-saving oriented drying temperature and humidity co-control method for the electric porcelain insulator according to claim 1, characterized in that, S1 specifically includes: Read the three-dimensional structural data, wall thickness distribution data, shed spacing data, porcelain head transition zone data, initial moisture content data, and loading position data of the porcelain insulator to be dried. Divide the structural regions according to axial height, radial thickness, and shed layer and configure region identifiers to generate a structural region model. Based on the structural region model, the range of wall thickness variation, the range of thickness transition, the degree of bending at the base of the umbrella skirt, and the degree of transition of the ceramic head connection are calculated. Structural regions that exceed the corresponding thresholds are marked as candidate weak regions, and a set of candidate weak regions is generated.

4. The energy-saving oriented drying temperature and humidity co-control method for the electric porcelain insulator according to claim 3, characterized in that, It also includes: matching the candidate weak area set with historical drying defect data, using a weighted calculation method to assign risk weights based on the frequency of defect occurrence, the severity of defect, and the degree of matching of defect location, and generating a set of structural weak areas and a structural weak area index table.

5. The energy-saving oriented drying temperature and humidity co-control method for the electric porcelain insulator according to claim 1, characterized in that, S2 specifically includes: After the porcelain insulators to be dried enter the initial drying stage, temperature response data, humidity response data, mass change data, dehumidification load data, and cumulative energy consumption data are collected, and a drying initial stage response data sequence is generated according to a unified sampling time identifier. The initial drying stage response data sequence is correlated with the index table of structural weak areas in time and space. The temperature and humidity response values ​​corresponding to each structural weak area are determined based on spatial distance weighting. The heating sequence, dehumidification path and moisture accumulation location corresponding to each structural weak area are determined according to the loading location data, and structural weak area response correlation data are generated. Based on the response correlation data of the structurally weak area, the rate of mass change, the rate of humidity decline, the degree of temperature lag, the degree of change of dehumidification load, and the contribution value of dehydration per unit energy consumption are calculated to generate a dry state feature set.

6. The energy-saving method for coordinated temperature and humidity control of porcelain insulators, as described in claim 1, is characterized in that... S3 specifically includes: A drying front migration model is established based on the structural region model, initial moisture content data, and drying state feature set. The front state value of each structural region is calculated by weighting based on the state data to determine the current position of the front and the local moisture content change trend. Based on the current position of the front, the rate of mass change, and the rate of humidity decline under continuous sampling time markers, the migration speed, migration direction, and expected area of ​​the drying front are predicted, generating the drying front migration prediction results. The predicted migration results of the drying front are matched with the set of structurally weak areas. When the predicted area overlaps with the structurally weak area and the migration speed exceeds the safe migration threshold, the front collision risk results are generated, including the risk area identification, the stage of risk occurrence, the risk intensity calculated by weight, and the risk cause type.

7. The energy-saving method for coordinated temperature and humidity control of porcelain insulators, as described in claim 1, is characterized in that... S4 specifically includes: Based on the risk area identification, risk intensity, and risk cause type in the frontal collision risk results, and combined with the frontal approach distance between the current frontal position and the structural weak area, the drying process is divided into a safe migration stage, a risk approach stage, and a weak area passage stage, thus obtaining the drying stage division results. By combining the results of the drying stage division, the unit energy consumption dehydration contribution value, and historical drying defect data, the temperature rise boundary, humidity maintenance boundary, moisture removal intensity boundary, and energy consumption constraint boundary of each stage are determined, and the stage boundary data are obtained.

8. The energy-saving method for coordinated temperature and humidity control of porcelain insulators, as described in claim 7, is characterized in that... Also includes: The stage boundary data is combined with risk area identifiers, risk intensity, and risk cause types to generate a parameter-generated constraint set, which includes temperature target range, humidity target range, temperature rise slope range, humidity holding duration range, dehumidification window range, dehumidification intensity range, energy consumption upper limit range, and constraint priority.

9. The energy-saving method for coordinated temperature and humidity control of porcelain insulators, as described in claim 2, is characterized in that... S5 specifically includes: Using the parameter generation constraint set as the search boundary, the target temperature value, target humidity value, temperature rise slope, humidity holding time, dehumidification window, dehumidification intensity and drying endpoint determination parameter are combined into a candidate parameter sequence, and this sequence is used as a candidate individual in the wolf pack algorithm to obtain a candidate parameter sequence group. The candidate parameter sequence group is input into the drying front migration model to calculate the reduction in front collision risk, the reduction in unit energy consumption, the reduction in historical drying defect risk, and the stability at the drying endpoint. The reinforcement learning parameter evaluator uses the drying state feature set as the state input, the candidate parameter sequence as the action input, and incorporates the penalty calculation for the constraint violation of the parameter generation constraint set by the candidate parameter sequence to output a comprehensive evaluation result.

10. The energy-saving method for coordinated temperature and humidity control of porcelain insulators, as described in claim 9, is characterized in that... Also includes: Based on the comprehensive evaluation results, the candidate parameter sequence group is iteratively updated, and the optimal candidate parameter sequence that meets the requirements of frontier collision risk threshold, energy consumption constraint boundary and drying endpoint determination is selected to generate temperature and humidity coordinated control parameter sequence.

Images