Water content prediction model parameter adjustment method and device and nonvolatile storage medium

Abstract

The invention discloses a water content prediction model parameter adjustment method and device and a nonvolatile storage medium. The method comprises the steps that cloud equipment sends a knowledge base to an optical line terminal, the knowledge base comprises a model parameter set of a water content prediction model corresponding to historical working conditions, and the water content prediction model is arranged in the optical line terminal; the equipment working condition of the target drying equipment sent by the optical line terminal through the uplink and a model parameter adjustment mode corresponding to the equipment working condition are received, and the model parameter adjustment mode is an adjustment mode determined by the optical line terminal according to the difference value between the water content prediction value and the water content measured value of the water content prediction model; and updating the model parameter set corresponding to the historical working condition in the knowledge base according to the equipment working condition and the model parameter adjustment mode corresponding to the equipment working condition. The technical problem of relatively high response delay caused by setting the model at the cloud end in related technologies is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of machine learning, and in particular, to a water content prediction model parameter adjustment method and device and a nonvolatile storage medium. BACKGROUND

[0002] In related technologies, a model is usually set in a cloud device, and the cloud device controls a device at an edge end to perform a corresponding action according to a model prediction result. However, this approach has the problem of high response delay.

[0003] To address the above problems, no effective solutions have been proposed so far. SUMMARY

[0004] Embodiments of the present application provide a water content prediction model parameter adjustment method and device and a nonvolatile storage medium to at least solve the technical problem of high response delay caused by setting a model in a cloud in related technologies.

[0005] According to an aspect of an embodiment of the present application, a water content prediction model parameter adjustment method is provided, including: a cloud device sending a knowledge base to an optical line terminal, wherein the knowledge base includes historical working conditions of a plurality of target drying devices and a model parameter set of a water content prediction model corresponding to the historical working conditions, the water content prediction model being set in the optical line terminal and used to predict the water content of materials in the target drying devices; receiving device working conditions of the target drying devices and a model parameter adjustment mode corresponding to the device working conditions sent by the optical line terminal through an uplink, wherein the model parameter adjustment mode is an adjustment mode determined by the optical line terminal according to a difference between a water content prediction value of the water content prediction model and a water content measured value; and updating the model parameter set corresponding to the historical working conditions in the knowledge base according to the device working conditions and the model parameter adjustment mode corresponding to the device working conditions.

[0006] Optionally, the updating of the model parameter set corresponding to the historical working condition in the knowledge base according to the device working condition and the model parameter adjustment mode corresponding to the device working condition comprises: determining a similarity between the device working condition and the historical device working condition in the knowledge base, wherein the knowledge base comprises a plurality of historical working conditions and a model parameter set of the water content prediction model corresponding to the historical working condition; in a case where there is a first historical device working condition corresponding to a similarity not less than a first preset similarity threshold in the knowledge base, adjusting the model parameter set corresponding to the first historical device working condition according to the model parameter adjustment mode; in a case where there is no first historical device working condition in the knowledge base and there is a second historical device working condition corresponding to a similarity less than the first preset similarity threshold and not less than a second preset similarity threshold, adjusting the model parameter set corresponding to the second historical device working condition according to the model parameter adjustment mode; and in a case where there is no first historical device working condition and no second historical device working condition corresponding in the knowledge base, determining the model parameter set corresponding to the device working condition according to the model parameter adjustment mode, and storing the device working condition and the model parameter set corresponding to the device working condition in the knowledge base.

[0007] Optionally, the cloud device determines the similarity between the device working condition and the historical device working condition in the knowledge base comprises: determining a discretized parameter level label corresponding to each working condition parameter in the device working condition, wherein the discretized parameter level label is used to indicate a preset parameter value range in which the parameter value of the working condition parameter is located, and the discretized parameter level label corresponding to each working condition parameter is determined by the optical line terminal according to the preset parameter value range in which the parameter value of the working condition parameter is located; determining a candidate historical device working condition in the knowledge base according to the discretized level label corresponding to the working condition parameter; and determining the similarity between the device working condition and the candidate historical device working condition according to the parameter values of each working condition parameter in the device working condition and the parameter values between each working condition parameter in the candidate historical device working condition.

[0008] Optionally, the determination of the similarity between the device working condition and the candidate historical device working condition according to the parameter values of each working condition parameter in the device working condition and the parameter values between each working condition parameter in the candidate historical device working condition comprises: determining a parameter type of the working condition parameter and a weight coefficient corresponding to the parameter type, wherein the weight coefficient is used to reflect the influence degree of the parameter type on the water content of the material, and the greater the weight coefficient, the greater the influence degree; and determining the similarity between the device working condition and the candidate historical device working condition according to the weight coefficient corresponding to the parameter type, the parameter values of each working condition parameter in the device working condition and the parameter values between each working condition parameter in the candidate historical device working condition.

[0009] Optionally, the determining the similarity between the device working condition and the historical device working condition in the knowledge base comprises: determining a discretized parameter level label corresponding to each working condition parameter in the historical device working condition, wherein the discretized parameter level label is used to indicate a preset parameter value range in which a parameter value of the working condition parameter is located; determining a value at a center point of the preset parameter value range as the parameter value of the working condition parameter in the historical device working condition; and determining the similarity between the device working condition and the historical device working condition according to the parameter value of the working condition parameter in the device working condition and the parameter value of the working condition parameter in the historical device working condition.

[0010] Optionally, before adjusting the model parameter set corresponding to the second historical device working condition according to the model parameter adjustment mode, the method further comprises: determining a discretized parameter level label corresponding to each working condition parameter in the second historical device working condition, wherein the discretized parameter level label is used to indicate a preset parameter value range in which a parameter value of the working condition parameter is located; determining a preset parameter value range corresponding to each working condition parameter in the device working condition according to the parameter type; in a case where a parameter value of a target working condition parameter in the device working condition is not in the corresponding preset parameter value range, determining that there is no second historical device working condition corresponding to the device working condition, wherein the target working condition parameter is a working condition parameter of the preset parameter type; in a case where there are more than a preset number of non-target working condition parameters in the device working condition, the parameter value of which is not in the corresponding preset parameter value range, determining that there is no second historical device working condition corresponding to the device working condition, wherein the non-target working condition parameter is a working condition parameter of a parameter type other than the preset parameter type, and the target working condition parameter has a greater influence on the moisture content of the material than the non-target working condition parameter; in a case where there is a non-target working condition parameter in the device working condition, the parameter value of which is not in the corresponding preset parameter range and the deviation amplitude between which and the corresponding preset parameter range is greater than a preset deviation amplitude threshold, determining that there is no second historical device working condition corresponding to the device working condition.

[0011] Optionally, the method further comprises: the optical line terminal sending the model parameter adjustment mode to the optical network unit through a downlink, wherein the model parameter adjustment mode is used to determine the device parameter adjustment mode of the target drying device.

[0012] Optionally, the method further comprises: the optical line terminal acquiring the device working condition collected by each sensor through the optical network unit, wherein the sensor and the optical network unit are directly connected.

[0013] According to another aspect of the embodiments of the present application, a water content prediction model parameter adjustment device is also provided, which comprises: a first processing module configured to send a knowledge base to an optical line terminal, wherein the knowledge base comprises historical working conditions of a plurality of target drying devices and a model parameter set of a water content prediction model corresponding to the historical working conditions, the water content prediction model being arranged in the optical line terminal and configured to predict the water content of materials in the target drying device; a second processing module configured to receive device working conditions of the target drying device and a model parameter adjustment mode corresponding to the device working conditions sent by the optical line terminal through an uplink, wherein the model parameter adjustment mode is an adjustment mode determined by the optical line terminal according to a difference between a water content prediction value and a water content measured value of the water content prediction model; and a third processing module configured to update the model parameter set corresponding to the historical working conditions in the knowledge base according to the device working conditions and the model parameter adjustment mode corresponding to the device working conditions.

[0014] According to another aspect of the embodiments of the present application, a non-volatile storage medium is also provided, which stores a program, wherein the program performs the water content prediction model parameter adjustment method when executed.

[0015] According to another aspect of the embodiments of the present application, an electronic device is also provided, which comprises a memory and a processor, the processor being configured to execute a program stored in the memory, wherein the program performs the water content prediction model parameter adjustment method when executed.

[0016] According to another aspect of the embodiments of the present application, a computer program product is also provided, which comprises a computer program, the computer program implementing the water content prediction model parameter adjustment method when executed by a processor.

[0017] In the embodiments of the present application, the cloud device is used to send a knowledge base to an optical line terminal, wherein the knowledge base comprises historical working conditions of a plurality of target drying devices and a model parameter set of a water content prediction model corresponding to the historical working conditions, the water content prediction model being arranged in the optical line terminal and configured to predict the water content of materials in the target drying device; the device working conditions of the target drying device and a model parameter adjustment mode corresponding to the device working conditions sent by the optical line terminal through an uplink are received; the model parameter set corresponding to the historical working conditions in the knowledge base is updated according to the device working conditions and the model parameter adjustment mode corresponding to the device working conditions; the model is deployed to the optical line terminal on the edge side, and the optical line terminal determines the model parameter adjustment mode, so as to reduce the response delay, thereby achieving the technical effect of responding in time when the device working conditions and the like change, and further solving the technical problem of high response delay caused by arranging the model in the cloud according to the related art. BRIEF DESCRIPTION OF DRAWINGS

[0018] The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the application and together with the description serve to explain the application. In the drawings:

[0019] Figure 1 is a structural schematic diagram of a drying equipment control system in the related art provided by an embodiment of the application;

[0020] Figure 2 is a structural schematic diagram of a drying equipment control system provided by an embodiment of the application;

[0021] Figure 3 is a flow schematic diagram of a water content prediction model parameter adjustment method provided by an embodiment of the application;

[0022] Figure 4 is a flow schematic diagram of a drying equipment control flow provided by an embodiment of the application;

[0023] Figure 5 is a structural schematic diagram of a water content prediction model parameter adjustment device provided by an embodiment of the application. DETAILED DESCRIPTION

[0024] In order to enable persons skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by persons skilled in the art without creative work should fall within the protection scope of the present application.

[0025] It should be noted that the terms "first", "second", and the like in the specification and claims of the present application and the above-described drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or a chronological sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes a series of steps or units does not necessarily have to include only those steps or units clearly listed, but can include other steps or units that are not clearly listed or inherent to the process, method, product or device.

[0026] In order to better understand the embodiments of the present application, the technical terms involved in the embodiments of the present application are explained as follows:

[0027] PON: PON is the English name of Passive Optical Network, and the Chinese name is Passive Optical Network.

[0028] OLT: English name is optical line terminal, and the Chinese name is optical line terminal.

[0029] ONU: English name is Optical Network Unit, and the Chinese name is Optical Network Unit.

[0030] In the production process of daily-use porcelain in the related art, the judgment of the drying state of the green body depends on the following method:

[0031] The temperature and humidity sensors are installed in the drying room, and the temperature / humidity sensors are connected to the Modbus RTU gateway through the RS485 bus (actual transmission rate 19.2 kbps-115.2 kbps), and then connected to the PLC. A line of experienced teachers observes the returned data on site; when the temperature is observed to rise to a certain interval, the teacher uses a tool to take out the green body, and according to the touch in the hand, the temperature and moisture content of the green body are judged by experience. The existing technology has the following key defects:

[0032] In terms of process mechanism, manual sampling is used in the drying process, which can easily interrupt the drying process and destroy the continuity of drying; at the same time, since the sampling breaks the continuity of drying, it is impossible to dynamically perceive the change of green body moisture content in the whole drying process. In terms of green body temperature control deviation, the temperature of the drying room environment ≠ the actual temperature of the green body, which has an error, which makes the temperature control of the green body distorted. And the related technology also has the problem of missing drying data dimension, which does not synchronize the monitoring of green body moisture content (W), green body surface temperature (Tsurface), green body weight (M), drying room average temperature (Toven) and outdoor meteorological parameters (humidity Hout), which cannot construct a multi-factor driven model, and the process control relies on manual experience, and the yield rate fluctuates greatly.

[0033] In terms of network architecture, in the related art, a large number of temperature / humidity sensors are connected to the Modbus RTU gateway through the RS485 bus, and then connected to the PLC, and the networking is realized through the industrial Ethernet switch. If according to the traditional network architecture, the mixed perception scheme adopted in the present scheme, i.e. microwave moisture meter, infrared temperature measuring instrument, high-temperature weighing instrument, outdoor weather station, and a large number of laid drying room temperature / humidity sensors, all need to pass through the Modbus RTU gateway to complete the protocol analysis, and are connected to the PLC, and the data forwarding between multiple PLCs is realized through the industrial switch, but this way has obvious defects:

[0034] ①Microwave moisture meter (sampling interval 1 minute), infrared temperature measuring instrument (sampling interval 10 seconds), weighing sensor (sampling interval 1 minute) and other heterogeneous devices are difficult to realize millisecond-level synchronous transmission of multi-source heterogeneous data (moisture content, temperature, weight, meteorological parameters) due to network time delay differences, resulting in poor spatio-temporal alignment of data and restricting the accuracy of dynamic modeling.

[0035] ②The traditional network architecture adopts a multi-hop transmission mode of "sensing layer→Modbus gateway→PLC→industrial switch→cloud", which causes:

[0036] End-to-end latency deterioration: Microwave moisture meter data needs to go through 4 layers of protocol conversion from collection to PLC processing: physical layer (RS485)→data link layer (Modbus RTU)→application layer (Modbus TCP)→cloud protocol (OPC UA), and the end-to-end latency is 500-800ms, and the moisture content control instruction lags behind 3-5 sampling periods;

[0037] Lack of edge computing capability: Lack of edge computing capability, raw data needs to be uploaded to the cloud for processing, high response delay (seconds), cannot meet the real-time modeling and closed-loop control requirements.

[0038] In addition, in terms of network scalability and security, there are certain problems in the related art, which hinder the upgrading of related processes. For example, densely arranged RS485 cables are prone to electromagnetic interference in high temperature and high humidity environments, resulting in distorted collected data.

[0039] And the model for predicting the moisture content of the material in the drying room in the related art is usually set in the cloud device, and the edge device is controlled to perform corresponding actions according to the model prediction result, which has a high response delay. It will lead to the inability to quickly adjust the drying room equipment according to the actual situation.

[0040] In order to solve the above problems, the related solutions in the embodiments of the present application are provided, which are described in detail below.

[0041] According to the embodiments of the present application, a method embodiment of a moisture content prediction model parameter adjustment method is provided. It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a group of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0042] The method embodiment provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device.

[0043] In some embodiments of the present application, the structure of the drying equipment control system in the related art is as followsFigure 1 As shown in the prior art, most of the RS485 bus and the industrial Ethernet multi-hop transmission architecture (sensing layer Modbus gateway PLC industrial switch cloud) network exist high end-to-end delay (500-800ms), poor data synchronization and other problems. The drying equipment control system in the embodiment of the present application is as shown in the prior art Figure 2 As shown in the prior art, based on the industrial passive optical network (PON) technology, the drying room data transmission network is constructed, the XGS-PON standard (supporting symmetric 10Gbps bandwidth, end-to-end delay ≤1ms, or delay in uplink and downlink) is adopted, the distributed optical network unit (ONU) is used to realize the convergence and transmission of multi-source heterogeneous sensor data, the network architecture is divided into three layers, including the sensing layer, the access layer and the transmission layer.

[0044] As can be seen from Figure 2 In the sensing layer, the microwave moisture meter, the infrared temperature meter, the weighing sensor and the drying room thermocouple are connected to the ONU through the RS485 interface, and are directly connected to the PON network through the Modbus RTU driver of the ONU, thereby reducing the traditional gateway level. The microwave moisture meter, the infrared temperature meter and other sensors can send Modbus RTU data frames (including register address, data value and CRC check code) through the RS485 interface.

[0045] In addition, in the system provided in the embodiment of the present application, the parsed data frame (including timestamp, sensor ID and data value) is directly encapsulated into the industrial PON protocol format, the ISO 8601 extended timestamp (accuracy ≤100μs) is embedded in the header, and the CRC-16 check code is appended to the tail. Therefore, the protocol conversion step (Modbus RTU Modbus TCP OPC UA) in the traditional multi-hop transmission is omitted, and the protocol stack level and data encapsulation overhead are reduced.

[0046] In addition, the ONU is built-in Modbus RTU driver, which can real-time analyze sensor register data (such as moisture content address 0x1001, temperature address 0x2001), and extract effective parameters (material moisture content, material surface temperature, material weight, etc.). The technical parameters, sampling methods and data acquisition and transmission methods of each sensing module in the sensing layer are as shown in the following table:

[0047]

[0048]

[0049] For the access layer, the ONU of the access layer is built-in Modbus RTU driver, which can real-time analyze sensor register data (such as moisture content address 0x1001, temperature address 0x2001), and is connected to the optical line terminal (OLT) through a single-mode optical fiber.

[0050] For the transmission layer, the OLT can upload the preprocessed data and the dynamic optimization parameters of the physical-data hybrid model to the local cloud platform through the XGS-PON technology, cooperate with the cloud computing power, realize the generation of control parameters, and then issue them to the drying room equipment (gas boiler, fan, air volume control valve).

[0051] In the system provided in the embodiments of the present application, a multi-source data synchronization mechanism is also provided, including timestamp generation and synchronization, dynamic delay correction, data frame format specification, timestamp storage and management, timestamp verification and fault tolerance mechanism, etc.

[0052] Optionally, in terms of timestamp generation and synchronization, the ONU can automatically embed a local timestamp in the data frame header when analyzing sensor data. The timestamp format is in the ISO 8601 extended format (including milliseconds and time zone information), for example: 2024-05-08T08:00:00.123+08:00, where:

[0053] YYYY-MM-DD: Date

[0054] T: Date and time separator

[0055] HH:MM:SS.sss: Time (hour: minute: second.millisecond)

[0056] ++08:00: Time zone (default time zone)

[0057] The timestamps of all sensor data on site can be stamped by the ONU (optical network unit). The ONU ensures that the ONU clock and the large network are synchronized through the NTP and PTPover PON protocols, with an accuracy of ≤100 μs (which can be improved to 100 ns as needed).

[0058] In terms of dynamic delay correction, the OLT can dynamically measure the transmission delay (RTT) based on the time slot scheduling algorithm and calculate the transmission path delay of each ONU. The specific correction formula is as follows:

[0059] Corrected timestamp = original timestamp (ONU side) + transmission delay (RTT)

[0060] Assuming that the original timestamp is 2024-05-08T08:00:00.000+08:00 and the transmission delay is 1 ms, the corrected timestamp is 2024-05-08T08:00:00.002+08:00.

[0061] In terms of data frame format specification, the ISO 8601 extended format timestamp (including milliseconds and time zone information) can be embedded in the data frame header, and the CRC-16 check code can be appended at the end, with the following format:

[0062]

[0063] In terms of storage and management of timestamps, the OLT built-in timing database (InfluxDB) stores timestamps and raw data from different ONUs by device ID, each timestamp being associated with a sensor ID, a data value, and a CRC check code. The storage period is 24 hours, which can support millisecond-level timing queries. A hash index can be constructed based on the device ID and time range to improve the retrieval efficiency of high-frequency data.

[0064] In terms of timestamp verification and fault tolerance mechanisms, the OLT performs the following timestamp verification procedures after receiving data frames:

[0065] Timing continuity check: Check whether the interval between adjacent data timestamps conforms to the sensor sampling period (e.g., 1 minute for a microwave moisture meter). If the deviation is > 10%, it is marked as abnormal.

[0066] CRC check code verification: Verify data integrity according to the CRC-16 check code at the end of the data frame. If the verification fails, trigger ONU retransmission.

[0067] Cross-device timestamp alignment: Synchronize the clocks of the OLT and the ONUs through the PTP protocol. If the timestamp of a certain ONU deviates from the OLT master clock by > 100 μs, automatically trigger the clock calibration instruction.

[0068] In terms of timestamp exception handling procedures, when a timestamp exception (e.g., deviation > 1 ms or verification failure) is detected, the OLT performs the following actions:

[0069] Abnormal data isolation: Temporarily store abnormal data in the repair buffer and trigger the data interpolation module of the edge computing container.

[0070] ONU clock calibration: Send a clock synchronization instruction through the PON downlink to force the ONU to align with the OLT master clock.

[0071] Log recording and alarm: Record abnormal events and push them to the cloud monitoring platform to prompt the operation and maintenance personnel to check the sensors or network links.

[0072] In some embodiments of the present application, the drying equipment control system can also align and store data. Optionally, the OLT layer stores raw data through the timing database (InfluxDB) and caches 24-hour data by device classification.

[0073] In addition, when building a single meter, the timestamps can be aligned at millisecond-level accuracy to avoid data misalignment caused by rounding errors. When abnormal data is detected, such as a timestamp deviation > 1 ms, an alarm is triggered and logs are recorded for manual calibration.

[0074] In some embodiments of the application, the OLT in the system also has the ability of containerized edge computing. Optionally, a lightweight container (Docker) can be built into the OLT to run the following application modules:

[0075] Data cleaning: Adopt sliding window median replacement method (window length 10 minutes), eliminate short-term sudden anomalies (such as temperature spikes);

[0076] Abnormal filtering: Based on μ±3σ threshold (mean ± 3 times standard deviation) to identify noise data, and replace with window median;

[0077] Missing value interpolation: Linear interpolation method is used to fill short-term missing data (1-3 time points), and long-term missing data is predicted by calling the heat conduction model;

[0078] Exponential smoothing: weight a = 0.3, generate smoothing parameters (W smoothed , T smoothed , M smoothed ), suppress random noise.

[0079] In addition, the OLT in the system can also generate a single meter at a minute granularity, and based on the single meter as a call source, generate a joint meter.

[0080] In some embodiments of the application, on the uplink of the system, the preprocessed data is uploaded to the cloud through XGS-PON technology (symmetric 10G bandwidth), the transmission delay is ≤1ms, the data format is unified as CSV / JSON, and the data is synchronized in real time in time sequence. The device state data of the drying equipment (such as fan speed, air supply temperature) can be real-time feedback to the OLT edge computing layer, the edge layer realizes data preprocessing, physical-data hybrid model parameter optimization, and uploads to the cloud model through the uplink, forming a "perception-modeling".

[0081] On the downlink of the system, some model parameters suitable for different working conditions stored locally by the cloud device can be cached to the OLT through the low-latency downlink (≤1ms) of the PON network, and then downloaded to the drying room equipment (gas boiler, fan, air source heat pump unit) by the OLT, forming a "real-time regulation" instruction transmission path: cloud→OLT→ONU→device controller, end-to-end delay ≤1ms.

[0082] In order to realize the adaptation to some old equipment, for sensors that only support second-level timestamps, the OLT layer automatically supplements the millisecond part (such as 2023-10-01 08:00:00→2023-1001T08:00:00.000+08:00). In addition, the OLT can be built-in Modbus RTU / TCP protocol conversion module to ensure seamless access of heterogeneous devices.

[0083] In some embodiments of the present application, the OLT built-in computing power board stores 24-hour raw data and pre-processed joint scales, and supports off-network continuous transmission. In addition, the computing power board can be deployed in the idle slot of the OLT device to realize the expansion of computing resources.

[0084] Under the above running environment, the embodiment of the present application provides a moisture content prediction model parameter adjustment method, which is suitable for a cloud device, as shown in the figure, the method comprises the following steps: Figure 3

[0085] Step S302, the cloud device sends a knowledge base to the optical line terminal, wherein the knowledge base includes historical working conditions of a plurality of target drying devices, and a model parameter set of a moisture content prediction model corresponding to the historical working conditions, the moisture content prediction model is set in the optical line terminal, and is used for predicting the moisture content of materials in the target drying device;

[0086] In some embodiments of the present application, the cloud device can also realize real-time screening of high-frequency running conditions and optimization of optimal control parameters based on the knowledge base set in the cloud, and synchronize the results to the OLT device computing board for data backup. When the network connection is interrupted, the OLT local redundant control strategy is automatically activated to ensure the low-latency call of key working condition parameters and the emergency control ability of the device.

[0087] Then, the OLT can dynamically select the global optimal control parameters issued by the cloud or the high-frequency working condition parameters (such as “medium weight + medium temperature” corresponding to K1=0.019) cached locally according to real-time working condition requirements, and issue them to the ONU (optical network unit) through the low-latency downlink (end-to-end delay ≤1ms) of the industrial PON network, and then the ONU parses and forwards them to the drying room equipment controller (such as a gas boiler or a variable frequency air blower), to complete the precise execution of the control instructions.

[0088] In some embodiments of the present application, the ONU can control the drying room equipment through the drying room equipment controller according to the adjustment result of the model parameters according to the following strategy:

[0089] First, the heat transfer efficiency coefficient (ΔK1) mapping:

[0090] Control object: gas boiler power or air source heat pump unit air supply temperature.

[0091] Adjustment logic: ΔK1 increases by 0.001→gas boiler power increases by 2% (or heat pump temperature +1℃).

[0092] Safety range:

[0093] K1∈(0.01,0.05), to ensure that the heat transfer efficiency is within the effective range.

[0094] ​Out-of-bound handling:

[0095] K1 out-of-bound: load the historical best (highest weighted score) under the same condition from the knowledge base.

[0096] Second, internal diffusion resistance coefficient (ΔK2) mapping:

[0097] Control object: green body stacking density or drying duration.

[0098] Adjustment logic: ΔK2 every decrease 0.003→green body stacking density decrease 10% (or drying duration increase 5 minutes).

[0099] Safety range: K2∈(0.1,0.5), match the material porosity (e.g. clay K2≈0.12, bone china K2≥0.45).

[0100] Out-of-bound handling:

[0101] K2 out-of-bound: trigger material porosity re-detection procedure, re-calibrate K2 value.

[0102] Third, external convection coefficient (ΔK3) mapping:

[0103] Control object: variable frequency centrifugal fan speed or air valve opening.

[0104] Adjustment logic: ΔK3 every increase 0.0005→fan speed increase 50rpm (or air valve opening +10%).

[0105] Safety range: K3≥0, negative value has no physical meaning (convection only promotes / maintains drying).

[0106] Out-of-bound handling:

[0107] K3<0: parameter auto-zero, trigger level 3 alarm:

[0108] Warning (-0.1≤K3<0): interface flicker prompt, parameter auto-zero.

[0109] Level 1 alarm (K3<-0.1): stop furnace and check the oven sealing.

[0110] Emergency alarm (3 consecutive times K3<-0.3): cut off the heat source, start the safety cooling mode.

[0111] In step S304, the optical line terminal receives the device working condition of the target drying device sent through the uplink, and a model parameter adjustment mode corresponding to the device working condition, wherein the model parameter adjustment mode is an adjustment mode determined by the optical line terminal according to the difference between the water content prediction value and the water content measured value of the water content prediction model.

[0112] In some embodiments of the present application, the optical line terminal sends a model parameter adjustment mode to the optical network unit through the downlink, wherein the model parameter adjustment mode is used to determine the device parameter adjustment mode of the target drying device.

[0113] In some embodiments of the present application, the optical line terminal obtains the device working conditions collected by each sensor through the optical network unit, wherein the sensors are directly connected to the optical network unit. The data collected by each sensor is as follows:

[0114] Microwave moisture meter: real-time measurement of green body moisture content W (precision ±0.3%, sampling interval can be set to 1 minute).

[0115] Infrared temperature measuring instrument: real-time measurement of green body surface temperature T surface (precision ±0.5℃, sampling interval can be set to 10 seconds).

[0116] Weighing sensor: real-time monitoring of green body weight M (range 0-50kg, resolution ±1g, sampling interval can be set to 1 minute).

[0117] Drying room temperature field: calculate the average temperature T oven of the drying room through distributed thermocouples (20 groups) (sampling interval can be set to 10 seconds).

[0118] Outdoor weather data: synchronous collection of outdoor temperature T out and humidity H out (sampling interval can be set to 5 minutes).

[0119] Data transmission: realize millisecond-level synchronization of multi-modal data through industrial PON network (XGS-PON technology), embed μs-level precision timestamp, complete edge computing through OLT, and ensure the spatio-temporal alignment of multi-modal data.

[0120] The optical line terminal can also use the following methods to preprocess the data collected by the sensors:

[0121] Abnormal filtering: eliminate abnormal data caused by sensor noise or communication interference (such as sudden temperature rise);

[0122] Missing interpolation: fill in the missing data due to device failure (such as linear interpolation or heat conduction model prediction);

[0123] Optionally, during the data collection process, the sensors may cause data loss due to communication interruption, device failure or signal interference. In order to ensure data continuity, the appropriate filling method needs to be selected according to the missing duration and parameter characteristics:

[0124] Linear interpolation method: when the variable (temperature, weight, moisture content) parameter changes smoothly without obvious mutation, the data is temporarily missing for 1-3 consecutive time points.

[0125] Sliding window median replacement method:

[0126] When high-frequency sampling data (such as temperature, moisture content) and short-term sudden anomalies (such as single-point spikes) appear abnormally, such abnormal values are usually caused by sensor noise, electromagnetic interference or equipment abnormalities, and need to be replaced according to the median by defining a sliding window and statistically identifying.

[0127] Exponential smoothing method: generate smoothing parameters (W smoothed , T smoothed , M smoothed ) through time series algorithm-exponential smoothing method.

[0128] After preprocessing, the OLT can integrate the preprocessed parameters (W smoothed , T smoothed , M smoothed ) and the original data (T oven , H out ) to generate dynamic trend parameters (ΔW, ΔM).

[0129] In some embodiments of the present application, the green body surface temperature (T t ), the green body weight (M t ), the green body moisture content (W t ), the air humidity (H out (t)), and the average temperature of the drying room (T over (t)) at time t can all be preprocessed using the same exponential smoothing method to eliminate random fluctuations in the data and highlight future trends. The essence is to achieve decay weight through historical data. The following is the exponential smoothing process:

[0130] Weighted average mechanism: each new data point W t at time t (such as the current moisture content) will be mixed with the historical smoothed value W t smoothed according to the preset weight, and the formula is:

[0131] X t smoothed = a * X t + (1-a) * X t-1 smoothed

[0132] Where:

[0133] X represents the parameter to be smoothed (W t , T t , M t , H out (t), T over(t)); a = 0.3, representing new data weight; 1-a = 0.7, representing historical smoothing value weight, the weight can be adjusted.

[0134] After that, the single scale data can be integrated in minute granularity, and X t smoothed The following parameter table is formed:

[0135]

[0136] In some embodiments of the present application, a moisture evaporation rate calculation model is included in the water content prediction model for calculating the instantaneous evaporation rate As an optional implementation, the model coefficients (K1, K2, K3) can be calibrated in real time based on Δw, W pred and the error rate of the measured value W meas The model coefficients (K1, K2, K3) can be calibrated in real time as follows:

[0137] Single-step prediction:

[0138] Error calculation: ε = (W pred -W meas ) / W meas × 100%;

[0139] Step adjustment rules:

[0140] ε>+5%: ΔK1=-0.002, ΔK2=+0.005, ΔK3=-0.0003 (suppress overheating risk);

[0141] +2%≤ε≤+5%: ΔK1=-0.001, ΔK2=+0.003 (mild correction);

[0142] |ε|≤2%: lock parameters (steady state operation);

[0143] -5%≤ε<-2%: ΔK1=+0.001, ΔK2=-0.003 (compensate for insufficient drying);

[0144] ε<-5%: ΔK1=+0.003, ΔK3=+0.0005 (emergency heat enhancement).

[0145] In some embodiments of the present application, the construction of the moisture evaporation rate calculation model in the water content prediction model is completed by the joint scale parameters called by the lightweight container (Docker) built in the OLT, and the instantaneous evaporation rate is calculated after the millisecond-level timestamp is aligned and input into the model. Among them, the evaporation rate calculation formula of the threshold value in the container is as follows:

[0146]

[0147] T equilibrium (t) = T oven (t) - β * (H out(t) -H ref )

[0148] In the above formula, T t : the green body surface temperature taken from the joint table. T equilibrium : the thermodynamic equilibrium temperature under the current environmental humidity (the higher the humidity, the more difficult it is for water to evaporate). M: the green body real-time weight taken from the joint table (measured by a weighing sensor) T oven : the average temperature of the drying room (collected by a thermocouple) H out (t): the real-time outdoor humidity (collected by a weather station) H ref = 50%: the reference humidity (an empirical set value) Coefficient: β = 0.005℃ / %: the humidity attenuation factor, indicating the influence of each 1% humidity deviation on the equilibrium temperature; k1, k2, k3: the material quality coefficients, which can be obtained through artificial testing or data analysis, and the influence, calibration method and actual significance of each material quality coefficient are shown in the following table:

[0149]

[0150] In some embodiments of the present application, is the internal diffusion influence item of the formula. The numerator is the difference between the surface temperature and the environmental equilibrium temperature, and the larger the difference, the faster the water evaporates. The larger the denominator M, the more heat is required to evaporate the water, and the lower the evaporation rate driven by the unit temperature difference. The exponential attenuation e -k2W indicates that the higher the water content W, the greater the diffusion resistance. k3 × (T oven -T t ) is the external heating influence item, indicating the temperature difference between the drying room temperature and the surface temperature of the material such as the green body. The larger the temperature difference, the faster the external heat will accelerate the evaporation of water on the surface of the material.

[0151] In some embodiments of the present application, the OLT can be built-in lightweight containers to perform five-level parameter step adjustment on the parameters of the model according to the real-time prediction error rate (ε = (W pred -W meas ) / W meas × 100%) to ensure the safety of the drying process. At the same time, the local cache optimizes the parameters under high-frequency working conditions, supporting emergency regulation in case of network interruption. In actual operation, the OLT of the edge layer synchronizes the ΔK adjustment parameters and working condition characteristics to the cloud through the PON uplink (delay ≤ 1 ms), the cloud performs knowledge base entry merging, weight redistribution and low-frequency data cleaning, generates an optimization strategy through a similarity matching algorithm, and returns it to the OLT, realizing continuous self-evolution of the model.

[0152] In some embodiments of the present application, the model parameter adjustment method includes the adjustment amount of each model parameter. The OLT can predict the next step moisture content by numerical integration based on the current material moisture content state Wt and evaporation rate, and the predicted value approximates the next time measured value, dynamically calibrating K1, K2, K3, the process is as follows:

[0153] Evaporation rate calculation process:

[0154] Moisture content single step prediction:

[0155] Error calculation:

[0156] a Moisture content measurement: the next minute combined gauge W t value (W t+1 )

[0157] b Error rate calculation:

[0158] Parameter step adjustment rule:

[0159] According to the numerical range of model prediction error (ε), the correction step of heat transfer efficiency coefficient (ΔK1), internal moisture diffusion resistance coefficient (ΔK2) and external heat convection coefficient (ΔK3) is dynamically adjusted, and the specific rules are as follows:

[0160] First, positive error emergency intervention.

[0161] Condition: error ε>+5% (actual drying speed is significantly faster than the predicted value, which may cause the green body to crack).

[0162] Adjustment action:

[0163] Reduce heat transfer efficiency: ΔK1=-0.002;

[0164] Increase internal diffusion resistance: ΔK2=+0.005;

[0165] Inhibit external convection: ΔK3=-0.0003.

[0166] Physical meaning: by inhibiting heat transfer and moisture diffusion, the drying speed is slowed down to prevent over-drying.

[0167] Second, positive error mild adjustment.

[0168] Condition: +2%≤ε≤+5% (drying speed is slightly faster than expected, which needs to be corrected slightly).

[0169] Adjustment action:

[0170] Small decrease in heat transfer efficiency: ΔK1=-0.001;

[0171] Moderately increase diffusion resistance: ΔK2 = +0.003;

[0172] External convection remains unchanged: ΔK3 = 0.

[0173] Objective: Balance drying efficiency and safety, avoid drastic adjustments causing process fluctuations.

[0174] Third, steady-state locking.

[0175] Condition: |ε| ≤ 2% (Model prediction accuracy is high, process state is stable).

[0176] Adjustment action: ΔK1 = ΔK2 = ΔK3 = 0 (All parameters remain current values).

[0177] Effect: Maintain current parameter configuration, ensure smooth operation of the drying process.

[0178] Fourth, negative error compensation adjustment.

[0179] Condition: -5% ≤ ε < -2% (Actual drying speed is slower than predicted value, there is a risk of insufficient drying).

[0180] Adjustment action:

[0181] Increase heat transfer efficiency: ΔK1 = +0.001;

[0182] Reduce diffusion resistance: ΔK2 = -0.003;

[0183] External convection remains unchanged: ΔK3 = 0;

[0184] Physical meaning: Accelerate heat penetration and moisture migration, compensate for insufficient drying.

[0185] Fifth, negative error abnormal recovery.

[0186] Condition: ε < -5% (Drying speed lags seriously, may be due to environmental mutation or equipment failure).

[0187] Adjustment action:

[0188] Significantly increase heat transfer efficiency: ΔK1 = +0.003;

[0189] External convection is strengthened: ΔK3 = +0.0005;

[0190] Internal diffusion resistance remains unchanged: ΔK2 = 0.

[0191] Objective: Emergency enhance external heat source and air flow, quickly restore drying efficiency.

[0192] Among the various parameters of the model, K1 is the heat transfer efficiency coefficient, the range is set in relation to the actual physical process, and exceeding the range may cause the model to be inaccurate, so the historical optimal value needs to be loaded to restore stability. K2 is related to the material porosity, and exceeding the limit means that the material properties change, so re-detection is required. The negative value of K3 has no physical meaning, and automatic zeroing and alarm can prevent incorrect calculation. The specific value range of each parameter is as follows:

[0193]

[0194] K1∈[0.01, 0.05], lower limit 0.01: ensure minimum effective heat transfer (lower than this value, drying stagnation), upper limit 0.05: prevent overheating leading to surface hard shell (higher than this value may cause cracking).

[0195] K2∈[0.1, 0.5], lower limit 0.1: minimum diffusion resistance (corresponding to high porosity materials such as clay), material properties: clay porosity > 35%, K2≈0.12, upper limit 0.5: maximum diffusion resistance (corresponding to low porosity materials such as bone china), industrial standard: bone china porosity < 8% requires K2≥0.45.

[0196] K3≥0, non-negative constraint: negative convection coefficient has no physical meaning (air convection only promotes / maintains drying) physical verification: when K3<0, the model predicts that the moisture content will rise against the trend, violating the laws of thermodynamics.

[0197] In some embodiments of the present application, after the OLT determines the model parameter adjustment mode and optimizes the model parameters according to the adjustment mode, it also records the current device operating conditions, including the material parameters of the material in the drying equipment, as shown in the following table:

[0198]

[0199]

[0200] Step S306, updating the model parameter set corresponding to the historical operating conditions in the knowledge base according to the device operating conditions and the model parameter adjustment mode corresponding to the device operating conditions.

[0201] In some embodiments of the present application, the knowledge base in the cloud can provide reliable initial parameter combinations (K1, K2, K3 and their adjustment steps ΔK1, ΔK2, ΔK3) for different operating conditions through accumulation and optimization of historical data, and support adaptive optimization of the model.

[0202] The knowledge base can support the index of working condition discretization, by mapping real-time working condition parameters (green body weight, oven temperature, green body surface temperature, external humidity) to a pre-defined discrete label combination (such as "medium weight + medium temperature + medium humidity + medium surface temperature"), and taking the discrete label combination as the index of the knowledge base.

[0203] In addition, the knowledge base also records the historical optimal parameter combination (K1, K2, K3) and its adjustment strategy (ΔK1, ΔK2, ΔK3) under each working condition combination, and marks the learning times (N, the cumulative number of successful applications of the working condition combination) and the weighted score (the comprehensive error rate and the good product rate).

[0204] In some embodiments of the present application, the parameter combination in the knowledge base can also be dynamically adjusted through model prediction error (ε) and feedback of actual production data to ensure its long-term effectiveness.

[0205] In some embodiments of the present application, the knowledge base can store the following data with the above discrete label combination as the index: historical optimal parameter combination: heat transfer efficiency coefficient (K1), internal diffusion resistance coefficient (K2), and external convection coefficient (K3), and its adjustment step (ΔK1, ΔK2, ΔK3).

[0206] Learning times (N): the cumulative number of successful applications of the working condition combination, each successful application of a working condition parameter combination (ε≤5% and Y≥95%) increases the value of N by 1; if ε>5% after continuous application for 3 times, the value of N is reset to 0, and the knowledge base is triggered to be cleaned up.

[0207] Weighted score (S): the score of the comprehensive model prediction error rate (ε) and the good product rate (Y), used to screen reliable strategies, the weighted score (S) is calculated as follows:

[0208]

[0209] W pred Model predicted moisture content, W meas : Next minute joint gauge measured moisture content

[0210] Good product rate (Y): the proportion of green bodies without cracking and deformation after drying, the calculation method is as follows:

[0211]

[0212] Scoring logic:

[0213] Error rate accounts for 70% (the lower the error, the higher the score); good product rate accounts for 30% (the higher the good product rate, the higher the score); full score is 100 points (ε=0% and Y=100%, S=100).

[0214] In some embodiments of the present application, the data stored in the knowledge base is shown in the following table:

[0215]

[0216] In the technical solution provided in step S206, updating the model parameter set corresponding to the historical working condition in the knowledge base according to the device working condition and the model parameter adjustment mode corresponding to the device working condition comprises: determining the similarity between the device working condition and the historical device working condition in the knowledge base, wherein the knowledge base includes a plurality of historical working conditions and a model parameter set of the water content prediction model corresponding to the historical working condition; in the case that there is a first device historical working condition corresponding to a similarity not less than a first preset similarity threshold in the knowledge base, adjusting the model parameter set corresponding to the first device historical working condition according to the model parameter adjustment mode; in the case that there is no first device historical working condition in the knowledge base, and there is a second historical device working condition corresponding to a similarity less than the first preset similarity threshold and not less than a second preset similarity threshold, adjusting the model parameter set corresponding to the second historical device working condition according to the model parameter adjustment mode; in the case that there is no corresponding first device historical working condition and second historical device working condition in the knowledge base, determining the model parameter set corresponding to the device working condition according to the model parameter adjustment mode, and storing the device working condition and the model parameter set corresponding to the device working condition in the knowledge base.

[0217] In some embodiments of the present application, the cloud device determines the similarity between the device working condition and the historical device working condition in the knowledge base, which comprises: determining the discrete parameter level label corresponding to each working condition parameter in the device working condition, wherein the discrete parameter level label is used to indicate the preset parameter value range in which the parameter value of the working condition parameter is located, and the discrete parameter level label corresponding to each working condition parameter is determined by the optical line terminal according to the preset parameter value range in which the parameter value of the working condition parameter is located; determining the candidate historical device working condition in the knowledge base according to the discrete level label corresponding to the working condition parameter; and determining the similarity between the device working condition and the candidate historical device working condition according to the parameter values of each working condition parameter in the device working condition and the parameter values between each working condition parameter in the candidate historical device working condition.

[0218] In some embodiments of the present application, the process of determining the discrete parameter level label is also the process of discretizing the parameter. It can be completed by the OLT or by the cloud device.

[0219] The purpose of discretizing the working condition parameter is to map the real-time parameter to a predefined discrete label combination, providing an index basis for knowledge base query. The working condition parameters that need to be processed include the current working condition parameters including the green body weight (M), the baking room temperature (T oven ), the external humidity (H out ), the green body surface temperature (Tt ). The specific discretization partition table can be shown in the following table, where the weights in the table are used to calculate the similarity, and the specific values in the table are only illustrative:

[0220]

[0221]

[0222] In some embodiments of the present application, the weights in the above table can reflect the influence degree of each working condition parameter on the model error rate. For example, the weight of the temperature (T oven ) of the drying room is set to the maximum because it directly affects the thermodynamic equilibrium and the evaporation rate. The weight of the second largest weight of the clay weight (M) is more related to the drying result than the process control.

[0223] The weight of the external humidity is set to the third largest because the external humidity is related to T equilibrium , which indirectly affects the evaporation rate. The weight of the surface temperature of the clay is the lowest because this parameter mainly reflects the local drying state of the material.

[0224] In addition, it should be noted that the above weights are not fixed values. Each quarter, the weights can be recalculated according to the latest production data.

[0225] As an optional implementation, determining the similarity between the device working condition and the candidate historical device working condition according to the parameter values of each working condition parameter in the device working condition and the parameter values of each working condition parameter in the candidate historical device working condition comprises: determining a parameter type of the working condition parameter and a weight coefficient corresponding to the parameter type, wherein the weight coefficient is used to reflect the influence degree of the parameter type on the moisture content of the material, and the greater the weight coefficient, the greater the influence degree; and determining the similarity between the device working condition and the candidate historical device working condition according to the weight coefficient corresponding to the parameter type, the parameter values of each working condition parameter in the device working condition, and the parameter values of each working condition parameter in the candidate historical device working condition.

[0226] In some embodiments of the present application, the cloud device determines the similarity between the device working condition and the historical device working condition in the knowledge base, comprising: determining a discretized parameter level label corresponding to each working condition parameter in the historical device working condition, wherein the discretized parameter level label is used to indicate a preset parameter value range in which the parameter value of the working condition parameter is located; determining a value at a center point of the preset parameter value range as the parameter value of the working condition parameter in the historical device working condition; and determining the similarity between the device working condition and the historical device working condition according to the parameter value of the working condition parameter in the device working condition and the parameter value of the working condition parameter in the historical device working condition.

[0227] As an optional implementation, before adjusting the model parameter set corresponding to the second historical device working condition according to the model parameter adjustment mode, the method further comprises: determining a discretized parameter level label corresponding to each working condition parameter in the second historical device working condition, wherein the discretized parameter level label is used to indicate a preset parameter value range in which the parameter value of the working condition parameter is located; determining a preset parameter value range corresponding to each working condition parameter in the device working condition according to the parameter type; in the case that the parameter value of a target working condition parameter in the device working condition is not in the corresponding preset parameter value range, determining that there is no second historical device working condition corresponding to the device working condition, wherein the target working condition parameter is a working condition parameter of the preset parameter type; in the case that there are more than a preset number of non-target working condition parameters in the device working condition, the parameter value of which is not in the corresponding preset parameter value range, determining that there is no second historical device working condition corresponding to the device working condition, wherein the non-target working condition parameter is a working condition parameter of a parameter type other than the preset parameter type, and the influence degree of the target working condition parameter on the moisture content of the material is greater than the influence degree of the non-target working condition parameter on the moisture content of the material; in the case that there are non-target working condition parameters in the device working condition, which are not in the corresponding preset parameter range and have a deviation amplitude from the corresponding preset parameter range greater than a preset deviation amplitude threshold, determining that there is no second historical device working condition corresponding to the device working condition.

[0228] In some embodiments of the present application, the process of matching the discretized working condition parameters and historical data can be completed by a cloud device. For example, after the OLT maps the real-time parameters (M, T oven , T t , H out ) to a predefined discrete label combination (such as “medium weight + medium temperature + medium humidity + medium surface temperature”), the cloud device can use the discrete label combination as an index to search for similar historical device working conditions in the knowledge base.

[0229] In some embodiments of the present application, since the working condition parameters can change continuously, the change of the working condition is discontinuous, and the parameters can be near the critical value of different gears (for example, M = 500 kg is just at the boundary of “light” and “medium”, and if there is a slight fluctuation, it can cause the matching result to jump), which affects the stability. Therefore, a dynamic similarity matching algorithm can be used to more accurately reflect the actual distance between the parameters and the levels.

[0230] Optionally, the dynamic similarity matching method comprises the following steps:

[0231] First step, definition of level center point

[0232] For each discretized level parameter interval, define its level center point as a representative value.

[0233] Second step, parameter normalization processing

[0234] To eliminate dimensional differences and facilitate similarity calculation, the real-time parameters are normalized.

[0235] Step 3: Parameter range definition

[0236] The following table is an example of the definition of the value range of each parameter.

[0237]

[0238] Step 4: Similarity calculation based on Euclidean geometry method

[0239] ①

[0240] Weight ωi: can be assigned according to the priority of the parameter, different values of i represent different parameters, such as i=1 represents the weight of the green body, i=2 represents the temperature of the drying room, i=3 represents the surface temperature of the green body, i=4 represents the external humidity, etc.

[0241] ② Similarity conversion: Where, D max means that each parameter takes the extreme value combination.

[0242] Step 5: Matching result determination

[0243] Complete match: if the similarity is ≥0.95 (threshold value can be adjusted), directly determine the corresponding label combination of the historical equipment working condition as the similar working condition.

[0244] Partial match: if 0.8≤similarity<0.95, trigger parameter priority verification process.

[0245] Low similarity: if the similarity is <0.8, it is determined as a new working condition, and the knowledge base optimization process is triggered.

[0246] Optionally, for the case of partial match, the weight of each parameter can be adjusted first, for example, the weight of the important parameter drying room temperature is increased, and then the similarity calculation is performed again.

[0247] In addition, the most important working condition parameter (i.e., the working condition parameter with the largest weight) can be determined according to the weight parameter, and it is determined whether the important working condition parameter in the device working condition matches the discrete label of the historical device working condition. If the most important working condition parameter does not match, manual confirmation is performed. Among the remaining working condition parameters other than the most important working condition parameter, if only one working condition parameter does not match and the deviation from the corresponding preset value range is less than a preset amplitude (such as 20%), it can be considered that the historical device working condition corresponding to the label combination is a similar working condition. In determining the deviation amplitude between the working condition parameter and the corresponding preset value range, the end point of the preset value range that is closer to the working condition parameter can be determined as the reference value. Then, the ratio of the absolute value of the difference between the working condition parameter and the reference value to the reference value can be calculated as the deviation amplitude.

[0248] In some embodiments of the present application, when a new device working condition is received, if the error rate and the good product rate of the working condition meet the preset requirements (for example, ε≤5% and Y≥95%), the model parameters corresponding to the device working condition and the historical device working condition are updated by weighting in the case that a similar historical device working condition matching the working condition is determined, and the specific update formula is:

[0249] K 新 = 0.7*K 历史 + 0.3*K 当前

[0250] In some embodiments of the present application, for the abnormal entry data in the knowledge base, if a certain entry triggers ε>5% for three times in succession, the entry is deleted and the model parameters corresponding to the original label combination of the entry are reset to the laboratory calibration value.

[0251] In addition, for similar working conditions in the knowledge base, if the Euclidean distance between two entries is ≤0.1 and the process characteristics are consistent (such as a difference of ≤1% in the drying end point), the two entries can be merged into one entry, and the weighted average parameter is taken as the updated parameter.

[0252] As an optional implementation, the data can also be cleaned according to the timeliness of the data. For example, for data that has not been matched for more than a certain time, a low evaluation score can be set for the data to calculate the weight. Then, according to the evaluation scores of the entries in the knowledge base, entries with low scores can be cleaned regularly. The specific score evaluation rules are not limited here.

[0253] In some embodiments of the present application, after the knowledge base is optimized, the cloud device can distribute the optimized knowledge base to the OLT of the edge layer through the PON network (downlink delay ≤1ms), the OLT dynamically adjusts the local cache according to the cloud strategy, and feeds back the execution effect to the cloud to form a closed loop of "edge execution-cloud learning-strategy iteration".

[0254] In some embodiments of the present application, the water content prediction model can also be calibrated offline in the following manner. Optionally, to ensure long-term accuracy of the model, a piece of green body sample can be taken every certain period of time, the true water content is determined by the oven method, and the model prediction error rate is calculated by comparing the predicted value with the measured value, with the following calculation formula:

[0255]

[0256] In the above formula, the superscript pred represents the predicted value, the superscript meas represents the measured value, and the subscript t+1 represents the corresponding time.

[0257] If the error rate is > 5%, the following correction process can be triggered:

[0258] Parameter reset: K1 / K2 / K3 is restored to the laboratory calibration value, and the current optimization step is cleared.

[0259] Knowledge base cleaning: delete historical records associated with the working condition (such as "medium weight + medium temperature" combination) to prevent abnormal data pollution.

[0260] Equipment linkage calibration: simultaneously adjust the oven temperature and fan speed.

[0261] In some embodiments of the present application, when maintaining the water content prediction model for a long time, the parameter weights can be updated based on quarterly production data through multivariate regression analysis. Or when the cumulative error rate increases by > 2% compared with the previous period, the weight redistribution is forcibly started.

[0262] For various sensors, laboratory calibration can be performed on various sensors according to a preset period to update the equipment error coefficient. And the sensor drift can be written into the preprocessing layer to correct the original data (such as Tsurface) in real time.

[0263] In some embodiments of the present application, before initializing and parameter calibrating the water content prediction model, the related parameters of the oven and the parameters of the green body and other materials also need to be determined. For example, the parameters of the oven include:

[0264] Size: 20m x 4m x 3m (length x width x height).

[0265] Heating method: dual-redundancy heating of electric circulating fan (power 0-50kW) and gas boiler (thermal efficiency ≥ 90%).

[0266] Temperature control accuracy: ±1.5℃ (based on feedback control of physical model in section 3.4).

[0267] Air convection system: variable frequency centrifugal fan (0-3000 rpm), supporting dynamic adjustment of external heat convection coefficient (K3).

[0268] Material parameters include:

[0269] Material: Kaolin mixture (porosity 23%, corresponding to K2 initial value 0.15).

[0270] Specifications: diameter 200 mm, thickness 15 mm.

[0271] Initial state: water content W0 = 22.5% (microwave moisture meter calibration), weight M0 = 0.6 Kg-1 kg.

[0272] Sensor configuration as follows:

[0273] Microwave moisture meter (MOSYE):

[0274] Measurement depth 5-15 mm, accuracy ±0.3%, sampling interval 1 minute.

[0275] Dual-probe symmetric installation, through-type measurement, eliminating surface interference.

[0276] Infrared temperature measurement instrument (GRS):

[0277] Temperature measurement range 0-200°C, accuracy ±0.5°C, sampling interval 10 seconds.

[0278] Three-directional three-dimensional layout, calculating the average of the surface temperature field of the mud body (T surface ).

[0279] Weighing sensor (Mettler Toledo SB01):

[0280] Range 0-50 kg, resolution ±1 g, real-time monitoring of weight (M).

[0281] Oven temperature field monitoring:

[0282] Distributed thermocouples (20 groups, uniformly distributed in three dimensions), calculating the average oven temperature (T oven ).

[0283] Outdoor weather station (Davis Vantage Pro2): temperature and humidity sampling interval 5 minutes, data synchronized in real time to the cloud platform through the industrial PON network.

[0284] For the industrial PON network architecture, the ONUs in the access layer can be configured to support Modbus RTU protocol parsing and timestamp alignment. The OLT in the transmission layer can be equipped with an edge computing container (Docker) to implement data preprocessing and joint metering construction. The data collected by the ONUs and the OLT can be uploaded to a private cloud via XGS-PON (10G bandwidth), and a water content prediction model can be deployed in the OLT, so that the model control instruction is issued with a delay of ≤1 ms.

[0285] In some embodiments of the present application, the model parameters of the water content prediction model can be initialized according to actual conditions, for example:

[0286] K1 (heat transfer efficiency coefficient): In a constant temperature 60℃ drying room, the rate of the center temperature of the body from 25℃ to 55℃ is measured, and K1 = 0.018 is fitted by a curve.

[0287] K2 (internal diffusion resistance coefficient): In a constant humidity 30% RH environment, the water content decay data (W(t)) of the body is recorded, and K2 = 0.15 is determined by fitting the exponential equation W(t) = 18.0e(-0.15t) + 2.0 through nonlinear least squares method.

[0288] K3 (external convection coefficient): Adjust the fan speed (1000-2500 rpm), measure the temperature difference ΔT and ΔW relationship, and calibrate K3 = 0.002. oven -T surface ΔW relationship, and calibrate K3 = 0.002.

[0289] In addition, based on historical data, the model parameter set corresponding to the discrete label combination matching the current actual environment can be screened and loaded. For example, the initial parameters K1 = 0.019, K2 = 0.148, and K3 = 0.002, and the data corresponding to the discrete label combination have a weighted score S = 92.5.

[0290] During the drying process, the real-time collected data are shown in the following table:

[0291]

[0292] When the real-time collected working condition parameters change, the real-time adjustment of the model parameters can be triggered. For example, it is detected that T oven from 80℃ to 82℃ (ΔT = +2℃), matching the discrete level "medium weight + medium temperature", calling the knowledge base parameters (K1 = 0.019, ΔK1 = -0.001).

[0293] Then the error calculation can be performed, assuming that W pred = 18.5%, W meas= 18.2%, then ε = +1.2%. Then, according to the error interval rule (|ε|≤2%) → lock the current parameters (ΔK1=ΔK2=ΔK3=0).

[0294] In the data collection process, if the error rate of each working condition parameter corresponding to the current model parameter combination fluctuates by ≤±3% for 10 minutes continuously and the good product rate Y≥95%, the model parameter combination is updated to the knowledge base. Before updating, the safety of the current working condition also needs to be verified. Specifically, it includes determining whether the current model parameters and each working condition parameter are within the allowed value range (such as the oven temperature ∈ [60℃, 150℃] and K3≥0). If it does not pass, it is not updated to the knowledge base.

[0295] In addition, the data in the knowledge base can be periodically checked according to the above standard and the data entries that do not pass the verification are deleted.

[0296] In order to reflect the advantages of the method provided in the present application over the related art, in some embodiments of the present application, the method provided in the present application and the method in the related art are used to control and schedule the drying device in two preset scenarios. Scenario 1 is constant temperature drying (T oven = 80℃), and scenario 2 is variable temperature drying (T oven from 70℃ to 90℃ in steps).

[0297] The error rate and control action of the method provided in the present application under scenario 1 are shown in the following table:

[0298]

[0299] The error rate and control action of the method provided in the present application under scenario 2 are shown in the following table:

[0300]

[0301] By comparing various key performance indicators, after using the method provided in the embodiments of the present application, the average error rate is 2.8%, while the method in the related art is 12.3%. In terms of process stability, the steady-state locking (|ε|≤2%) of the method provided in the embodiments of the present application accounts for 85%. In terms of quality improvement, the cracking defect rate of the material is reduced from 5.7% to 0.9%. In terms of energy efficiency optimization, the drying cycle is shortened by 21.4% and the energy consumption is reduced by 18%.

[0302] In some embodiments of the present application, a method for controlling a drying device is also provided, which includes the following steps: Figure 4The drying equipment control flow is shown. In the flow, the sensors such as the microwave moisture meter and the infrared temperature meter of the production equipment layer collect data, which is transmitted to the OLT after protocol analysis and time calibration are completed by the ONU; the OLT first performs preprocessing such as time alignment, bad point interpolation, and data smoothing on the data, generates joint parameters, and then calculates the crystal point rate error and predicts the moisture content based on the physical / data model, and outputs the initial control parameters (ΔK1, ΔK2, ΔK3, etc.) to link and control the gas boiler power, air supply temperature, fan speed, and the like of the production equipment; after the control command is fed back to the local cloud, the working condition similarity is calculated, and the branching processing is performed according to ≥0.95 (accurate matching, calling historical optimal parameters), 0.8-0.95 (fault tolerance matching, allowing a 20% deviation in minor parameters, combining historical and current values to generate new parameters), and <0.8 (new working condition self-learning, executing initial parameters); at the same time, the knowledge base is stored in the database according to the error ≤5% and the success rate ≥95%, and the working conditions without data for 90 days or with an error exceeding 5% for 3 times are eliminated. The rules are dynamically maintained, the model is calibrated regularly and the rule base is updated, the working condition simplification level is adaptively adjusted, and finally a closed-loop system of "data collection→edge control→knowledge iteration→optimization control" is formed.

[0303] It should be noted that the specific numerical values mentioned in the embodiments of the present application are only for reference and do not represent a limitation on the present application.

[0304] By adopting the cloud device to send the knowledge base to the optical line terminal, the knowledge base includes historical working conditions of a plurality of target drying equipment and a model parameter set corresponding to a moisture content prediction model of the historical working conditions, the moisture content prediction model is set in the optical line terminal and is used to predict the moisture content of the material in the target drying equipment; the device working condition of the target drying equipment and a model parameter adjustment mode corresponding to the device working condition are received by the optical line terminal through the uplink, the model parameter adjustment mode is an adjustment mode determined by the optical line terminal according to the difference between the moisture content prediction value and the moisture content measured value of the moisture content prediction model; the model parameter set corresponding to the historical working condition in the knowledge base is updated according to the device working condition and the model parameter adjustment mode corresponding to the device working condition, the model is deployed to the optical line terminal on the edge side, and the model parameter adjustment mode is determined by the optical line terminal, so as to reduce the response delay, thereby realizing the technical effect of timely response when the device working condition and the like change, and further solving the technical problem of high response delay caused by setting the model in the cloud in the related technology.

[0305] According to the embodiments of the present application, a moisture content prediction model parameter adjustment device is provided, as shown in Figure 5 The structure of the moisture content prediction model parameter adjustment device is shown in FIG. 1. The device working condition of the target drying equipment is received by the optical line terminal through the uplink, and the model parameter adjustment mode corresponding to the device working condition is determined by the optical line terminal according to the difference between the moisture content prediction value and the moisture content measured value of the moisture content prediction model. Figure 5As can be seen, the device comprises: a first processing module 50 configured to send a knowledge base to an optical line terminal, wherein the knowledge base comprises historical working conditions of a plurality of target drying devices and a model parameter set of a moisture content prediction model corresponding to the historical working conditions, and the moisture content prediction model is arranged in the optical line terminal and configured to predict the moisture content of materials in the target drying device; a second processing module 52 configured to receive device working conditions of the target drying device and a model parameter adjustment mode corresponding to the device working conditions sent by the optical line terminal through an uplink, wherein the model parameter adjustment mode is an adjustment mode determined by the optical line terminal according to a difference between a predicted value and a measured value of the moisture content of the moisture content prediction model; and a third processing module 54 configured to update the model parameter set corresponding to the historical working conditions in the knowledge base according to the device working conditions and the model parameter adjustment mode corresponding to the device working conditions.

[0306] In some embodiments of the present application, the third processing module 54 updates the model parameter set corresponding to the historical working conditions in the knowledge base according to the device working conditions and the model parameter adjustment mode corresponding to the device working conditions, including: determining a similarity between the device working conditions and historical device working conditions in the knowledge base, wherein the knowledge base comprises a plurality of historical working conditions and a model parameter set of a moisture content prediction model corresponding to the historical working conditions; in a case where there is a first device historical working condition corresponding to a similarity not less than a first preset similarity threshold in the knowledge base, adjusting the model parameter set corresponding to the first device historical working condition according to the model parameter adjustment mode; in a case where there is no first device historical working condition in the knowledge base, and there is a second historical device working condition corresponding to a similarity less than the first preset similarity threshold and not less than a second preset similarity threshold, adjusting the model parameter set corresponding to the second historical device working condition according to the model parameter adjustment mode; and in a case where there is no first device historical working condition and second historical device working condition corresponding in the knowledge base, determining a model parameter set corresponding to the device working conditions according to the model parameter adjustment mode, and storing the device working conditions and the model parameter set corresponding to the device working conditions in the knowledge base.

[0307] In some embodiments of the present application, the third processing module 54 determines the similarity between the device working conditions and the historical device working conditions in the knowledge base, including: determining a discretized parameter level label corresponding to each working condition parameter in the device working conditions, wherein the discretized parameter level label is used to indicate a preset parameter value range in which a parameter value of the working condition parameter is located, and the discretized parameter level label corresponding to each working condition parameter is determined by the optical line terminal according to the preset parameter value range in which the parameter value of the working condition parameter is located; determining a candidate historical device working condition in the knowledge base according to the discretized level label corresponding to the working condition parameter; and determining the similarity between the device working conditions and the candidate historical device working conditions according to the parameter values of each working condition parameter in the device working conditions and the parameter values between each working condition parameter in the candidate historical device working conditions.

[0308] In some embodiments of the present application, the third processing module 54 determines the similarity between the device working condition and the historical device working condition in the knowledge base comprises: determining the discrete parameter level label corresponding to each working condition parameter in the historical device working condition, wherein the discrete parameter level label is used to indicate the preset parameter value range in which the parameter value of the working condition parameter is located; determining the value at the center point of the preset parameter value range as the parameter value of the working condition parameter in the historical device working condition; determining the similarity between the device working condition and the historical device working condition according to the parameter value of the working condition parameter in the device working condition and the parameter value of the working condition parameter in the historical device working condition.

[0309] In some embodiments of the present application, the third processing module 54 determines the similarity between the device working condition and the historical device working condition in the knowledge base comprises: determining the discrete parameter level label corresponding to each working condition parameter in the historical device working condition, wherein the discrete parameter level label is used to indicate the preset parameter value range in which the parameter value of the working condition parameter is located; determining the value at the center point of the preset parameter value range as the parameter value of the working condition parameter in the historical device working condition; determining the similarity between the device working condition and the historical device working condition according to the parameter value of the working condition parameter in the device working condition and the parameter value of the working condition parameter in the historical device working condition.

[0310] In some embodiments of the present application, before adjusting the model parameter set corresponding to the second historical device working condition according to the model parameter adjustment mode, the third processing module 54 is further configured to: determine the discrete parameter level label corresponding to each working condition parameter in the second historical device working condition, wherein the discrete parameter level label is used to indicate the preset parameter value range in which the parameter value of the working condition parameter is located; determine the preset parameter value range corresponding to each working condition parameter in the device working condition according to the parameter type; in the case that the parameter value of the target working condition parameter in the device working condition is not in the corresponding preset parameter value range, determine that there is no second historical device working condition corresponding to the device working condition, wherein the target working condition parameter is the working condition parameter of the preset parameter type; in the case that there are more than a preset number of non-target working condition parameters in the device working condition whose parameter values are not in the corresponding preset parameter value range, determine that there is no second historical device working condition corresponding to the device working condition, wherein the non-target working condition parameter is the working condition parameter whose type is not the preset parameter type, and the influence degree of the target working condition parameter on the moisture content of the material is greater than the influence degree of the non-target working condition parameter on the moisture content of the material; in the case that there are non-target working condition parameters in the device working condition which are not in the corresponding preset parameter range and the deviation amplitude between the corresponding preset parameter range is greater than a preset deviation amplitude threshold, determine that there is no second historical device working condition corresponding to the device working condition.

[0311] In some embodiments of the present application, the optical line terminal sends a model parameter adjustment mode to the optical network unit through a downlink, where the model parameter adjustment mode is used to determine a device parameter adjustment mode for the target drying device.

[0312] In some embodiments of the present application, the optical line terminal obtains the device working conditions collected by each sensor through the optical network unit, where the sensors are directly connected to the optical network unit.

[0313] It should be noted that each module in the above water content prediction model parameter adjustment method can be a program module (for example, a set of program instructions that implement a certain specific function) or a hardware module. For the latter, it can be in the following form, but is not limited to this: the form of each module is a processor, or the functions of each module are implemented by a processor.

[0314] According to an embodiment of the present application, a non-volatile storage medium is also provided, and the non-volatile storage medium stores a program. When the program runs, it performs the following water content prediction model parameter adjustment method: a cloud device sends a knowledge base to an optical line terminal, where the knowledge base includes historical working conditions of a plurality of target drying devices and a model parameter set of a water content prediction model corresponding to the historical working conditions, the water content prediction model is set in the optical line terminal and is used to predict the water content of materials in the target drying device; the optical line terminal sends device working conditions of the target drying device and a model parameter adjustment mode corresponding to the device working conditions through an uplink, where the model parameter adjustment mode is an adjustment mode determined by the optical line terminal according to a difference between a water content prediction value of the water content prediction model and a water content measured value; and the historical working conditions corresponding to the model parameter set in the knowledge base are updated according to the device working conditions and the model parameter adjustment mode corresponding to the device working conditions.

[0315] According to an embodiment of the present application, an electronic device is also provided, which includes a memory and a processor. The processor is used to run a program stored in the memory. When the program runs, it performs the following water content prediction model parameter adjustment method: a cloud device sends a knowledge base to an optical line terminal, where the knowledge base includes historical working conditions of a plurality of target drying devices and a model parameter set of a water content prediction model corresponding to the historical working conditions, the water content prediction model is set in the optical line terminal and is used to predict the water content of materials in the target drying device; the optical line terminal sends device working conditions of the target drying device and a model parameter adjustment mode corresponding to the device working conditions through an uplink, where the model parameter adjustment mode is an adjustment mode determined by the optical line terminal according to a difference between a water content prediction value of the water content prediction model and a water content measured value; and the historical working conditions corresponding to the model parameter set in the knowledge base are updated according to the device working conditions and the model parameter adjustment mode corresponding to the device working conditions.

[0316] According to the embodiment of the present application, a computer program product is also provided, comprising a computer program which, when executed by a processor, implements the following moisture content prediction model parameter adjustment method: a cloud device sends a knowledge base to an optical line terminal, wherein the knowledge base comprises historical working conditions of a plurality of target drying devices and a model parameter set of a historical working condition corresponding moisture content prediction model, the moisture content prediction model is set in the optical line terminal and is used to predict the moisture content of materials in the target drying device; the device working condition of the target drying device and the model parameter adjustment mode corresponding to the device working condition are received, which are sent by the optical line terminal through an uplink, wherein the model parameter adjustment mode is an adjustment mode determined by the optical line terminal according to the difference between the moisture content prediction value and the moisture content measured value of the moisture content prediction model; the model parameter set corresponding to the historical working condition in the knowledge base is updated according to the device working condition and the model parameter adjustment mode corresponding to the device working condition.

[0317] In the above-described embodiments of the present application, the description of each embodiment has its own focus, and the parts not described in detail in a certain embodiment can be referred to the relevant description of other embodiments.

[0318] In the several embodiments provided by the present application, it should be understood that the disclosed technology can be implemented in other ways. Of course, the unit embodiment described above is only schematic. For example, the division of the units can be different, and each unit or some features can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual coupling or direct coupling or communication connection between each represented or discussed element can be indirect coupling or communication connection through some interface, unit or module, and can be electrical or other form.

[0319] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e. they can be located in one place or distributed on multiple units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0320] In addition, each functional unit in each embodiment of the present application can be integrated into a processing unit, or each unit can exist physically, or two or more units can be integrated into one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0321] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application or the part that essentially contributes to the related art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the method described in the embodiments of the present application. The aforementioned storage medium includes a U disk, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store program codes.

[0322] The above only describes the preferred embodiments of the present application. It should be noted that, for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, which should also be considered as the protection scope of the present application.

Claims

1. A method for adjusting parameters of a water content prediction model, characterized in that, include: The cloud device sends a knowledge base to the optical line terminal. The knowledge base includes the historical operating conditions of multiple target drying equipment and the model parameter set of the moisture content prediction model corresponding to the historical operating conditions. The moisture content prediction model is set in the optical line terminal and is used to predict the moisture content of the materials in the target drying equipment. The optical line terminal receives the equipment operating status of the target drying equipment and the model parameter adjustment method corresponding to the equipment operating status, which is sent by the optical line terminal through the uplink. The model parameter adjustment method is an adjustment method determined by the optical line terminal based on the difference between the moisture content prediction value and the measured moisture content value of the moisture content prediction model. The set of model parameters corresponding to the historical operating conditions in the knowledge base is updated according to the equipment operating conditions and the model parameter adjustment method corresponding to the equipment operating conditions.

2. The method for adjusting parameters of the water content prediction model according to claim 1, characterized in that, Updating the set of model parameters corresponding to the historical operating conditions in the knowledge base based on the equipment operating conditions and the corresponding model parameter adjustment methods includes: Determine the similarity between the equipment operating condition and historical equipment operating conditions in the knowledge base, wherein the knowledge base includes multiple historical operating conditions and a set of model parameters for the water content prediction model corresponding to the historical operating conditions; If a first device historical operating condition with a similarity not less than a first preset similarity threshold exists in the knowledge base, the model parameter set corresponding to the first device historical operating condition is adjusted according to the model parameter adjustment method. If the first historical operating condition of the device does not exist in the knowledge base, and there is a corresponding second historical operating condition of the device with a similarity less than the first preset similarity threshold and not less than the second preset similarity threshold, the model parameter set corresponding to the second historical operating condition is adjusted according to the model parameter adjustment method. If the first historical operating condition and the second historical operating condition of the equipment do not exist in the knowledge base, the model parameter set corresponding to the operating condition of the equipment is determined according to the model parameter adjustment method, and the operating condition of the equipment and the model parameter set corresponding to the operating condition of the equipment are stored in the knowledge base.

3. The method for adjusting parameters of the water content prediction model according to claim 2, characterized in that, The cloud-based device determines the similarity between the device's operating condition and historical device operating conditions in the knowledge base, including: The discretized parameter level label corresponding to each operating condition parameter in the equipment operating condition is determined, wherein the discretized parameter level label is used to indicate the preset parameter value range in which the parameter value of the operating condition parameter is located, and the discretized parameter level label corresponding to each operating condition parameter is determined by the optical line terminal according to the preset parameter value range in which the parameter value of the operating condition parameter is located. Candidate historical equipment operating conditions are determined in the knowledge base based on the discretized level labels corresponding to the operating condition parameters. Based on the parameter values ​​of each operating condition parameter in the equipment operating condition and the parameter values ​​between each operating condition parameter in the candidate historical equipment operating conditions, the similarity between the equipment operating condition and the candidate historical equipment operating conditions is determined.

4. The method for adjusting parameters of the water content prediction model according to claim 3, characterized in that, Determining the similarity between the equipment operating condition and the candidate historical equipment operating condition based on the parameter values ​​of each operating condition parameter in the equipment operating condition and the parameter values ​​between each operating condition parameter in the candidate historical equipment operating condition includes: The parameter type of the operating condition parameter and the weighting coefficient corresponding to the parameter type are determined. The weighting coefficient is used to reflect the degree of influence of the parameter type on the moisture content of the material, and the larger the weighting coefficient, the greater the degree of influence. Based on the weight coefficients corresponding to the parameter types, the parameter values ​​of each operating condition parameter in the equipment operating condition, and the parameter values ​​between each operating condition parameter in the candidate historical equipment operating conditions, the similarity between the equipment operating condition and the candidate historical equipment operating conditions is determined.

5. The method for adjusting parameters of the water content prediction model according to claim 2, characterized in that, The cloud-based device determines the similarity between the device's operating condition and historical device operating conditions in the knowledge base, including: Determine the discretized parameter level label corresponding to each operating condition parameter in the historical equipment operating conditions, wherein the discretized parameter level label is used to indicate the preset parameter value range in which the parameter value of the operating condition parameter is located; The value at the center point of the preset parameter value range is determined to be the parameter value of the operating condition parameter in the historical equipment operating conditions; Based on the parameter values ​​of the operating parameters in the current equipment condition and the parameter values ​​of the operating parameters in the historical equipment condition, the similarity between the current equipment condition and the historical equipment condition is determined.

6. The method for adjusting parameters of the water content prediction model according to claim 2, characterized in that, Before adjusting the set of model parameters corresponding to the second historical equipment operating condition according to the model parameter adjustment method, the method further includes: Determine the discretized parameter level label corresponding to each operating condition parameter in the second historical equipment operating condition, wherein the discretized parameter level label is used to indicate the preset parameter value range in which the parameter value of the operating condition parameter is located; Based on the parameter type, determine the range of preset parameter values ​​corresponding to each operating condition parameter in the equipment operating conditions; If the parameter value of the target operating condition parameter in the equipment operating condition is not within the corresponding preset parameter value range, it is determined that there is no second historical equipment operating condition corresponding to the equipment operating condition, wherein the target operating condition parameter is an operating condition parameter of type preset parameter type; If, in the equipment operating condition, there are more than a preset number of non-target operating condition parameters whose parameter values ​​are not within the corresponding preset parameter value range, it is determined that there is no second historical equipment operating condition corresponding to the equipment operating condition. Here, the non-target operating condition parameters are operating condition parameters whose type is not the preset parameter type, and the influence of the target operating condition parameters on the moisture content of the material is greater than the influence of the non-target operating condition parameters on the moisture content of the material. If, in the equipment operating condition, there is a non-target operating condition parameter that is not within the corresponding preset parameter range and whose deviation from the corresponding preset parameter range is greater than a preset deviation threshold, it is determined that there is no second historical equipment operating condition corresponding to the equipment operating condition.

7. The method for adjusting parameters of the water content prediction model according to claim 1, characterized in that, The method further includes: The optical line terminal sends the model parameter adjustment method to the optical network unit via the downlink, wherein the model parameter adjustment method is used to determine the equipment parameter adjustment method for the target drying equipment.

8. The method for adjusting parameters of the water content prediction model according to claim 1, characterized in that, The method further includes: The optical line terminal acquires the device operating conditions collected by each sensor through the optical network unit, wherein the sensor and the optical network unit are directly connected.

9. A parameter adjustment device for a moisture content prediction model, suitable for cloud-based devices, characterized in that, include: The first processing module is used to send a knowledge base to the optical line terminal. The knowledge base includes the historical operating conditions of multiple target drying equipment and the model parameter set of the moisture content prediction model corresponding to the historical operating conditions. The moisture content prediction model is set in the optical line terminal and is used to predict the moisture content of the materials in the target drying equipment. The second processing module is used to receive the equipment operating status of the target drying equipment sent by the optical line terminal through the uplink, and the model parameter adjustment method corresponding to the equipment operating status, wherein the model parameter adjustment method is an adjustment method determined by the optical line terminal based on the difference between the moisture content prediction value and the measured moisture content value of the moisture content prediction model. The third processing module is used to update the set of model parameters corresponding to the historical operating conditions in the knowledge base according to the equipment operating conditions and the model parameter adjustment method corresponding to the equipment operating conditions.

10. A non-volatile storage medium, characterized in that, The non-volatile storage medium stores a program, wherein when the program is executed, it controls the device containing the non-volatile storage medium to perform the water content prediction model parameter adjustment method according to any one of claims 1 to 8.

11. An electronic device, characterized in that, include: A memory and a processor, the processor being configured to run a program stored in the memory, wherein the program, when running, executes the water content prediction model parameter adjustment method according to any one of claims 1 to 8.

12. A computer program product, characterized in that, The method includes a computer program that, when executed by a processor, implements the method for adjusting parameters of a water content prediction model according to any one of claims 1 to 8.

Images