Automatic discharging system

Abstract

The invention discloses an automatic discharging system, which belongs to the technical field of automatic control, and is characterized in that a plurality of edge computing nodes are arranged, then the edge computing nodes are used for identifying parameters of a drying machine acquired by data sensing nodes in real time, and the drying machine is controlled according to an identification result, so that the parameters can be set in an optimal mode; according to the method, the monitoring effect on the drying process is improved, the later operation and maintenance cost is reduced, then the water content prediction model is pre-trained by adopting a powerful algorithm of the cloud computing center, and the water content prediction model is deployed on the edge computing node, so that the data processing pressure of the edge node can be effectively reduced, and the discharging process control is more accurate.

Description

Technical Field

[0001] The invention belongs to the technical field of automatic control, and in particular relates to an automatic material discharging system. Background Art

[0002] At present, how to manage the sludge generated by sewage treatment plants is one of the major challenges facing environmental protection departments. Whether it is traditional treatment (landfill) or turning waste into treasure (agricultural use, building materials), sludge drying is an important link. With the increasing amount of sludge, sludge drying technology is also constantly seeking breakthroughs and innovations. At present, industrial sludge drying equipment mainly adopts offline stand-alone control mode. Although the operation is simple, the offline single control mode will inevitably have problems such as unreasonable equipment parameter settings, low efficiency of manual monitoring, and high operation and maintenance costs. Summary of the invention

[0003] The present invention provides an automatic discharging system, which is used to solve the problems of low control efficiency and high operation and maintenance cost of a drying machine in the prior art.

[0004] An automatic unloading system, comprising: an edge computing node, a cloud computing center, and a data sensor node corresponding to each edge computing node;

[0005] The data sensor node is used to collect basic operating parameters of the dryer and transmit the basic operating parameters to the corresponding edge computing node;

[0006] The edge computing node is used to accept the control of the cloud computing center to deploy the moisture content prediction model, and control the moisture content of the dryer discharge based on the deployed moisture content prediction model and basic operating parameters;

[0007] Among them, the dryer discharge is used to characterize the discharge of materials from the dryer;

[0008] The cloud computing center is used to pre-train a moisture content prediction model using historical sample data, and deploy the moisture content prediction model on an edge computing node.

[0009] Furthermore, the basic operating parameters of the dryer include part or all of the drying chamber inlet air volume, drying chamber outlet air temperature, wet material moisture content, drying chamber drying temperature, drying chamber relative humidity, water removal amount, and baffle lifting height of the discharge port.

[0010] Furthermore, the edge computing node includes a model deployment module, a control parameter adjustment module and a moisture content prediction module;

[0011] The model deployment module is used to accept the control of the cloud computing center to deploy the moisture content prediction model;

[0012] The moisture content prediction module is used to use the basic operating parameters as input of the moisture content prediction model to obtain the predicted moisture content;

[0013] The control parameter adjustment module is used to obtain the target moisture content set by the staff through human-computer interaction, and determine whether the error between the target moisture content and the predicted moisture content is less than a preset threshold value. If so, the current discharge control parameters of the dryer are maintained and the discharge operation is continued. Otherwise, the annealing genetic algorithm is used to adjust the discharge control parameters in real time until the error between the target moisture content and the predicted moisture content is less than the preset threshold value, and the target control parameters are obtained, and the dryer is controlled by the target control parameters.

[0014] Furthermore, the annealing genetic algorithm is used to adjust the discharge control parameters in real time until the error between the target moisture content and the predicted moisture content is less than a preset threshold, and the target control parameters are obtained, and the dryer is controlled by the target control parameters, including:

[0015] Determine the control parameters to be optimized; wherein the control parameters to be optimized include the air volume at the drying chamber inlet, the drying chamber drying temperature, and the baffle lifting height of the discharge port;

[0016] Performing random initialization between the upper limit and the lower limit of the control parameter to be optimized, and obtaining multiple parameter codes corresponding to the control parameter to be optimized; wherein the value corresponding to each parameter code is different;

[0017] Based on the parameter code, the target operating parameter with the highest similarity is matched in the historical basic operating parameters, and the target operating parameter is used as the input of the water content prediction model to obtain the actual water content;

[0018] According to the actual moisture content and the target moisture content, obtain the fitness value corresponding to each parameter code;

[0019] According to the fitness value corresponding to each parameter encoding, the initial temperature of simulated annealing is initialized to T0;

[0020] According to the fitness value corresponding to each parameter code, the parameter code is updated by using an adaptive crossover strategy, an adaptive mutation strategy, and an adaptive selection strategy to determine the update value corresponding to the parameter code;

[0021] When the fitness corresponding to the update value corresponding to the parameter encoding increases, the update is accepted; when the fitness corresponding to the update value corresponding to the parameter encoding decreases, the adaptive selection probability is used to select whether to accept the update;

[0022] Determine whether the current annealing temperature is less than the preset temperature threshold. If so, select the parameter code with the highest fitness as the target control parameter, and control the dryer through the target control parameter. Otherwise, perform annealing operation and return to the step of obtaining the updated value corresponding to the parameter code.

[0023] Furthermore, according to the actual moisture content and the target moisture content, the fitness value corresponding to each parameter code is obtained as follows:

[0024] Get the error value between the actual moisture content and the target moisture content as E;

[0025] According to the error value E, the fitness value is obtained as f=1 / (E+0.01);

[0026] Traverse all parameter encodings and determine the fitness value corresponding to each parameter encoding.

[0027] Furthermore, according to the fitness value corresponding to each parameter encoding, the initial temperature of simulated annealing is initialized to T0, including:

[0028] According to the fitness value corresponding to each parameter code, the parameter code with the largest fitness value is determined as the current optimal parameter code;

[0029] According to the fitness value corresponding to the current optimal parameter encoding, the initial temperature of simulated annealing is determined as:

[0030]

[0031] in, Indicates the fitness value corresponding to the current optimal parameter encoding during the first cycle.

[0032] Further, according to the fitness value corresponding to each parameter code, the parameter code is updated by using an adaptive crossover strategy, an adaptive mutation strategy, and an adaptive selection strategy to determine the update value corresponding to the parameter code, including:

[0033] According to the fitness value corresponding to each parameter encoding, the adaptive crossover probability and adaptive mutation probability are determined as follows:

[0034]

[0035] Among them, pc i represents the crossover probability corresponding to the i-th parameter encoding, i = 1, 2, ..., I, I represents the total number of parameter encodings, represents the fitness value corresponding to the current optimal parameter encoding in the tth cycle, f i t Indicates the fitness value corresponding to the i-th parameter encoding during the t-th cycle, represents the average fitness value corresponding to all parameter encodings in the tth cycle, α1 represents the first constant, α2 represents the second constant, and the first constant is less than the second constant, pm i represents the mutation probability corresponding to the i-th parameter encoding, α3 represents the third constant, α4 represents the fourth constant, and the third constant is less than the fourth constant;

[0036] Based on the adaptive crossover probability, the roulette wheel method is used to crossover each parameter encoding to obtain the first offspring after the crossover;

[0037] Based on the adaptive mutation probability, the roulette wheel method is used to mutate each parameter encoding to obtain the second offspring after mutation;

[0038] According to the first offspring after crossover and the second offspring after mutation, a selection operation is performed on each parameter individual to obtain an updated value corresponding to the parameter encoding.

[0039] Furthermore, based on the adaptive crossover probability, a roulette wheel method is used to crossover each parameter encoding to obtain the first offspring after the crossover, including:

[0040] Based on the adaptive crossover probability, a roulette wheel method is used to determine whether crossover is needed. If so, crossover is performed, otherwise no crossover is performed.

[0041] Cross-cutting process, including:

[0042] For any parameter encoding, determine the first historical optimal value corresponding to the parameter encoding, and randomly select a second historical optimal value corresponding to another parameter encoding;

[0043] When the fitness of the first historical optimal value is less than the second historical optimal value, the parameter encoding is crossed as follows:

[0044]

[0045]

[0046] in, represents the i-th parameter encoding during the t-th cycle, β represents the adaptive cross coefficient, represents the current optimal parameter encoding, represents the first offspring after crossover, f i t Indicates the fitness value corresponding to the i-th parameter encoding, represents the fitness value corresponding to the current optimal parameter encoding, and λ represents a constant term between (0,1);

[0047] When the fitness of the first historical optimal value is greater than that of the second historical optimal value, the second historical optimal value is used as the first child after crossover.

[0048] Furthermore, based on the adaptive mutation probability, the roulette wheel method is used to mutate each parameter encoding to obtain the second offspring after mutation, including:

[0049] Based on the adaptive mutation probability, the roulette wheel method is used to determine whether mutation is needed. If so, mutation is performed, otherwise no mutation is performed;

[0050] The mutation process includes:

[0051] Randomly generate a random number A uniformly distributed between (0,1);

[0052] When the random number A is greater than 0.5, the parameter encoding cross is:

[0053]

[0054] in, represents the jth parameter code in the tth cycle, j = 1, 2, ..., I, I represents the total number of parameter codes, UB represents the upper limit code corresponding to the parameter code; tamx represents the maximum number of cycles, that is, the number of times required to reach the preset temperature threshold from the initial temperature; Represents the second offspring after mutation;

[0055] When the random number A is less than or equal to 0.5, the parameter encoding cross is:

[0056]

[0057] Wherein, LB represents the lower limit code corresponding to the parameter code.

[0058] Furthermore, an adaptive selection probability is used to select whether to accept the update, including:

[0059] Get the adaptive selection probability as:

[0060]

[0061] Among them, ps i represents the adaptive selection probability corresponding to the i-th parameter encoding during the t-th cycle, e represents the natural constant, k represents the Boltzmann constant, fit represents the fitness value corresponding to the i-th parameter encoding during the t-th cycle, fi,t represents the fitness value corresponding to the updated value corresponding to the i-th parameter encoding, and T represents the current annealing temperature.

[0062] An automatic discharging system provided by the present invention sets multiple edge computing nodes, and then uses the edge computing nodes to identify the parameters of the dryer collected by the data sensor nodes in real time, and controls the dryer according to the identification results. The parameters can be set in the best way, the monitoring effect of the drying process can be improved, and the subsequent operation and maintenance costs can be reduced. Secondly, the powerful algorithm of the cloud computing center is used to pre-train the moisture content prediction model, and the moisture content prediction model is deployed on the edge computing nodes, which can effectively reduce the data processing pressure of the edge nodes and make the discharging process control more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0064] Figure 1 A schematic structural diagram of an automatic material discharging system provided in an embodiment of the present invention.

[0065] Figure 2 A schematic diagram of the structure of an edge computing node provided in an embodiment of the present invention.

[0066] The above drawings have shown clear embodiments of the present invention, which will be described in more detail below. These drawings and text descriptions are not intended to limit the scope of the present invention in any way, but to illustrate the concept of the present invention for those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0067] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0068] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0069] like Figure 1 As shown, an embodiment of the present invention provides an automatic unloading system, including: an edge computing node, a cloud computing center, and a data sensor node corresponding to each edge computing node;

[0070] The data sensor node is used to collect basic operating parameters of the dryer and transmit the basic operating parameters to the corresponding edge computing node;

[0071] The data sensing node can be connected to the IoT sensor or the control system of the dryer to obtain parameters, and the basic operating parameters of the dryer can be collected according to the preset data resampling frequency. When the data sampling frequency is high, the dryer can be better monitored and controlled, but it will increase the data transmission pressure and data processing pressure. When the data sampling frequency is low, it can effectively reduce the data processing pressure and data transmission pressure of the edge node, but the control effect will also be reduced. Therefore, the staff can set the data sampling frequency according to actual needs.

[0072] The edge computing node is used to accept the control of the cloud computing center to deploy the moisture content prediction model, and control the moisture content of the dryer discharge based on the deployed moisture content prediction model and basic operating parameters;

[0073] Among them, the dryer discharge is used to characterize the discharge of materials from the dryer;

[0074] Generally speaking, the data processing capabilities of edge computing nodes are relatively general, and the training effect of the moisture content prediction model is relatively general. Therefore, the embodiment of the present invention adopts a cloud computing center for pre-training, and then deploys the pre-trained moisture content prediction model. Then, the edge computing node can predict the moisture content according to the deployed moisture content prediction model, and perform real-time detection and control of the dryer according to the set target moisture content and the predicted moisture content, thereby effectively improving the monitoring efficiency of the dryer and reducing the subsequent operation and maintenance costs.

[0075] The cloud computing center is used to pre-train a moisture content prediction model using historical sample data, and deploy the moisture content prediction model on an edge computing node.

[0076] Optionally, the moisture content prediction model may be trained in advance using an intelligent optimization algorithm so that the moisture content prediction model can predict the moisture content according to basic operating parameters.

[0077] An automatic material discharge system provided by the present invention sets multiple edge computing nodes, and then uses the edge computing nodes to identify the parameters of the dryer collected by the data sensor nodes in real time, and controls the dryer according to the identification results. The parameters can be set in the best way, the monitoring effect of the drying process can be improved, and the subsequent operation and maintenance costs can be reduced. Secondly, the powerful algorithm of the cloud computing center is used to pre-train the moisture content prediction model, and the moisture content prediction model is deployed on the edge computing nodes, which can effectively reduce the data processing pressure of the edge nodes and provide more accurate material discharge services.

[0078] In an embodiment of the present invention, the basic operating parameters of the dryer include part or all of the drying chamber inlet air volume, the drying chamber outlet air temperature, the moisture content of the wet material, the drying chamber drying temperature, the drying chamber relative humidity, the water removal amount, and the baffle lifting height of the discharge port.

[0079] Optionally, the preferred basic operating parameters of the embodiment of the present invention are the drying chamber inlet air volume, the drying chamber outlet air temperature, the moisture content of the wet material, the drying chamber drying temperature, the drying chamber relative humidity, and the baffle lifting height of the discharge port. These parameters can be obtained by measurement or docking with the dryer control system, which is more convenient.

[0080] like Figure 2 As shown, the edge computing node includes a model deployment module, a control parameter adjustment module and a water content prediction module;

[0081] The model deployment module is used to accept the control of the cloud computing center to deploy the moisture content prediction model;

[0082] The moisture content prediction module is used to use the basic operating parameters as input of the moisture content prediction model to obtain the predicted moisture content;

[0083] The control parameter adjustment module is used to obtain the target moisture content set by the staff through human-computer interaction, and determine whether the error between the target moisture content and the predicted moisture content is less than a preset threshold value. If so, the current discharge control parameters of the dryer are maintained and the discharge operation is continued. Otherwise, the annealing genetic algorithm is used to adjust the discharge control parameters in real time until the error between the target moisture content and the predicted moisture content is less than the preset threshold value, and the target control parameters are obtained, and the dryer is controlled by the target control parameters.

[0084] The embodiment of the present invention can realize automatic regulation of the dryer by setting a control parameter adjustment module, thereby controlling the moisture content of the sludge and improving the automation level.

[0085] Optionally, using the annealing genetic algorithm to adjust the discharge control parameters in real time is only a preferred implementation of the embodiment of the present invention, and other algorithms can also be used to adjust these control parameters to achieve intelligent control.

[0086] In an embodiment of the present invention, an annealing genetic algorithm is used to adjust the discharge control parameters in real time until the error between the target moisture content and the predicted moisture content is less than a preset threshold, and the target control parameters are obtained, and the dryer is controlled by the target control parameters, including:

[0087] Determine the control parameters to be optimized; wherein the control parameters to be optimized include the air volume at the drying chamber inlet, the drying chamber drying temperature, and the baffle lifting height of the discharge port;

[0088] It is worth mentioning that, in addition to the drying chamber inlet air volume, the drying chamber drying temperature and the baffle lifting height of the discharge port, other parameters of the dryer can also be used as control parameters to be optimized.

[0089] Performing random initialization between the upper limit and the lower limit of the control parameter to be optimized, and obtaining multiple parameter codes corresponding to the control parameter to be optimized; wherein the value corresponding to each parameter code is different;

[0090] Based on the parameter code, the target operating parameter with the highest similarity is matched in the historical basic operating parameters, and the target operating parameter is used as the input of the water content prediction model to obtain the actual water content;

[0091] Since manual measurement of samples is time-consuming and laborious, it is impossible to summarize all corresponding relationships. Therefore, the embodiment of the present invention uses manual measurement samples for training to obtain a relationship between an operating parameter and water content, and then constructs a complete input through historical basic operating parameters so that the water content prediction model can accurately identify data.

[0092] Based on the parameter code, matching the target operating parameter with the highest similarity among the historical basic operating parameters may include:

[0093] The Euclidean distances between the three parameters corresponding to the parameter code and the historical basic operating parameters are obtained, and the historical basic operating parameter with the smallest Euclidean distance is used as the target operating parameter with the highest similarity.

[0094] According to the actual moisture content and the target moisture content, obtain the fitness value corresponding to each parameter code;

[0095] According to the fitness value corresponding to each parameter encoding, the initial temperature of simulated annealing is initialized to T0;

[0096] According to the fitness value corresponding to each parameter code, the parameter code is updated by using an adaptive crossover strategy, an adaptive mutation strategy, and an adaptive selection strategy to determine the update value corresponding to the parameter code;

[0097] When the fitness corresponding to the update value corresponding to the parameter encoding increases, the update is accepted; when the fitness corresponding to the update value corresponding to the parameter encoding decreases, the adaptive selection probability is used to select whether to accept the update;

[0098] Determine whether the current annealing temperature is less than the preset temperature threshold. If so, select the parameter code with the highest fitness as the target control parameter, and control the dryer through the target control parameter. Otherwise, perform annealing operation and return to the step of obtaining the updated value corresponding to the parameter code.

[0099] It is difficult to search for the best parameter combination in the prior art, so the embodiment of the present invention provides an annealing genetic algorithm to solve the technical problem of low search efficiency in the prior art.

[0100] In the embodiment of the present invention, the fitness value corresponding to each parameter code is obtained according to the actual moisture content and the target moisture content:

[0101] Get the error value between the actual moisture content and the target moisture content as E;

[0102] According to the error value E, the fitness value is obtained as f=1 / (E+0.01);

[0103] Traverse all parameter encodings and determine the fitness value corresponding to each parameter encoding.

[0104] In the embodiment of the present invention, according to the fitness value corresponding to each parameter encoding, the initial temperature of the simulated annealing is initialized to T0, including:

[0105] According to the fitness value corresponding to each parameter code, the parameter code with the largest fitness value is determined as the current optimal parameter code;

[0106] According to the fitness value corresponding to the current optimal parameter encoding, the initial temperature of simulated annealing is determined as:

[0107]

[0108] in, Indicates the fitness value corresponding to the current optimal parameter encoding during the first cycle.

[0109] In the embodiment of the present invention, according to the fitness value corresponding to each parameter code, an adaptive crossover strategy, an adaptive mutation strategy, and an adaptive selection strategy are used to update the parameter code to determine the update value corresponding to the parameter code, including:

[0110] According to the fitness value corresponding to each parameter encoding, the adaptive crossover probability and adaptive mutation probability are determined as follows:

[0111]

[0112] Among them, pc i represents the crossover probability corresponding to the i-th parameter encoding, i = 1, 2, ..., I, I represents the total number of parameter encodings, represents the fitness value corresponding to the current optimal parameter encoding in the tth cycle, f i t Indicates the fitness value corresponding to the i-th parameter encoding during the t-th cycle, represents the average fitness value corresponding to all parameter encodings in the tth cycle, α1 represents the first constant, α2 represents the second constant, and the first constant is less than the second constant, pm i represents the mutation probability corresponding to the i-th parameter encoding, α3 represents the third constant, α4 represents the fourth constant, and the third constant is less than the fourth constant;

[0113] The crossover process is one of the key operations in the genetic algorithm. Through the crossover operation, a new generation of individuals is obtained, and the new individuals have the characteristics of the parent individuals. The individuals in the group are randomly paired, and for each individual, part of the chromosomes between them are exchanged with a certain probability (crossover probability). The greater the crossover probability, the more and faster the new individuals generated by crossover. If the crossover probability is too large, there is a high probability that excellent individuals will be destroyed; if the probability is too small, the search process will become slow. Therefore, the crossover probability should be based on the adaptive parameters of the individual fitness. For individuals with a fitness higher than the average fitness, the crossover probability should be reduced so that the individuals can be protected; for individuals with a fitness lower than the average fitness, the crossover probability should be increased so that the individuals are eliminated with a high probability.

[0114] Mutation is to change the gene value of each individual in the group to the corresponding allele with a low probability (called mutation probability). If the mutation probability is too small, the chance of generating new individuals is too small; if the mutation probability is too large, the genetic algorithm will be too random. Therefore, the mutation probability, like the crossover probability, must also be based on the adaptive parameters of individual fitness. For individuals with fitness higher than the average fitness, the mutation probability is reduced, which reduces calculations while ensuring population diversity and speeds up the algorithm; for individuals with fitness lower than the average fitness, the mutation probability is increased, which to a certain extent avoids evolutionary stagnation and premature convergence to the local optimum.

[0115] Based on the adaptive crossover probability, the roulette wheel method is used to crossover each parameter encoding to obtain the first offspring after the crossover;

[0116] Based on the adaptive mutation probability, the roulette wheel method is used to mutate each parameter encoding to obtain the second offspring after mutation;

[0117] According to the first offspring after crossover and the second offspring after mutation, a selection operation is performed on each parameter individual to obtain an updated value corresponding to the parameter encoding.

[0118] In the embodiment of the present invention, based on the adaptive crossover probability, a roulette wheel method is used to crossover each parameter code to obtain the first offspring after the crossover, including:

[0119] Based on the adaptive crossover probability, a roulette wheel method is used to determine whether crossover is needed. If so, crossover is performed, otherwise no crossover is performed.

[0120] Cross-cutting process, including:

[0121] For any parameter encoding, determine the first historical optimal value corresponding to the parameter encoding, and randomly select a second historical optimal value corresponding to another parameter encoding;

[0122] When the fitness of the first historical optimal value is less than the second historical optimal value, the parameter encoding is crossed as follows:

[0123]

[0124] in, represents the i-th parameter encoding during the t-th cycle, β represents the adaptive cross coefficient, represents the current optimal parameter encoding, represents the first offspring after crossover, Indicates the fitness value corresponding to the i-th parameter encoding, represents the fitness value corresponding to the current optimal parameter encoding, and λ represents a constant term between (0,1);

[0125] When the fitness of the first historical optimal value is greater than that of the second historical optimal value, the second historical optimal value is used as the first child after crossover.

[0126] By crossing the encoded historical optimal position with the current optimal parameter encoding, we can effectively avoid falling into the local optimum, thereby improving the convergence effect and speed of the algorithm.

[0127] In the embodiment of the present invention, based on the adaptive mutation probability, a roulette wheel method is used to mutate each parameter encoding to obtain a second offspring after mutation, including:

[0128] Based on the adaptive mutation probability, the roulette wheel method is used to determine whether mutation is needed. If so, mutation is performed, otherwise no mutation is performed;

[0129] The mutation process includes:

[0130] Randomly generate a random number A uniformly distributed between (0,1);

[0131] When the random number A is greater than 0.5, the parameter encoding cross is:

[0132]

[0133] in, represents the jth parameter code in the tth cycle, j = 1, 2, ..., I, I represents the total number of parameter codes, UB represents the upper limit code corresponding to the parameter code; tamx represents the maximum number of cycles, that is, the number of times required to reach the preset temperature threshold from the initial temperature; Represents the second offspring after mutation;

[0134] When the random number A is less than or equal to 0.5, the parameter encoding cross is:

[0135]

[0136] Wherein, LB represents the lower limit code corresponding to the parameter code.

[0137] In the embodiment of the present invention, the adaptive selection probability is used to select whether to accept the update, including:

[0138] Get the adaptive selection probability as:

[0139]

[0140] Among them, ps i represents the adaptive selection probability corresponding to the i-th parameter encoding in the t-th cycle, e represents the natural constant, k represents the Boltzmann constant, and f i t Indicates the fitness value corresponding to the i-th parameter encoding during the t-th cycle, represents the fitness value corresponding to the updated value corresponding to the i-th parameter encoding, and T represents the current annealing temperature.

[0141] Optionally, using historical sample data to pre-train a moisture content prediction model and deploying the moisture content prediction model on an edge computing node may include:

[0142] The RBF (Radial Basis Function) neural network is used as the moisture content prediction model, and the network parameters of the moisture content prediction model are initialized;

[0143] Obtaining historical sample data; wherein the historical sample data includes historical basic operating parameters and corresponding discharge moisture content, the moisture content can be obtained by manual measurement, after which a data relationship can be formed to facilitate subsequent data prediction;

[0144] The historical basic operating parameters are used as the actual input of the moisture content prediction model to obtain the predicted output of the moisture content prediction model, and multiple historical sample data are used to obtain the root mean square error value between the predicted output and the moisture content of the discharge;

[0145] According to the error value, the network parameters of the water content prediction model are adjusted as follows:

[0146]

[0147] in, represents the adjusted network weight, that is, the network weight corresponding to the jth neuron in the hidden layer during the t+1th training; represents the network weight before adjustment, that is, the network weight corresponding to the jth neuron in the hidden layer during the tth training; η w represents the first learning rate; It represents the root mean square error value of the moisture content prediction model during the t-th training; Indicates that during the tth training, the jth neuron in the hidden layer of the moisture content prediction model responds to the input historical sample data x n Response; β w represents the first operation parameter value corresponding to the network weight, and β w The value range of is [0,1]; Indicates the adjustment amount corresponding to the network weight during the tth training; represents the network weight corresponding to the jth neuron in the hidden layer during the t-1th training; when t = 1, represents the optimal network weight; j = 1, 2, ..., J, J represents the total number of neurons in the moisture content prediction model; represents the adjusted network center, that is, the network center corresponding to the jth neuron in the hidden layer during the t+1th training; represents the network center before adjustment, that is, the network center corresponding to the jth neuron in the hidden layer during the tth training; η c represents the second learning rate; β c represents the second operation parameter value corresponding to the network center, and β c The value range of is [0,1]; Indicates the adjustment amount corresponding to the network center during the tth training, Indicates the network center corresponding to the jth neuron in the hidden layer during the t-1th training; when t = 1, represents the optimal network center; represents the adjusted network width, that is, the network width corresponding to the jth neuron in the hidden layer during the t+1th training; represents the network width before adjustment, that is, the network width corresponding to the jth neuron in the hidden layer during the tth training; η σ represents the third learning rate, β c represents the third operation parameter value corresponding to the network width, and β c The value range of is [0,1]; Indicates the adjustment amount corresponding to the network width during the tth training; represents the network width corresponding to the jth neuron in the hidden layer during the t-1th training; when t=1, represents the optimal network width;

[0148] The root mean square error value is re-obtained, and it is determined whether the re-obtained root mean square error value is less than the preset mean square error threshold. If so, the trained moisture content prediction model is obtained, and the trained moisture content prediction model is deployed on the edge computing node, otherwise the network parameters are re-adjusted.

[0149] Optionally, after obtaining the trained moisture content prediction model, the trained moisture content prediction model may be pruned, and the pruned moisture content prediction model may be deployed on an edge computing node.

[0150] Those skilled in the art will readily appreciate other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary techniques in the art that are not disclosed by the present invention. It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims

1. An automatic unloading system, characterized in that: include: Edge computing nodes, cloud computing centers, and data sensor nodes corresponding to each edge computing node; The data sensor node is used to collect basic operating parameters of the dryer and transmit the basic operating parameters to the corresponding edge computing node; The edge computing node is used to accept the control of the cloud computing center to deploy the moisture content prediction model, and control the moisture content of the dryer discharge based on the deployed moisture content prediction model and basic operating parameters; Among them, the dryer discharge is used to characterize the discharge of materials from the dryer; The cloud computing center is used to pre-train a moisture content prediction model using historical sample data, and deploy the moisture content prediction model on an edge computing node.

2. The automatic unloading system according to claim 1, characterized in that: The basic operating parameters of the dryer include part or all of the drying chamber inlet air volume, drying chamber outlet air temperature, wet material moisture content, drying chamber drying temperature, drying chamber relative humidity, water removal volume, and baffle lifting height of the discharge port.

3. The automatic unloading system according to claim 1, characterized in that: The edge computing node includes a model deployment module, a control parameter adjustment module and a moisture content prediction module; The model deployment module is used to accept the control of the cloud computing center to deploy the moisture content prediction model; The moisture content prediction module is used to use the basic operating parameters as input of the moisture content prediction model to obtain the predicted moisture content; The control parameter adjustment module is used to obtain the target moisture content set by the staff through human-computer interaction, and determine whether the error between the target moisture content and the predicted moisture content is less than a preset threshold value. If so, the current discharge control parameters of the dryer are maintained and the discharge operation is continued. Otherwise, the annealing genetic algorithm is used to adjust the discharge control parameters in real time until the error between the target moisture content and the predicted moisture content is less than the preset threshold value, and the target control parameters are obtained, and the dryer is controlled by the target control parameters.

4. The automatic unloading system according to claim 3, characterized in that: The annealing genetic algorithm is used to adjust the discharge control parameters in real time until the error between the target moisture content and the predicted moisture content is less than the preset threshold, and the target control parameters are obtained. The dryer is controlled by the target control parameters, including: Determine the control parameters to be optimized; wherein the control parameters to be optimized include the air volume at the drying chamber inlet, the drying chamber drying temperature, and the baffle lifting height of the discharge port; Performing random initialization between the upper limit and the lower limit of the control parameter to be optimized, and obtaining multiple parameter codes corresponding to the control parameter to be optimized; wherein the value corresponding to each parameter code is different; Based on the parameter coding, the target operating parameter with the highest similarity is matched in the historical basic operating parameters, and the target operating parameter is used as the input of the water content prediction model to obtain the actual water content; According to the actual moisture content and the target moisture content, obtain the fitness value corresponding to each parameter code; According to the fitness value corresponding to each parameter encoding, the initial temperature of simulated annealing is initialized to T0; According to the fitness value corresponding to each parameter code, the parameter code is updated by using an adaptive crossover strategy, an adaptive mutation strategy, and an adaptive selection strategy to determine the update value corresponding to the parameter code; When the fitness corresponding to the update value corresponding to the parameter encoding increases, the update is accepted; when the fitness corresponding to the update value corresponding to the parameter encoding decreases, the adaptive selection probability is used to select whether to accept the update; Determine whether the current annealing temperature is less than the preset temperature threshold. If so, select the parameter code with the highest fitness as the target control parameter, and control the dryer through the target control parameter. Otherwise, perform annealing operation and return to the step of obtaining the updated value corresponding to the parameter code.

5. The automatic unloading system according to claim 4, characterized in that: According to the actual moisture content and the target moisture content, the fitness value corresponding to each parameter encoding is obtained: Get the error value between the actual moisture content and the target moisture content as E; According to the error value E, the fitness value is obtained as f=1 / (E+0.01); Traverse all parameter encodings and determine the fitness value corresponding to each parameter encoding.

6. The automatic unloading system according to claim 4, characterized in that: According to the fitness value corresponding to each parameter encoding, the initial temperature of simulated annealing is initialized to T0, including: According to the fitness value corresponding to each parameter code, the parameter code with the largest fitness value is determined as the current optimal parameter code; According to the fitness value corresponding to the current optimal parameter encoding, the initial temperature of simulated annealing is determined as: in, Indicates the fitness value corresponding to the current optimal parameter encoding during the first cycle.

7. The automatic unloading system according to claim 6, characterized in that: According to the fitness value corresponding to each parameter code, the parameter code is updated by using an adaptive crossover strategy, an adaptive mutation strategy, and an adaptive selection strategy to determine the update value corresponding to the parameter code, including: According to the fitness value corresponding to each parameter encoding, the adaptive crossover probability and adaptive mutation probability are determined as follows: Among them, pc i represents the crossover probability corresponding to the i-th parameter encoding, i = 1, 2, ..., I, I represents the total number of parameter encodings, Indicates the fitness value corresponding to the current optimal parameter encoding during the t-th cycle, Indicates the fitness value corresponding to the i-th parameter encoding during the t-th cycle, represents the average fitness value corresponding to all parameter encodings in the tth cycle, α1 represents the first constant, α2 represents the second constant, and the first constant is less than the second constant, pm i represents the mutation probability corresponding to the i-th parameter encoding, α3 represents the third constant, α4 represents the fourth constant, and the third constant is less than the fourth constant; Based on the adaptive crossover probability, the roulette wheel method is used to crossover each parameter encoding to obtain the first offspring after the crossover; Based on the adaptive mutation probability, the roulette wheel method is used to mutate each parameter encoding to obtain the second offspring after mutation; According to the first offspring after crossover and the second offspring after mutation, a selection operation is performed on each parameter individual to obtain an updated value corresponding to the parameter encoding.

8. The automatic unloading system according to claim 7, characterized in that: Based on the adaptive crossover probability, the roulette wheel method is used to crossover each parameter encoding to obtain the first offspring after the crossover, including: Based on the adaptive crossover probability, a roulette wheel method is used to determine whether crossover is needed. If so, crossover is performed, otherwise no crossover is performed. Cross-cutting process, including: For any parameter encoding, determine the first historical optimal value corresponding to the parameter encoding, and randomly select a second historical optimal value corresponding to another parameter encoding; When the fitness of the first historical optimal value is less than the second historical optimal value, the parameter encoding is crossed as follows: in, represents the i-th parameter encoding during the t-th cycle, β represents the adaptive cross coefficient, represents the current optimal parameter encoding, represents the first offspring after crossover, Indicates the fitness value corresponding to the i-th parameter encoding, represents the fitness value corresponding to the current optimal parameter encoding, and λ represents a constant term between (0,1); When the fitness of the first historical optimal value is greater than that of the second historical optimal value, the second historical optimal value is used as the first child after crossover.

9. The automatic unloading system according to claim 8, characterized in that: Based on the adaptive mutation probability, the roulette wheel method is used to mutate each parameter encoding to obtain the second offspring after mutation, including: Based on the adaptive mutation probability, the roulette wheel method is used to determine whether mutation is needed. If so, mutation is performed, otherwise no mutation is performed; The mutation process includes: Randomly generate a random number A uniformly distributed between (0,1); When the random number A is greater than 0.5, the parameter encoding cross is: in, represents the jth parameter code in the tth cycle, j = 1, 2, ..., I, I represents the total number of parameter codes, UB represents the upper limit code corresponding to the parameter code; tamx represents the maximum number of cycles, that is, the number of times required to reach the preset temperature threshold from the initial temperature; Represents the second offspring after mutation; When the random number A is less than or equal to 0.5, the parameter encoding cross is: Wherein, LB represents the lower limit code corresponding to the parameter code.

10. The automatic unloading system according to claim 4, characterized in that: Adopt adaptive selection probability to choose whether to accept the update, including: Get the adaptive selection probability as: Among them, ps i represents the adaptive selection probability corresponding to the i-th parameter encoding in the t-th cycle, e represents the natural constant, k represents the Boltzmann constant, Indicates the fitness value corresponding to the i-th parameter encoding during the t-th cycle, represents the fitness value corresponding to the updated value corresponding to the i-th parameter encoding, and T represents the current annealing temperature.

Images