A Deep Learning-Based Intelligent Control Method for Grain Storage Environment
By verifying the action path of the equipment through an improved MTGNN and a balanced motion probe, a sealing chain is formed, which solves the problem of deviation in the judgment of the action range of equipment in the grain storage environment control, and improves the accuracy and stability of the control.
Patent Information
- Application Number
- CN202611472886.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-21
- Publication Date
- 2026-11-17
AI Technical Summary
In existing grain storage environment control methods, the correlation between grain condition nodes learned by the model lacks verification of the actual action path of environmental regulation equipment, which leads to deviations in the judgment of the equipment's action range and state changes that do not match the equipment's action path after the control action is executed, affecting the accuracy and stability of the control action.
An improved MTGNN is used to construct the correlation between grain condition nodes. The action path of the equipment is verified by balancing action probes to form an equipment action verification chain. Candidate control actions are screened and actual grain condition changes are written back to ensure that the control actions propagate along the accurate path and are stable.
It reduces the interference of external environmental changes and natural heat dissipation from grain piles on the equipment's action path, improves the accuracy of the correspondence between the action range of environmental regulation equipment and the grain pile points to be regulated, and enhances the stability and continuous correction capability of grain storage environment control.
Smart Images

Figure CN123113377A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent control technology for grain storage environment, and in particular to an intelligent control method for grain storage environment based on deep learning. Background Technology
[0002] With the application of automated monitoring equipment in grain silos, the grain storage process can now continuously collect data on grain temperature, grain moisture, internal temperature and humidity, gas concentration, and the operating status of environmental control equipment. Existing control methods mostly rely on fixed thresholds, single-point detection results, or historical trends to start and stop ventilation, cooling, and dehumidification equipment. Some solutions introduce recurrent neural networks, graph neural networks, and time-series prediction models to predict changes in grain conditions and then generate equipment control commands based on the prediction results.
[0003] In existing methods, the correlations between grain condition nodes learned by the model mainly come from the correlation of historical data, lacking verification of the actual action paths of environmental regulation equipment. When changes occur in the external environment, natural heat dissipation from the grain pile, and the combined effects of different equipment, the model easily identifies synchronous changes as equipment control propagation relationships, leading to biases in the judgment of the equipment's scope of action. Existing methods also lack segment-by-segment verification of candidate correlation edges, allowing control actions to propagate along correlation paths not supported by actual responses, resulting in a lack of reliable correspondence between the grain condition nodes to be regulated and the environmental regulation equipment. The actual changes in grain condition after the execution of control actions are often directly used to update the overall model, and state changes without matching equipment action paths can easily rewrite the original propagation relationships, affecting the accuracy and stability of subsequent control actions.
[0004] Therefore, how to provide a deep learning-based intelligent control method for grain storage environment is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention
[0005] One objective of this invention is to propose an intelligent control method for grain storage environment based on deep learning. This invention uses an improved MTGNN to construct the correlation relationship between grain condition nodes, verifies the action path of equipment through balanced action probes, forms an equipment action sealing chain, and completes the screening of control actions and writing back the sealing edge. It has the advantages of accurate action path, low false control rate, and high control stability.
[0006] A method for intelligent control of grain storage environment based on deep learning according to an embodiment of the present invention includes the following steps: The grain storage environment status is obtained, and an improved MTGNN is used to map the grain condition detection location and the location of the environmental regulation equipment to grain condition nodes, generating a candidate association graph of grain condition nodes and determining the grain condition nodes to be regulated. The control environment regulation equipment performs a balancing action probe consisting of continuous micro-movements in opposite directions, with the cumulative adjustment amount returning to zero; Track the balance action probe along the candidate correlation graph of grain condition nodes, and identify the grain condition nodes that reproduce the sequence of continuous micro-actions as response witness nodes; According to the arrival order of the response witness nodes, the continuous candidate association edges that connect the response witness nodes sequentially from the grain condition node corresponding to the location of the environmental regulation equipment are converted into sealing edges to form the equipment action sealing chain. Generate candidate control actions for the grain plot points to be regulated. After binding the candidate control actions with the source of the corresponding balance action probe, push them into the equipment action sealing chain. The action only propagates along the sealing edge that is consistent with the source of the balance action probe bound to the candidate control action and has the same propagation direction and arrival order. Propagation stops when encountering an unsealed candidate associated edge. The target control action is determined from the candidate control actions that run along the equipment's sealing chain to the grain condition point to be regulated, and the operation of the environmental regulation equipment is controlled. Obtain actual changes in grain conditions, write back the actual changes in grain conditions that match the sealing chain of the equipment action to the sealing edge, isolate the actual changes in grain conditions that do not match from the sealing edge update, and update the sealing chain of the equipment action.
[0007] Optionally, determining the grain condition node to be regulated specifically includes: Input the grain storage environment status into the improved MTGNN to generate candidate association edges with propagation direction between grain condition nodes; The state that exceeds the grain storage safety boundary is transmitted in reverse along the candidate associated edge to the grain condition node corresponding to the location of the environmental regulation equipment, forming a reverse transmission chain; Freeze the state of the grain situation nodes outside the reverse propagation chain, and exchange the arrival order of adjacent propagation states within the reverse propagation chain. The state that exceeds the grain storage safety boundary is reversed according to the adjustment direction of the environmental adjustment equipment, and then returned along the reverse transmission chain according to the arrival order before and after the exchange. The grain condition nodes where the pre-exchange return status is matched with the corresponding grain warehouse storage environment status, and where the post-exchange return status is not matched with the corresponding grain warehouse storage environment status, are identified as grain condition nodes to be regulated.
[0008] Optionally, the execution process of the balancing action probe specifically includes: Based on the current operating level of the environmental control equipment, the first micro-action, the reverse micro-action, and the reset micro-action are applied sequentially. Set the first micro-action and the reset micro-action to be in the same direction and equal in amount, and set the reverse micro-action to be opposite in direction to the first micro-action and the adjustment amount equal to the sum of the first micro-action and the reset micro-action. Maintain a constant operating level of the environmental control equipment between adjacent continuous micro-actions; The directional sequence of continuous micro-movements is bound to the holding duration as a balance motion probe.
[0009] Optionally, determining the grain condition node that reproduces the sequence of continuous micro-actions as the response witness node specifically includes: Based on the starting time of the balancing action probe, the state changes of each grain condition node are segmented according to continuous micro-actions and adjacent holding times; Extract the direction of change and the time of the first change of each segment, determine the change interval based on the adjacent times of the first change, determine the earliest time of the first change as the first response time of the grain condition node, and form the response order of the grain condition node according to the order of the generation of the direction of change. Reverse the direction of each micro-action in the direction sequence of the balancing action probe to generate a reverse direction sequence, and align the grain condition node response sequence segment by segment with the direction sequence and reverse direction sequence of the balancing action probe. Remove grain condition nodes that have interrupted directional alignment, inconsistent change intervals with corresponding retention times, or completed reverse directional alignment. The grain situation nodes that have completed the alignment of the directional order and whose reverse directional order alignment is interrupted are identified as response witness nodes. The response witness nodes are arranged according to the time of the first response to form the arrival order of the response witness nodes.
[0010] Optionally, the forming device acts as a sealing chain, specifically including: Starting from the grain condition node corresponding to the location of the environmental regulation equipment, candidate related edges are sequentially connected according to the first response time of the response witness node; Candidate associated edges between the grain condition node corresponding to the location of the environmental regulation equipment and the first response witness node, between adjacent response witness nodes, and with the same source of the balancing action probes and the same propagation direction and arrival order, are converted into sealing edges. The continuation process will stop if any of the following occurs: candidate associated edges are missing, propagation direction is reversed, or arrival order is reversed. Connect the continuously formed sealing edges before the discontinuation to form the equipment function sealing chain.
[0011] Optionally, the step of binding the candidate control action with the source of the corresponding balance action probe and then pressing it into the device action certification chain specifically includes: The candidate control action is bound to the source of the corresponding balance action probe and written into the grain condition node corresponding to the location of the environmental regulation equipment. Set the first sealing edge of the sealing chain to the unlocked state and the remaining sealing edges to the locked state. Propagate along the edges of the sealing chain in the order of their succession to form the forward propagation state corresponding to the candidate control action. After the forward propagation state reaches the response witness node, the source, propagation direction and arrival order of the balance action probe of the next sealing edge and the previous sealing edge are checked. If the check matches, the next sealing edge is unlocked and the remaining subsequent sealing edges are kept locked. If the check does not match, the subsequent sealing edges are locked. Retain candidate control actions that connect to the grain processing point to be regulated, and eliminate candidate control actions that interrupt the transmission.
[0012] Optionally, determining the target control action specifically includes: The forward propagation state corresponding to the candidate control action that runs through to the grain level to be regulated is reversed in the opposite adjustment direction to form a return state, and is returned in reverse order along the corresponding equipment action sealing chain. At each response witness node, the returned state is sequentially matched with the forward propagation state corresponding to the candidate control action; Eliminate candidate control actions that trigger an interruption at any response witness node; The target control action is determined from the candidate control actions that have reached the grain condition node corresponding to the location of the environmental control equipment in the return state and have completed segment-by-segment engagement.
[0013] Optionally, the step of writing back the actual grain condition changes of the matching device's sealing chain to the sealing edge specifically includes: The actual grain situation changes are arranged according to the moment of the first change of the response witness node; Align the actual changes in grain conditions with the forward propagation states corresponding to the target control actions node by node; At each response witness node, verify the direction and order of change of the actual grain situation and the direction and order of propagation of the corresponding sealing edge; The actual changes in grain conditions that are continuously verified and consistent are defined as matching actual changes in grain conditions; Set the sealing edge in the sealing chain to the locked state, unlock the sealing edge corresponding to the actual grain condition change and write back the actual grain condition change, and keep the sealing edge not corresponding to the actual grain condition change locked.
[0014] Optionally, isolating unmatched actual grain condition changes from the update of the sealing certificate specifically includes: Align the response witness nodes in reverse order along the equipment action sealing chain that do not match the actual grain condition changes; Determine the direction of change and the first inconsistent response witness node in the arrival order, and cut off the writeback to the corresponding sealing edge for the unmatched actual grain situation change; Keep the corresponding seals and locks closed, and write back the grain storage environment status of the corresponding grain storage point if there are no matching actual changes in grain conditions. When subsequent changes in actual grain conditions are continuously verified against the corresponding sealing edge from the first discrepancy response witness node, the corresponding sealing edge will be unlocked.
[0015] Optionally, the updating device acts as a sealing chain, specifically including: According to the order of the sealing edges, retain the sealing edges that have completed the writing of actual grain condition changes; Disconnect the first locked sealing edge and subsequent sealing edges from the sealing chain that functions with the equipment; According to the first change time of the response witness node, the unlocked sealing edge will be reconnected to the device function sealing chain one by one; Connecting the retained sealing edges with the reconnected sealing edges forms a device-controlled sealing chain that extends to the grain condition nodes to be regulated.
[0016] The beneficial effects of this invention are: (1) The present invention applies continuous micro-actions with the cumulative adjustment amount to zero at the action position of the environmental regulation equipment by balancing action probe, identifies the response witness node that reproduces the micro-action sequence along the candidate correlation graph of grain condition node, and transforms the data correlation obtained by the model learning into the propagation relationship verified by the actual response of the equipment, thereby reducing the interference of changes in the external environment and natural heat dissipation of the grain pile on the judgment of the action path of the equipment.
[0017] (2) According to the arrival order of the response witness nodes, the present invention converts the continuous candidate associated edges into sealed edges, establishes the equipment action sealed chain, and the candidate control action only propagates along the sealed edges that match the source, propagation direction and arrival order. The unsealed position is immediately blocked, avoiding the control action from entering the grain condition area that lacks actual response support, and improving the correspondence accuracy between the range of action of the environmental regulation equipment and the grain condition node to be regulated.
[0018] (3) The present invention checks the actual grain condition changes with the equipment function seal chain node by node, only writes back the corresponding seal edge when matching the actual grain condition changes, isolates the unmatched actual grain condition changes from the seal edge update, and performs partial disconnection and edge-by-edge reconnection on the locked seal edge to avoid irrelevant state changes from rewriting the stable propagation relationship, thereby improving the stability and continuous correction capability of the grain storage environment control process. Attached Figure Description
[0019] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a deep learning-based intelligent control method for grain storage environment proposed in this invention. Figure 2 This is a flowchart illustrating the generation process of the device action seal chain driven by the balance action probe proposed in this invention. Figure 3 This is a flowchart of the candidate control action verification and equipment function certification chain update proposed in this invention. Detailed Implementation
[0020] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.
[0021] refer to Figures 1-3 A deep learning-based intelligent control method for grain storage environment includes the following steps: The grain storage environment status is obtained, and an improved MTGNN is used to map the grain condition detection location and the location of the environmental regulation equipment to grain condition nodes, generating a candidate association graph of grain condition nodes and determining the grain condition nodes to be regulated. MTGNN is a multivariate time series graph neural network used to learn the connection weights and directions between grain condition nodes based on the continuous state sequences of multiple grain condition nodes. The improved MTGNN adds the aforementioned reverse propagation chain generation, state freezing of grain condition nodes outside the reverse propagation chain, and exchange and return-to-connection operations for the arrival order of adjacent propagation states within the reverse propagation chain, based on the graph learning layer of MTGNN. The model training data uses the grain storage environment states obtained during the historical operation of the grain depot according to the sampling period. The training samples are divided into 13 consecutive sampling times. The grain storage environment states of the first 12 sampling times are used as the model input, and the grain storage environment states of the 13th sampling time are used as the training target. After the model training is completed, the grain storage environment states of the current 12 consecutive sampling times are input into the model to obtain the corresponding adjacency matrix. The improved MTGNN retains the graph learning layer of MTGNN. After the graph learning layer outputs the adjacency matrix, candidate associated edges are selected based on the adjacency matrix. Based on the candidate associated edges, continuous state sequences, grain storage safety boundaries, the location of environmental regulation equipment and the regulation direction of environmental regulation equipment, the reverse transmission chain is generated, the state of grain condition nodes outside the reverse transmission chain is frozen, the arrival order of adjacent propagation states within the reverse transmission chain is exchanged and the return deduction is performed. The grain condition nodes to be regulated are determined based on the return deduction results. The control environment regulation equipment performs a balancing action probe consisting of continuous micro-movements in opposite directions, with the cumulative adjustment amount returning to zero; Track the balance action probe along the candidate correlation graph of grain condition nodes, and identify the grain condition nodes that reproduce the sequence of continuous micro-actions as response witness nodes; According to the arrival order of the response witness nodes, the continuous candidate association edges that connect the response witness nodes sequentially from the grain condition node corresponding to the location of the environmental regulation equipment are converted into sealing edges to form the equipment action sealing chain. Read the status items that exceed the grain storage safety boundary in the grain condition node to be regulated, and read the environmental regulation equipment, regulation direction, minimum single regulation amount, minimum operation amount and maximum operation amount that can adjust the corresponding status item from the equipment control configuration table. When the status value exceeds the corresponding upper limit, select the regulation direction to decrease the corresponding status item. When the status value is lower than the corresponding lower limit, select the regulation direction to increase the corresponding status item. Divide the minimum single adjustment amount by the difference between the maximum and minimum operating amounts to obtain the proportional adjustment amount. Starting from a proportional adjustment amount, increase the adjustment amount incrementally. Add or subtract the adjustment amount after each increase from the current operating amount of the environmental control equipment according to the corresponding adjustment direction. Keep the adjustment amount that is within the range of 0 to 1 after calculation. Stop increasing when the next adjustment amount causes the current operating amount to exceed the range of 0 to 1. Combine each kept adjustment amount with the corresponding adjustment direction to form a candidate control action for the grain condition point to be controlled. Generate candidate control actions for the grain plot points to be regulated. After binding the candidate control actions with the source of the corresponding balance action probe, push them into the equipment action sealing chain. The action only propagates along the sealing edge that is consistent with the source of the balance action probe bound to the candidate control action and has the same propagation direction and arrival order. Propagation stops when encountering an unsealed candidate associated edge. The target control action is determined from the candidate control actions that run along the equipment's sealing chain to the grain condition point to be regulated, and the operation of the environmental regulation equipment is controlled. After determining the target control action, the current operating quantity of the environmental control equipment is added to or subtracted from the adjustment quantity of the target control action according to the adjustment direction of the target control action to obtain the target proportional operating quantity; when the target proportional operating quantity is zero, a shutdown command is written to the environmental control equipment controller; when the target proportional operating quantity is greater than zero, the target proportional operating quantity is multiplied by the difference between the maximum and minimum operating quantities read from the equipment control configuration table, and then the minimum operating quantity is added to obtain the target equipment operating quantity. The target equipment operating quantity is written to the environmental control equipment controller to control the operation of the environmental control equipment. Obtain actual changes in grain conditions, write back the actual changes in grain conditions that match the sealing chain of the equipment action to the sealing edge, isolate the actual changes in grain conditions that do not match from the sealing edge update, and update the sealing chain of the equipment action.
[0022] In this implementation method, determining the grain condition node to be regulated specifically includes: Input the grain storage environment status into the improved MTGNN to generate candidate association edges with propagation direction between grain condition nodes; Grain condition nodes correspond to the grain condition detection locations and the operating locations of environmental control equipment within the grain warehouse. The grain condition nodes corresponding to the grain condition detection locations are established based on the detection location numbers and spatial coordinates, which are read from the grain condition detection equipment layout table. The grain condition nodes corresponding to the operating locations of environmental control equipment are established based on the equipment installation locations and the equipment operating areas, which are read from the grain warehouse ventilation duct layout diagram and the equipment control configuration table. Environmental control equipment refers to controllable equipment used to change the environmental conditions of grain storage, including fans, grain coolers, dehumidifiers, air supply valves, and exhaust valves; the operating locations of environmental control equipment correspond to air supply outlets, air return outlets, air duct outlets, and the equipment's operating area. The environmental status of the grain storage area is composed of the status items of each grain condition node at the same sampling time. The status items of each grain condition node include grain temperature, grain moisture, internal temperature, internal humidity, oxygen concentration, carbon dioxide concentration, and equipment operating status. The equipment operating status of the grain condition node corresponding to the grain condition detection location is entered as zero; zero values are not included in the comparison of the grain storage safety boundary. The environmental status item of the grain condition node corresponding to the location of the environmental control equipment is the measurement value at the same sampling time from the grain condition detection location closest to the location of the environmental control equipment within the equipment's operating area. The equipment operating status is read from the environmental control equipment controller; the equipment operating status includes start / stop status and current operating level. Fan frequency, valve opening degree, and braking speed are read from the environmental control equipment controller. The corresponding values in the cold setpoint and dehumidification setpoint are recorded as the original operating quantity. When the start / stop state is stopped, the current operating quantity is zero. When the start / stop state is running, the original operating quantity is subtracted from the minimum operating quantity in the equipment control configuration table, and then divided by the difference between the maximum and minimum operating quantities. The resulting ratio of 0 to 1 is used as the current operating quantity. The grain storage safety boundary includes the upper limit of grain temperature, the upper limit of warehouse humidity, the upper and lower limits of grain moisture, the lower limit of oxygen concentration, and the upper limit of carbon dioxide concentration. The boundary values are read from the grain warehouse control parameter table corresponding to the current grain variety. Grain temperature, grain moisture, warehouse humidity, oxygen concentration, and carbon dioxide concentration are used in the comparison of the grain storage safety boundary. Warehouse temperature and equipment operating status are not used in the comparison of the grain storage safety boundary. The continuous state sequence of each grain condition node is obtained by arranging the state items of the corresponding grain condition node in chronological order from twelve consecutive sampling times. The twelve sampling times constitute an input time window of the improved MTGNN. Each time a new sampling time is obtained, the earliest sampling time in the input time window is deleted and the new sampling time is added to form the next continuous state sequence. The graph learning layer of the improved MTGNN receives the continuous state sequence and outputs an adjacency matrix with connection weights and connection directions. Among the connections that start from the grain condition node and have connection weights greater than zero, the top three connections are retained in descending order of connection weights. The retained connections are used as candidate associated edges, and the connection directions are used as propagation directions. The state that exceeds the grain storage safety boundary is transmitted in reverse along the candidate associated edge to the grain condition node corresponding to the location of the environmental regulation equipment, forming a reverse transmission chain; The state of exceeding the grain storage safety boundary includes the difference between the state value and the corresponding lower limit. Reverse propagation starts from the grain condition node corresponding to the state exceeding the grain storage safety boundary and searches backward along candidate associated edges to the grain condition node corresponding to the location of the environmental regulation equipment. When multiple reachable paths are obtained through the reverse search, the sum of the connection weights of the candidate associated edges in each reachable path is calculated, and the reachable path with the largest sum of connection weights is retained. When the sum of connection weights is the same, the reachable path with fewer candidate associated edges is retained. When both the sum of connection weights and the number of candidate associated edges are the same, the reachable path is compared sequentially according to the grain condition node numbers traversed by the reachable path, and the reachable path with the smallest grain condition node number appearing first is retained. The grain condition nodes and candidate associated edges traversed by the retained reachable paths are determined as the reverse propagation chain. If no reachable path to the grain condition node corresponding to the location of the environmental regulation equipment is found, no reverse propagation chain is formed for the corresponding state exceeding the grain storage safety boundary. Freeze the state of the grain situation nodes outside the reverse propagation chain, and exchange the arrival order of adjacent propagation states within the reverse propagation chain. The freeze operation keeps the state of the grain condition nodes outside the reverse propagation chain unchanged; the adjacent propagation state is the state of the adjacent candidate associated edges on the reverse propagation chain that have exceeded the grain storage safety boundary; before the exchange, the source grain condition node, state item and arrival order corresponding to each propagation state are recorded; the arrival order exchange only changes the order in which the adjacent propagation states enter the next grain condition node, and the source grain condition node, state item, arrival order before the exchange, state value and propagation direction remain unchanged. The state that exceeds the grain storage safety boundary is reversed according to the adjustment direction of the environmental adjustment equipment, and then returned along the reverse transmission chain according to the arrival order before and after the exchange. When the state changes direction, the difference exceeding the corresponding upper limit is converted to a decreasing direction, and the difference below the corresponding lower limit is converted to an increasing direction. The returned state is aligned item by item according to the source grain condition node, state item, and arrival order before the exchange. When the returned arrival order is consistent with the arrival order before the exchange, the corresponding state value is increased in the increasing direction and decreased in the decreasing direction within the improved MTGNN. The deregulation is considered complete when the calculated state value enters the grain storage safety boundary. The deregulation is considered interrupted when the returned arrival order is inconsistent with the arrival order before the exchange, or when the calculated state value does not enter the grain storage safety boundary. The grain condition nodes where the pre-exchange return status is matched with the corresponding grain warehouse storage environment status, and where the post-exchange return status is not matched with the corresponding grain warehouse storage environment status, are identified as grain condition nodes to be regulated.
[0023] In this embodiment, the execution process of the balancing action probe specifically includes: Based on the current operating level of the environmental control equipment, the first micro-action, the reverse micro-action, and the reset micro-action are applied sequentially. The current operating quantity of the environmental control equipment is taken from the current operating state of the aforementioned equipment, and the 0 to 1 proportional value is obtained from the environmental control equipment controller before the execution of the first micro-action. The first micro-action, reverse micro-action, and reset micro-action are each an adjustment performed by the environmental control equipment controller on the operating quantity of the environmental control equipment. The minimum single adjustment amount recorded in the equipment control configuration table is divided by the difference between the maximum and minimum operating quantities, and the resulting proportional value is used as the adjustment amount of the first micro-action. Before the first micro-action is executed, the operating quantity obtained by adding one first micro-action adjustment amount to the current operating quantity of the environmental control equipment and the operating quantity obtained by subtracting one first micro-action adjustment amount are calculated respectively. When both operating quantities are within the range of 0 to 1, the first micro-action is executed in the direction of increasing operating quantity. When either operating quantity exceeds the range of 0 to 1, the execution of this balancing action probe is stopped. When issuing a micro-action to the environmental control equipment controller, the micro-action adjustment amount is multiplied by the difference between the maximum and minimum operating quantities to obtain the adjustment amount under the corresponding equipment unit. Set the first micro-action and the reset micro-action to be in the same direction and equal in amount, and set the reverse micro-action to be opposite in direction to the first micro-action and the adjustment amount equal to the sum of the first micro-action and the reset micro-action. The adjustment amount of the reverse micro-action is twice the adjustment amount of the first micro-action, and the adjustment amount of the reset micro-action is the same as the adjustment amount of the first micro-action; after the three consecutive micro-actions are executed, the operating value of the environmental control equipment returns to the current operating value of the environmental control equipment before the execution. Maintain a constant operating level of the environmental control equipment between adjacent continuous micro-actions; After the first micro-action is executed, the operating rate of the environmental control equipment is maintained for two sampling cycles. After the reverse micro-action is executed, the operating rate of the environmental control equipment is maintained for three sampling cycles. The sampling cycles are read from the aforementioned grain condition monitoring equipment layout table, and the operating rate of the environmental control equipment is not changed during the maintenance period. The directional sequence of continuous micro-movements is bound to the holding duration as a balance motion probe.
[0024] The binding operation involves recording the direction of each micro-action in the execution order of the first micro-action, the reverse micro-action, and the reset micro-action, and recording the corresponding holding duration after the first micro-action and the reverse micro-action. The direction record and the holding duration record are combined into a balance action probe.
[0025] In this embodiment, the grain condition node that reproduces the sequence of continuous micro-actions is determined as the response witness node, specifically including: Based on the starting time of the balancing action probe, the state changes of each grain condition node are segmented according to continuous micro-actions and adjacent holding times; The starting time of the balancing action probe is taken as the execution time of the first micro-action; the first segment extends from the execution time of the first micro-action to the execution time of the reverse micro-action, the second segment extends from the execution time of the reverse micro-action to the execution time of the reset micro-action, and the third segment extends from the execution time of the reset micro-action to the first sampling time after the environmental control equipment operation amount returns to the value before execution. Extract the direction of change and the time of the first change of each segment, determine the change interval based on the adjacent times of the first change, determine the earliest time of the first change as the first response time of the grain condition node, and form the response order of the grain condition node according to the order of the generation of the direction of change. For each grain condition monitoring point, the state changes within three segments are extracted for grain temperature, grain moisture, warehouse temperature, warehouse humidity, oxygen concentration, and carbon dioxide concentration. The equipment operating status is not included in the response witness node determination. Starting from the second sampled value within a segment, it is compared sequentially with the previous sampled value. When the absolute value of the difference between the current sampled value and the previous sampled value reaches the detection resolution of the corresponding grain condition monitoring equipment, the current sampling time is determined as the first change time. A positive difference is determined as an upward direction, and a negative difference is determined as a downward direction. The detection resolution is read from the aforementioned grain condition monitoring equipment layout table. When the absolute value of the difference between adjacent sampled values within a segment does not reach the detection resolution, the corresponding segment is considered to have no response. When the first segment responds, the direction of change of the first segment is recorded as the first direction; when the direction of change of subsequent segments is the same as the first direction, it is recorded as the first direction, and when it is opposite to the first direction, it is recorded as the reverse direction; the first direction and the reverse direction are arranged in the order of the three segments to form the response order of the grain condition nodes for the corresponding status items; when any segment does not respond, it is determined that the alignment of the direction order of the corresponding status items is interrupted, and the change interval involving the unresponsive segment is not calculated; When all three segments respond, the change interval is the number of sampling periods contained between the first change times of adjacent segments; the earliest first change time among the three segments of the corresponding state item is taken as the first response time of the corresponding grain condition node under the corresponding state item. Reverse the direction of each micro-action in the direction sequence of the balancing action probe to generate a reverse direction sequence, and align the grain condition node response sequence segment by segment with the direction sequence and reverse direction sequence of the balancing action probe. During segment-by-segment alignment, the directions of the first micro-action and the reset micro-action are recorded as the first direction, and the direction of the reverse micro-action is recorded as the reverse direction. The response order of the grain condition nodes is compared sequentially according to their arrangement positions. When all three positions are consistent, the direction order alignment is considered complete. When any position is inconsistent or there is no response, the direction order alignment is considered interrupted. In the reverse direction order, the first direction is converted to the reverse direction, and the reverse direction is converted to the first direction. Segment-by-segment alignment is performed in the same way. The two change intervals are compared with the number of sampling periods contained in the hold duration after the first micro-action and the hold duration after the reverse micro-action, respectively. When the values are equal, the change interval is considered to be consistent with the corresponding hold duration. The aforementioned determination is performed on each state item; when at least one state item has completed directional sequence alignment, the change interval is consistent with the corresponding maintenance duration, and the reverse direction sequence alignment is interrupted, it is determined that the directional sequence alignment of the corresponding grain condition node is completed, the change interval is consistent with the corresponding maintenance duration, and the reverse direction sequence alignment is interrupted; when no state item satisfies the aforementioned conditions, it is determined that the corresponding grain condition node has at least one of the following situations: directional sequence alignment interruption, change interval inconsistent with the corresponding maintenance duration, and reverse direction sequence alignment completed; when at least one state item satisfies the aforementioned conditions, the earliest time among the first response times of each state item that satisfies the aforementioned conditions is taken as the first response time of the grain condition node; Remove grain condition nodes that have interrupted directional alignment, inconsistent change intervals with corresponding retention times, or completed reverse directional alignment. The grain situation nodes that have completed the alignment of the directional order and whose reverse directional order alignment is interrupted are identified as response witness nodes. The response witness nodes are arranged according to the time of the first response to form the arrival order of the response witness nodes.
[0026] When arranging according to the first response time, the order is from earliest to latest; when the first response times are the same, they are arranged in ascending order according to the grain situation node number; the grain situation node number is generated according to the node establishment order when the grain situation node is established.
[0027] In this embodiment, the formation of the equipment function certification chain specifically includes: Starting from the grain condition node corresponding to the location of the environmental regulation equipment, candidate related edges are sequentially connected according to the first response time of the response witness node; According to the arrival order of the aforementioned response witness nodes, the grain condition node corresponding to the location of the environmental control equipment and the first response witness node are taken as the previous and next grain condition nodes of the current continuation; after the current continuation is completed, the next grain condition node is taken as the new previous grain condition node, and the next response witness node is taken as the new next grain condition node. Candidate association edges connecting the previous and next grain condition nodes are searched sequentially in the grain condition node candidate association graph. Candidate associated edges between the grain condition node corresponding to the location of the environmental regulation equipment and the first response witness node, between adjacent response witness nodes, and with the same source of the balancing action probes and the same propagation direction and arrival order, are converted into sealing edges. In the first candidate association edge, the location of the environmental regulation equipment corresponds to the grain condition node and the environmental regulation equipment executing the balancing action probe, and the first response witness node is determined by the same balancing action probe; in subsequent candidate association edges, the preceding and following grain condition nodes are both determined by the same balancing action probe; when the above conditions are met, the balancing action probes are considered to have the same source; the source of the balancing action probe is jointly determined by the environmental regulation equipment executing the balancing action probe and the corresponding start time; in the arrival order determination, the grain condition node corresponding to the location of the environmental regulation equipment is located before the first response witness node, and the adjacent response witness nodes are determined according to the aforementioned arrival order; when a candidate association edge points from the preceding grain condition node to the following grain condition node, the propagation direction is considered to be consistent with the arrival order; When a candidate associated edge meets the aforementioned conditions, the grain condition nodes at both ends and the propagation direction of the candidate associated edge are retained. The source, propagation direction and arrival order of the corresponding balance action probe are recorded in the candidate associated edge, and the candidate associated edge that has completed the recording is converted into a sealing edge. The continuation process will stop if any of the following occurs: candidate associated edges are missing, propagation direction is reversed, or arrival order is reversed. Missing candidate association edges means there are no candidate association edges connecting the current previous grain condition node and the next grain condition node in the candidate association graph; reversed propagation direction means the candidate association edge points from the next grain condition node to the previous grain condition node; reversed arrival order means the starting point of the candidate association edge is after the ending point of the candidate association edge in the aforementioned sequence; if any of the above situations occur, the current candidate association edge is not transformed, and the candidate association edges after the current candidate association edge are not continued; if none of the above situations occur, the continuation ends after the last response witness node; Connect the continuously formed sealing edges before the discontinuation to form the equipment function sealing chain.
[0028] When the first candidate associated edge is not converted into a sealing edge, no equipment action sealing chain is formed; when at least one sealing edge is formed, the continuously formed sealing edges are connected according to the arrival order of the corresponding response witness nodes of the sealing edge. The starting point after connection is the grain condition node corresponding to the location of the environmental regulation equipment, and the ending point is the last response witness node before the connection is stopped; when no connection is stopped, the ending point is the last response witness node in the arrival order, and the connection result is the equipment action sealing chain.
[0029] In this embodiment, the candidate control action is bound to the source of the corresponding balance action probe and then pushed into the device action certification chain, specifically including: The candidate control action is bound to the source of the corresponding balance action probe and written into the grain condition node corresponding to the location of the environmental regulation equipment. The source of the balance action probe corresponding to the candidate control action is determined according to the aforementioned environmental regulation equipment and the corresponding start time; when the source is bound, the source of the corresponding balance action probe is recorded in the candidate control action, and the candidate control action that has been recorded is written into the grain condition node corresponding to the action position of the environmental regulation equipment. Set the first sealing edge of the sealing chain to the unlocked state and the remaining sealing edges to the locked state. The unlocked state indicates that candidate control actions are allowed to pass through the sealing edge, and the locked state indicates that candidate control actions are prevented from passing through the sealing edge. According to the sequence of the sealing chain of the device function, the sealing edge of the grain condition node corresponding to the function position of the connected environmental regulation device and the first response witness node is set to the unlocked state, and the subsequent sealing edges are set to the locked state. Propagate along the edges of the sealing chain in the order of their succession to form the forward propagation state corresponding to the candidate control action. Before propagation, the grain condition node to be controlled is compared one by one with the grain condition nodes in the equipment action certification chain arranged in sequential order according to the grain condition node number. The location of the grain condition node with the same number is determined as the propagation termination position. If no grain condition node with the same number is found in the equipment action certification chain, the propagation of the corresponding candidate control action is considered to be interrupted. Candidate control actions consist of the adjustment direction and adjustment amount determined when the candidate control action is generated; edge-by-edge propagation starts from the grain condition node corresponding to the location of the environmental control equipment and passes the candidate control action along the unlocked sealing edge to the next grain condition node; different candidate control actions are written separately; each time a candidate control action passes through a sealing edge, the corresponding adjustment direction, adjustment amount, source of the bound balance action probe, and the sequence of the sealing edges that have been passed are written to the currently arrived grain condition node; the contents written for the same candidate control action at each grain condition node are arranged according to the sequence of the sealing edges to form the corresponding forward propagation state; After the forward propagation state reaches the response witness node, the source, propagation direction and arrival order of the balance action probe of the next sealing edge and the previous sealing edge are checked. If the check matches, the next sealing edge is unlocked and the remaining subsequent sealing edges are kept locked. If the check does not match, the subsequent sealing edges are locked. The balance action probe source verification is consistent if the environmental control equipment and corresponding start time recorded by the next and previous verification edges are the same; the propagation direction verification is consistent if the end point of the previous verification edge and the start point of the next verification edge are both the current response witness node, and the next verification edge points to the next response witness node in the arrival order; the arrival order verification is consistent if the response witness node corresponding to the end point of the next verification edge is located after the current response witness node; after reaching the response witness node in the forward propagation state, first check if the current response witness node is located at the propagation termination position; if it is located at the propagation termination position, stop the propagation and keep the subsequent verification edges locked; if it is not located at the propagation termination position, perform the next verification edge verification; if all three verifications are consistent, unlock the next verification edge; if any verification is inconsistent, keep the next verification edge and the subsequent verification edges locked. Retain candidate control actions that connect to the grain processing point to be regulated, and eliminate candidate control actions that interrupt the transmission.
[0030] When the forward propagation state reaches the propagation termination position along the continuously unlocked sealing edge, the corresponding candidate control action is considered to have been connected to the grain condition node to be regulated; when the grain condition node to be regulated is not found in the equipment action sealing chain, the forward propagation state cannot pass through the next sealing edge, or the sealing edge verification is inconsistent, the propagation of the corresponding candidate control action is considered to be interrupted.
[0031] In this embodiment, determining the target control action specifically includes: The forward propagation state corresponding to the candidate control action that runs through to the grain level to be regulated is reversed in the opposite adjustment direction to form a return state, and is returned in reverse order along the corresponding equipment action sealing chain. Keeping the adjustment amount, balance action probe source, grain condition node and sealing edge connection sequence record unchanged in the forward propagation state, the adjustment direction is changed to the opposite adjustment direction to form a return state; the return starts from the grain condition node to be controlled, and the sealing edge is read from back to front according to the sealing edge connection sequence record, and the return state is sequentially transmitted to each response witness node passed through in the forward propagation state. At each response witness node, the returned state is sequentially matched with the forward propagation state corresponding to the candidate control action; During segment-by-segment engagement, the feedback state and the corresponding forward propagation state are aligned according to the same candidate control action, the same balance action probe source, the same response witness node, and the same sealing edge connection sequence; the adjustment amount of the feedback state at the same response witness node is canceled out with the adjustment amount of the forward propagation state in opposite adjustment directions; when the adjustment amount after cancellation is zero, the current response witness node is considered to have completed engagement; when the adjustment amount after cancellation is not zero, the current response witness node is considered to have interrupted engagement. If any of the following situations occur: the candidate control actions corresponding to the return state and the forward propagation state are inconsistent, the source of the balance action probe is inconsistent, the response witness node is inconsistent, the sequence of sealing edge is inconsistent, or the adjustment direction is not reversed, the current response witness node is considered to be interrupted. Eliminate candidate control actions that trigger an interruption at any response witness node; For each candidate control action, reverse feedback and segmented locking are performed respectively; when locking is interrupted at any response witness node, the reverse feedback of the corresponding feedback state is stopped and the corresponding candidate control action is removed; when all response witness nodes through which the forward propagation state passes have completed locking, the feedback state is continued to be fed back to the grain condition node corresponding to the location of the environmental control equipment. The target control action is determined from the candidate control actions that have reached the grain condition node corresponding to the location of the environmental control equipment in the return state and have completed segment-by-segment engagement.
[0032] When the return status reaches the grain condition node corresponding to the location of the environmental control equipment, and all response witness nodes passed through by the forward propagation status have completed engagement, the corresponding candidate control action is retained. If only one candidate control action is retained, the retained candidate control action is determined as the target control action. If multiple candidate control actions are retained, they are arranged in ascending order of adjustment amount, and the candidate control action with the smallest adjustment amount is determined as the target control action. If the adjustment amounts are the same, the candidate control actions are determined as the first one written to the grain condition node corresponding to the location of the environmental control equipment.
[0033] In this embodiment, writing back the actual grain condition changes of the matching device's sealing chain to the sealing edge specifically includes: The actual grain situation changes are arranged according to the moment of the first change of the response witness node; Since the start of the target control action, the grain storage environment status of each response witness node passed through by the forward propagation state corresponding to the target control action is acquired according to the aforementioned sampling cycle. After the current operating quantity fed back by the environmental adjustment equipment controller reaches the value obtained by increasing or decreasing the corresponding adjustment quantity according to the adjustment direction of the target control action, sampling continues. The duration of continued sampling is the difference between the first response time of the last response witness node passed through by the forward propagation state corresponding to the target control action and the start time of the balance action probe. Sampling stops when the duration expires. For each state item that meets the judgment condition when determining the response witness node, the first sampled value after the start of the target control action is compared with the last sampled value before the start of the target control action. Subsequent sampled values are compared with the previous sampled value in turn. When the absolute value of the difference between the current sampled value and the previous sampled value of the state item first reaches the detection resolution of the corresponding grain condition detection equipment, the corresponding state item, the current difference, and the current sampling time are recorded. The difference of the corresponding state item is not repeated for subsequent sampled values. The state items, differences, and sampling times recorded in the same response witness node are merged into the corresponding actual grain condition change, and the earliest sampling time is determined as the first change time of the response witness node. The actual grain condition changes are arranged in order of arrival of the forward propagation state corresponding to the target control action, from earliest to latest, through each response witness node. If the initial change times are the same, they are arranged in order of ascending grain condition node number. Align the actual changes in grain conditions with the forward propagation states corresponding to the target control actions node by node; Read the response witness nodes passed through in the forward propagation state according to the sealing edge continuation order in the forward propagation state, and compare the grain condition node number corresponding to the actual grain condition change with the grain condition node number in the forward propagation state one by one; if the grain condition node number is consistent, align the corresponding actual grain condition change with the forward propagation state at the current response witness node; if the grain condition node number is inconsistent, or if the current response witness node has not obtained the actual grain condition change, it is determined that the current response witness node has not completed the node-by-node alignment. At each response witness node, verify the direction and order of change of the actual grain situation and the direction and order of propagation of the corresponding sealing edge; The direction of change in actual grain condition in this step refers to the transmission direction of the actual grain condition change between grain condition nodes, and does not indicate the direction of increase or decrease of the status item; the direction of change in actual grain condition is determined according to the arrival order of the actual grain condition changes; the direction of change of the first actual grain condition change is from the grain condition node corresponding to the location of the environmental regulation equipment to the first response witness node, and the direction of change of subsequent actual grain condition changes is from the previous response witness node to the current response witness node; the corresponding sealing edge is the sealing edge in the sealing chain of equipment action with the current response witness node as the endpoint; If the direction of change of actual grain conditions is consistent with the direction of propagation of the corresponding sealing edge, and the arrival order of the corresponding response witness node is consistent with the arrival order of the corresponding sealing edge endpoint, the current response witness node is considered to be consistent; if either the direction of change or the arrival order is inconsistent, or if the current response witness node has not completed node-by-node alignment, the current response witness node is considered to be inconsistent. The actual changes in grain conditions that are continuously verified and consistent are defined as matching actual changes in grain conditions; Starting from the first response witness node passed through in the forward propagation state corresponding to the target control action, the verification is performed continuously according to the sealing edge continuation order in the forward propagation state; when the current response witness node is verified to be consistent, the verification continues to the next response witness node; when the first verification discrepancy occurs, the continuous verification stops; the actual grain condition changes that are consistent with the continuous verification before the first verification discrepancy are considered matched actual grain condition changes, and the actual grain condition changes that are inconsistent with the first verification and those that are after the first verification discrepancy are considered unmatched actual grain condition changes; Set the sealing edge in the sealing chain to the locked state, unlock the sealing edge corresponding to the actual grain condition change and write back the actual grain condition change, and keep the sealing edge not corresponding to the actual grain condition change locked.
[0034] Set all sealing edges in the equipment sealing chain to the locked state; unlock the sealing edges whose endpoints are the corresponding response witness nodes in the order of arrival of the actual grain condition changes, and write the grain condition node number, status item, difference, first change time, change direction and arrival order corresponding to the actual grain condition changes into the unlocked sealing edges; do not keep the sealing edges corresponding to the actual grain condition changes locked.
[0035] In this implementation, the isolation of unmatched actual grain condition changes from the border closure update specifically includes: Align the response witness nodes in reverse order along the equipment action sealing chain that do not match the actual grain condition changes; Read the unmatched actual grain condition changes from back to front according to the arrival order of the unmatched actual grain condition changes. Search for the corresponding response witness node with the same number in the opposite direction of the sealing edge continuation order according to the grain condition node number. When the grain condition node numbers are consistent, align the unmatched actual grain condition changes with the corresponding response witness node. If no response witness node with the same number is found, the corresponding unmatched actual grain condition change will not participate in the sealing edge write-back. Determine the direction of change and the first inconsistent response witness node in the arrival order, and cut off the writeback to the corresponding sealing edge for the unmatched actual grain situation change; After completing the reverse alignment, starting from the side of the grain condition node corresponding to the location of the environmental regulation equipment, check the direction of change and arrival order of the unmatched actual grain condition changes with the propagation direction and arrival order of the corresponding sealing edge according to the sequence of the equipment action sealing chain; determine the earliest response witness node with inconsistent change direction or arrival order as the first mismatch response witness node; the corresponding sealing edge is the sealing edge connecting the first mismatch response witness node; from the first mismatch position, do not write unmatched actual grain condition changes to the corresponding sealing edge and subsequent sealing edges; Keep the corresponding seals and locks closed, and write back the grain storage environment status of the corresponding grain storage point if there are no matching actual changes in grain conditions. Maintain the locked state of the sealing edge corresponding to the first mismatched response witness node and subsequent sealing edges; according to the grain condition node number, status item, and first change time of the unmatched actual grain condition change record, read the sampled value of the corresponding status item at the first change time from the grain storage environment status obtained by the aforementioned sampling, and write the read sampled value into the same status item record of the corresponding grain condition node at the first change time, without overwriting the status item records of the corresponding grain condition node at other sampling times; when the unmatched actual grain condition change includes multiple status items, read and write the sampled value of each status item at the first change time respectively; When subsequent changes in actual grain conditions are continuously verified against the corresponding sealing edge from the first discrepancy response witness node, the corresponding sealing edge will be unlocked.
[0036] Starting from the next sampling moment corresponding to the actual grain condition change of the first mismatched response witness node, subsequent actual grain condition changes are generated according to the aforementioned sampling period, detection resolution, and actual grain condition change acquisition method. The collection duration of this subsequent actual grain condition change is the collection duration after the aforementioned target control action is executed, and the collection stops when the collection duration expires. If no subsequent actual grain condition changes are obtained during this collection period, the sealing edge corresponding to the first mismatched response witness node remains locked. When subsequent actual grain condition changes are obtained, the subsequent actual grain condition changes obtained this time are aligned with the response witness nodes according to the grain condition node number. Starting from the first mismatched response witness node, the change direction and arrival order are checked node by node according to the sequence of the sealing chain of equipment function. When all responses from the first mismatched response witness node to the last response witness node for which subsequent actual grain condition changes are obtained match, the locking of the sealing edge corresponding to the first mismatched response witness node is released. If any response witness node does not match, the corresponding sealing edge remains locked.
[0037] In this embodiment, updating the device function authentication chain specifically includes: According to the order of the sealing edges, retain the sealing edges that have completed the writing of actual grain condition changes; The sealing edge that has completed the actual grain condition change write-back is the sealing edge that has been unlocked in the aforementioned manner and has been written with the grain condition node number, status item, difference, first change time, change direction and arrival order that match the actual grain condition change; starting from the first sealing edge of the equipment action sealing chain, the sealing edges are checked continuously in the order of sealing edge continuation, and the sealing edges that have completed the actual grain condition change write-back are retained in the equipment action sealing chain; Disconnect the first locked sealing edge and subsequent sealing edges from the sealing chain that functions with the equipment; The first locked sealing edge detected according to the sealing edge sequence is identified as the first locked sealing edge. If the first locked sealing edge is the first sealing edge in the equipment-operated sealing chain, this sealing edge and subsequent sealing edges are directly disconnected from the equipment-operated sealing chain. If the first locked sealing edge is not the first sealing edge in the equipment-operated sealing chain, the connection between the first locked sealing edge and the previous sealing edge is canceled, and the first locked sealing edge and subsequent sealing edges are disconnected from the equipment-operated sealing chain. After disconnection, the grain condition nodes at both ends, propagation direction, arrival order, and balance action probe source of each sealing edge are retained. If no locked sealing edge is detected, the disconnection operation is not performed. According to the first change time of the response witness node, the unlocked sealing edge will be reconnected to the device function sealing chain one by one; The sealed edge that is unlocked is the sealed edge that is unlocked based on the continuous verification of the subsequent actual grain condition changes mentioned above; the first change time of the corresponding witness node at the end of the sealed edge is taken as the first change time of the subsequent actual grain condition changes used to unlock the corresponding sealed edge at that witness node; the sealed edges that are unlocked are arranged from earliest to latest according to the first change time; if the first change times are the same, they are arranged from smallest to largest according to the grain condition node number corresponding to the end of the sealed edge; Starting from the last sealing edge in the reserved sealing edge, connect the unlocking sealing edge with the same starting point as the previous sealing edge one by one; when there are no reserved sealing edges, connect the unlocking sealing edge with the same starting point from the grain condition node corresponding to the location of the environmental regulation equipment. Connecting the retained sealing edges with the reconnected sealing edges forms a device-controlled sealing chain that extends to the grain condition nodes to be regulated.
[0038] When the endpoint of the previous sealing edge and the starting point of the next sealing edge are the same grain condition node, and the next sealing edge points from the grain condition node to the next response witness node, and the arrival order of the response witness node corresponding to the endpoint of the next sealing edge is after the response witness node corresponding to the endpoint of the previous sealing edge, the two sealing edges are connected; the edges are connected one by one in the aforementioned manner until the endpoint of the sealing edge is the grain condition node to be regulated, and the continuously connected sealing edges are determined as the updated equipment function sealing chain; If any of the following situations occur during the reconnection process: the starting point of the next sealing edge is inconsistent with the ending point of the previous sealing edge, the propagation direction is reversed, or the arrival order is reversed, the corresponding sealing edge will not be connected, and the corresponding sealing edge will remain disconnected from the equipment-operated sealing chain. If the connection result does not reach the grain condition node to be regulated, the connection result will not be recognized as the updated equipment-operated sealing chain, and the unconnected sealing edge will remain disconnected.
[0039] Example 1: To verify the feasibility of this invention in practice, it was applied to the grain storage environment control of a grain warehouse. The grain warehouse is a single-story structure, storing approximately 5200 tons of wheat, with an average grain pile height of 6.2 meters. 240 grain temperature monitoring points, 24 grain moisture monitoring points, 12 warehouse temperature and humidity monitoring points, and 4 gas concentration monitoring points were installed inside the warehouse. The sampling cycle of the grain condition monitoring equipment was 5 minutes. The environmental control equipment consisted of 4 fans, 1 grain cooler, 2 dehumidifiers, 8 air supply valves, and 4 exhaust valves. The equipment control configuration table recorded the start / stop status, minimum operating volume, maximum operating volume, and minimum single adjustment amount for each device. The minimum single adjustment amount for the fan frequency was 2 Hz, and the minimum single adjustment amount for the valve opening was 5%. The setpoints for the grain cooler and dehumidifier were recorded according to the minimum adjustment step size allowed by the equipment controller. The grain storage control parameter table sets the upper limit of grain temperature at 25.0℃, the grain moisture range at 11.5% to 13.5%, the upper limit of humidity in the storage at 75%RH, the lower limit of oxygen concentration at 19.0%, and the upper limit of carbon dioxide concentration at 0.50%.
[0040] The grain silo originally used a threshold-triggered method to control the environmental conditioning equipment. When the detected value exceeded the safety boundary, the controller would activate the ventilation fan, grain cooler, or dehumidifier according to a fixed priority, with the equipment operation levels set according to the extent of the exceedance. In actual operation, changes in heat, moisture, and gas within the grain pile have a propagation lag, and abnormal changes at some detection points are not directly caused by the current equipment. Fixed-priority control easily leads to multiple devices operating simultaneously, continuing to adjust even after local conditions have recovered, resulting in excessive grain temperature drops, reverse humidity fluctuations within the silo, and high ventilation energy consumption. When adjacent detection points are affected by air ducts, grain pile porosity, and local condensation, the arrival order of abnormal changes will also shift, and conventional control cannot distinguish between the actual path of equipment action and the path of occasional fluctuations.
[0041] When applying this invention, the grain condition detection locations and the operating locations of environmental control equipment are established as grain condition nodes. The spatial coordinates of the nodes corresponding to the grain condition detection locations are read from the grain condition detection equipment layout table, and the installation location and operating area of the nodes corresponding to the operating locations of the environmental control equipment are read from the grain silo ventilation duct layout diagram and equipment control configuration table. Each grain condition node uses state items from 12 consecutive sampling times to form an input time window. The state items include grain temperature, grain moisture, silo temperature, silo humidity, oxygen concentration, carbon dioxide concentration, and equipment operating status. The improved MTGNN outputs the connection weights and propagation directions between nodes based on the continuous state sequence. Each node retains the three connections with connection weights greater than 0 and ranked first, forming a candidate association graph of grain condition nodes.
[0042] When the grain temperature in the upper part of the grain pile reached 27.4℃ and the humidity inside the storage silo reached 81.8%RH, both the grain temperature and the humidity exceeded the safe storage boundary. An improved MTGNN was used to search for the location of environmental control equipment along candidate associated edges, forming reverse transmission chains related to the ventilator, grain cooler, and dehumidifier. When multiple reachable paths exist, the retained path is determined according to the sum of the connection weights of the candidate associated edges, the number of candidate associated edges, and the grain condition node number. The state of nodes outside the reverse transmission chain remains unchanged, and the arrival order of adjacent propagation states within the chain is exchanged before a return connection is executed. Grain condition nodes whose return states before the exchange can be matched with the corresponding grain storage environment state, or whose return states are interrupted after the exchange, are identified as grain condition nodes to be controlled. This process avoids directly using short-term fluctuations far from the equipment's operating path as the basis for control.
[0043] After identifying the grain condition node to be regulated, the ventilator is controlled to execute a balancing probe. Before execution, the ventilator's current operating value is 0.45, and the proportional adjustment of the first micro-action is 0.05. The controller sequentially executes continuous micro-actions increasing by 0.05, decreasing by 0.10, and increasing by 0.05. After the first micro-action, it maintains a 2-sampling-cycle period; after the reverse micro-action, it maintains a 3-sampling-cycle period. After the 3 micro-actions, the ventilator's operating value returns to 0.45. The grain temperature, grain moisture, warehouse temperature, warehouse humidity, oxygen concentration, and carbon dioxide concentration at each grain condition node are extracted according to three segments, including the direction of change and the time of the first change. A change in the status item that reaches the resolution of the detection equipment is recorded as a valid response. Grain condition nodes where the direction sequence of the three segments is consistent with the balancing probe, the change interval is consistent with the holding time, and the reverse direction sequence is aligned with the interruption are identified as response witness nodes.
[0044] The response witness nodes are arranged from earliest to latest according to the time of their first response. Starting from the grain condition node corresponding to the location of the ventilator's action, candidate related edges connecting each response witness node are searched sequentially. When the source, propagation direction, and arrival order of the balancing action probes of the candidate related edges are consistent, the candidate related edges are converted into sealing edges and connected to form a device action sealing chain. The device action sealing chain formed in this control operation runs through the four response witness nodes and reaches the grain condition node to be regulated. Grain condition nodes outside the chain do not participate in the propagation of this candidate control action.
[0045] Candidate control actions are generated based on the out-of-bounds status item and the equipment control configuration table. For grain temperatures that are too high, ventilators and grain coolers that can lower the grain temperature are selected; for humidity levels that are too high in the storage area, ventilators, dehumidifiers, and air supply valves that can lower the humidity are selected. Each device starts with a proportional adjustment amount and increases it sequentially. The corresponding adjustment amount is retained when the current operating amount after calculation remains within the range of 0 to 1. Candidate control actions are bound to the environmental adjustment equipment executing the balance action probe and the start time, and are written to the grain condition node corresponding to the location of the environmental adjustment equipment. The first sealing edge of the sealing chain is unlocked, while the remaining sealing edges remain locked. Each time a candidate control action reaches a response witness node, the source, propagation direction, and arrival order of the balance action probe of the next sealing edge are checked. After the check matches, each edge is unlocked sequentially. Candidate control actions that do not reach the grain condition node to be regulated are eliminated.
[0046] After the connection is established, the candidate control actions are sent back in reverse order. The return state maintains the same order of adjustment amount, balancing action probe source, grain condition node, and sealing edge sequence, only changing the adjustment direction to the opposite direction. The return state is considered complete when each response witness node and the forward propagation state are sequentially matched, and the adjustment amount is completely offset while the source, node, and sequence are consistent. This operation retained the combined action of increasing the fan operation by 0.15, the grain cooler operation by 0.10, and the dehumidifier operation by 0.10; other candidate actions were eliminated due to propagation interruption or matching interruption. After determining the target control action according to the rule of prioritizing the smaller adjustment amount, the proportional operation amount is converted into the fan frequency, cooling setpoint, and dehumidification setpoint that the equipment controller can execute and then issued.
[0047] After the target control action is executed, response witness node data is continuously acquired according to a 5-minute sampling cycle. The grain temperature gradually decreased from 27.4℃ to 24.8℃, the humidity inside the warehouse decreased from 81.8%RH to 73.6%RH, the oxygen concentration remained above 19.2%, and the carbon dioxide concentration did not exceed 0.50%. Actual grain condition changes are arranged according to the time of the first change and aligned node by node with the forward propagation state. Actual grain condition changes that continuously match in direction and arrival order are written to the corresponding sealing edge; unmatched changes do not participate in the sealing edge update, but only the status record of the corresponding grain condition node is written back. After subsequent changes are continuously verified to be consistent, the corresponding sealing edge is unlocked and reconnected to the device's sealing chain.
[0048] To obtain comparable data, nine representative abnormal states were selected for repeated control under the same grain warehouse, similar grain storage quantity, and the same safety boundary conditions. Each abnormal state was run five times using both conventional threshold control and the method of this invention. The data in the table is the average of these five runs. The stabilization time was calculated from the start of the control action until all out-of-bounds states returned to the safety boundary and remained there for three consecutive sampling periods. The maximum overshoot was expressed as the ratio of the maximum deviation from the safety boundary target value after state recovery to the initial out-of-bounds amplitude. Control energy consumption was accumulated from the electrical energy metering values of the ventilator, grain cooler, dehumidifier, and valve actuators. The number of invalid actions was the number of actions that, after equipment execution, failed to create a matching effect on the actual grain condition change in the corresponding equipment's operational chain.
[0049] Table 1: Comparison of the effects of grain storage environment control under different abnormal conditions
[0050] In terms of stabilization time, this invention is faster than conventional threshold control under all nine types of abnormal conditions. The average stabilization time of conventional control is 150.8 min, while that of this invention is 113.1 min, a reduction of 25.2% on average. When the humidity in a single storage silo exceeds the limit, the stabilization time decreases from 118 min to 84 min, a reduction of 28.8%. When grain temperature, storage humidity, and carbon dioxide concentration all exceed the limits simultaneously, the number of controlled objects increases, and the state propagation relationship becomes more complex. Even so, this invention still reduces the stabilization time from 218 min to 166 min. These results demonstrate that by using a balance action probe to identify the true response node and limit the propagation path of candidate control actions, the controller no longer waits for changes in irrelevant nodes, nor does it need to repeatedly try equipment actions that have no interaction with the node to be controlled.
[0051] The maximum over-adjustment range decreased from an average of 7.8% under conventional control to 2.5%. Conventional control directly increases equipment operation based on the extent of over-adjustment. However, even when grain temperature or humidity approaches the safety boundary, the state continues to change due to equipment inertia and the lag in propagation within the grain pile. This invention first uses the equipment action sealing chain to screen actions that can reach the node of grain condition to be regulated. Then, it eliminates actions with unstable propagation relationships by segmentally matching the return state and the forward propagation state. The control quantity is generated starting from a small adjustment amount, and the target control action is determined from the candidate actions that meet the propagation closure conditions, so that the equipment operation range matches the actual action path, and excessive ventilation, excessive cooling, and excessive dehumidification are all suppressed.
[0052] Energy consumption data shows that conventional control consumes an average of 107.8 kWh, while the present invention consumes an average of 87.9 kWh, resulting in an average energy saving rate of 18.5%. The energy saving rates under various abnormal conditions range from 16.8% to 19.1%, with no instances of energy saving only under a single operating condition and energy consumption increasing under combined operating conditions. In combined out-of-range conditions, the present invention does not simply put all relevant equipment into operation simultaneously. Instead, it selects operating equipment and adjustment amounts based on the grain condition node to be regulated, the equipment action sealing chain, and the results of candidate action matching. Some equipment actions are blocked in advance during the sealing chain propagation, and the control amount is not further increased after the equipment reaches the target operating ratio, reducing the energy consumption caused by high-frequency operation of fans, repeated start-stop of coolers, and long-term full-load operation of dehumidifiers.
[0053] The number of invalid actions decreased from an average of 3.1 times to 0.4 times. When single grain temperature, warehouse humidity, oxygen concentration, and carbon dioxide concentration were abnormal, this invention did not exhibit any invalid actions. However, when grain moisture content or multiple states were coupled, a single mismatch still occurred in some operations. This was because grain moisture content changes slowly, and changes in local airflow channels could cause the initial change time to shift. The mismatch was isolated from the update of the sealing edge and did not further alter the equipment's sealing chain. Only after subsequent continuous verification of actual grain condition changes was the corresponding sealing edge unlocked and reconnected, preventing a single offset from causing continuous errors in the subsequent control path.
[0054] The above data demonstrates that this invention solves the problems in conventional grain warehouse environmental control, such as difficulty in confirming the effective range of equipment, inconsistent responses from multiple detection points, easy over-adjustment of fixed control quantities, and miswriting of abnormal changes into control relationships. An improved MTGNN is used to form a candidate correlation graph, a balanced action probe is used to obtain the actual response sequence of equipment, an equipment action verification chain is used to limit the propagation range of candidate control actions, and actual grain condition change write-back is used to update the verified action relationships. Each stage uses the same grain condition node, propagation direction, and arrival order for connection, resulting in improvements in control time, over-adjustment amplitude, energy consumption, and the number of invalid actions. The implementation results support the feasibility of this invention in intelligent control of grain warehouse storage environments.
[0055] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.
Claims
1. A method for intelligent control of grain storage environment based on deep learning, characterized in that, Includes the following steps: The grain storage environment status is obtained, and an improved MTGNN is used to map the grain condition detection location and the location of the environmental regulation equipment to grain condition nodes, generating a candidate association graph of grain condition nodes and determining the grain condition nodes to be regulated. The control environment regulation equipment performs a balancing action probe consisting of continuous micro-movements in opposite directions, with the cumulative adjustment amount returning to zero; Track the balance action probe along the candidate correlation graph of grain condition nodes, and identify the grain condition nodes that reproduce the sequence of continuous micro-actions as response witness nodes; According to the arrival order of the response witness nodes, the continuous candidate association edges that connect the response witness nodes sequentially from the grain condition node corresponding to the location of the environmental regulation equipment are converted into sealing edges to form the equipment action sealing chain. Generate candidate control actions for the grain plot points to be regulated. After binding the candidate control actions with the source of the corresponding balance action probe, push them into the equipment action sealing chain. The action only propagates along the sealing edge that is consistent with the source of the balance action probe bound to the candidate control action and has the same propagation direction and arrival order. Propagation stops when encountering an unsealed candidate associated edge. The target control action is determined from the candidate control actions that run along the equipment's sealing chain to the grain condition point to be regulated, and the operation of the environmental regulation equipment is controlled. Obtain actual changes in grain conditions, write back the actual changes in grain conditions that match the sealing chain of the equipment action to the sealing edge, isolate the actual changes in grain conditions that do not match from the sealing edge update, and update the sealing chain of the equipment action.
2. The intelligent control method for grain storage environment based on deep learning according to claim 1, characterized in that, The determination of the grain condition node to be regulated specifically includes: Input the grain storage environment status into the improved MTGNN to generate candidate association edges with propagation direction between grain condition nodes; The state that exceeds the grain storage safety boundary is transmitted in reverse along the candidate associated edge to the grain condition node corresponding to the location of the environmental regulation equipment, forming a reverse transmission chain; Freeze the state of the grain situation nodes outside the reverse propagation chain, and exchange the arrival order of adjacent propagation states within the reverse propagation chain. The state that exceeds the grain storage safety boundary is reversed according to the adjustment direction of the environmental adjustment equipment, and then returned along the reverse transmission chain according to the arrival order before and after the exchange. The grain condition nodes where the pre-exchange return status is matched with the corresponding grain warehouse storage environment status, and where the post-exchange return status is not matched with the corresponding grain warehouse storage environment status, are identified as grain condition nodes to be regulated.
3. The intelligent control method for grain storage environment based on deep learning according to claim 1, characterized in that, The execution process of the balancing action probe specifically includes: Based on the current operating level of the environmental control equipment, the first micro-action, the reverse micro-action, and the reset micro-action are applied sequentially. Set the first micro-action and the reset micro-action to be in the same direction and equal in amount, and set the reverse micro-action to be opposite in direction to the first micro-action and the adjustment amount equal to the sum of the first micro-action and the reset micro-action. Maintain a constant operating level of the environmental control equipment between adjacent continuous micro-actions; The directional sequence of continuous micro-movements is bound to the holding duration as a balance motion probe.
4. The intelligent control method for grain storage environment based on deep learning according to claim 3, characterized in that, The determination of the grain condition node that reproduces the sequence of continuous micro-actions as the response witness node specifically includes: Based on the starting time of the balancing action probe, the state changes of each grain condition node are segmented according to continuous micro-actions and adjacent holding times; Extract the direction of change and the time of the first change of each segment, determine the change interval based on the adjacent times of the first change, determine the earliest time of the first change as the first response time of the grain condition node, and form the response order of the grain condition node according to the order of the generation of the direction of change. Reverse the direction of each micro-action in the direction sequence of the balancing action probe to generate a reverse direction sequence, and align the grain condition node response sequence segment by segment with the direction sequence and reverse direction sequence of the balancing action probe. Remove grain condition nodes that have interrupted directional alignment, inconsistent change intervals with corresponding retention times, or completed reverse directional alignment. The grain situation nodes that have completed the alignment of the directional order and whose reverse directional order alignment is interrupted are identified as response witness nodes. The response witness nodes are arranged according to the time of the first response to form the arrival order of the response witness nodes.
5. The intelligent control method for grain storage environment based on deep learning according to claim 4, characterized in that, The forming device function sealing chain specifically includes: Starting from the grain condition node corresponding to the location of the environmental regulation equipment, candidate related edges are sequentially connected according to the first response time of the response witness node; Candidate associated edges between the grain condition node corresponding to the location of the environmental regulation equipment and the first response witness node, between adjacent response witness nodes, and with the same source of the balancing action probes and the same propagation direction and arrival order, are converted into sealing edges. The continuation process will stop if any of the following occurs: candidate associated edges are missing, propagation direction is reversed, or arrival order is reversed. Connect the continuously formed sealing edges before the discontinuation to form the equipment function sealing chain.
6. The intelligent control method for grain storage environment based on deep learning according to claim 5, characterized in that, The step of binding the candidate control action with the source of the corresponding balance action probe and then pressing it into the device action verification chain specifically includes: The candidate control action is bound to the source of the corresponding balance action probe and written into the grain condition node corresponding to the location of the environmental regulation equipment. Set the first sealing edge of the sealing chain to the unlocked state and the remaining sealing edges to the locked state. Propagate along the edges of the sealing chain in the order of their succession to form the forward propagation state corresponding to the candidate control action. After the forward propagation state reaches the response witness node, the source, propagation direction and arrival order of the balance action probe of the next sealing edge and the previous sealing edge are checked. If the check matches, the next sealing edge is unlocked and the remaining subsequent sealing edges are kept locked. If the check does not match, the subsequent sealing edges are locked. Retain candidate control actions that connect to the grain processing point to be regulated, and eliminate candidate control actions that interrupt the transmission.
7. The intelligent control method for grain storage environment based on deep learning according to claim 6, characterized in that, The determination of the target control action specifically includes: The forward propagation state corresponding to the candidate control action that runs through to the grain level to be regulated is reversed in the opposite adjustment direction to form a return state, and is returned in reverse order along the corresponding equipment action sealing chain. At each response witness node, the returned status is sequentially matched with the forward propagation status corresponding to the candidate control action; Eliminate candidate control actions that trigger an interruption at any response witness node; The target control action is determined from the candidate control actions that have reached the grain condition node corresponding to the location of the environmental control equipment in the return state and have completed segment-by-segment engagement.
8. The intelligent control method for grain storage environment based on deep learning according to claim 7, characterized in that, The step of writing back the actual grain condition changes of the matching device's sealing chain to the sealing edge specifically includes: The actual grain situation changes are arranged according to the moment of the first change of the response witness node; Align the actual changes in grain conditions with the forward propagation states corresponding to the target control actions node by node; At each response witness node, verify the direction and order of change of the actual grain situation and the direction and order of propagation of the corresponding sealing edge; The actual changes in grain conditions that are continuously verified and consistent are defined as matching the actual changes in grain conditions; Set the sealing edge in the sealing chain to the locked state, unlock the sealing edge corresponding to the actual grain condition change and write back the actual grain condition change, and keep the sealing edge not corresponding to the actual grain condition change locked.
9. The intelligent control method for grain storage environment based on deep learning according to claim 8, characterized in that, The method of isolating unmatched changes in actual grain conditions from the update of the sealing certificate specifically includes: Align the response witness nodes in reverse order along the equipment action sealing chain that do not match the actual grain condition changes; Determine the direction of change and the first inconsistent response witness node in the arrival order, and cut off the writeback to the corresponding sealing edge for the unmatched actual grain situation change; Keep the corresponding seals and locks closed, and write back the grain storage environment status of the corresponding grain storage point if there are no matching actual changes in grain conditions. When subsequent changes in actual grain conditions are continuously verified against the corresponding sealing edge from the first discrepancy response witness node, the corresponding sealing edge will be unlocked.
10. The intelligent control method for grain storage environment based on deep learning according to claim 9, characterized in that, The updated device acts as a verification chain, specifically including: According to the order of the sealing edges, retain the sealing edges that have completed the writing of actual grain condition changes; Disconnect the first locked sealing edge and subsequent sealing edges from the sealing chain that functions with the equipment; According to the first change time of the response witness node, the unlocked sealing edge will be reconnected to the device function sealing chain one by one; Connecting the retained sealing edges with the reconnected sealing edges forms a device-controlled sealing chain that extends to the grain condition nodes to be regulated.
