Intelligent decision-making method and system based on computing-network integrated industrial PLC controller

Abstract

The present invention relates to the field of intelligent decision-making technology, and discloses an intelligent decision-making method and system based on an integrated computing and network industrial PLC controller. The method comprises: allocating time slices and dividing the PLC controller decision cycle for the control parameters of water supply equipment to obtain an initial control parameter queue; determining the probability distribution of normal operation, efficient control, and decision conflict states to obtain a water supply parameter conflict probability table; reallocating parameter control time resources to obtain an optimized control sequence for water supply parameters; converting parameter groups that still have conflicts into independent control variables and performing optimization calculations on each of them to obtain multi-parameter parallel control instructions for water supply equipment; and dynamically scheduling resources for the multi-parameter parallel control instructions for water supply equipment to generate an energy-saving control solution for water supply equipment. This method effectively optimizes the dynamic allocation of control resources, significantly shortens the response time of control parameters, and improves the overall throughput of the system.

Description

Technical Field

[0001] The present invention relates to the field of intelligent decision-making technology, and in particular to an intelligent decision-making method and system based on an integrated computing and network industrial PLC controller. Background Art

[0002] Traditional PLC controllers use fixed priorities or round-robin scheduling for resource scheduling. This often leads to decision conflicts and control delays when multiple water supply control parameters (such as water pressure, flow, water level, and water quality) compete for limited controller processing resources. In actual water supply system operation, the request frequency and priority of each control parameter change dynamically. Fixed scheduling strategies are unable to adapt to this complex operating environment, often resulting in high-priority parameters not being processed promptly or low-priority parameters not receiving a response for extended periods of time.

[0003] While computing-network integration offers a new approach to addressing these challenges, practical applications still present numerous challenges. Complex correlations exist between parameters in water supply equipment control systems, making traditional independent parameter optimization methods ineffective in addressing conflicts and coordination issues. Furthermore, water supply equipment is typically distributed across diverse locations and has varying energy states, posing a pressing challenge in minimizing system energy consumption while ensuring control performance. Furthermore, as the system scales up, the number of control parameters increases dramatically, creating a bottleneck for traditional approaches in achieving efficient multi-parameter parallel decision-making within limited computing resources. Summary of the Invention

[0004] The present invention provides an intelligent decision-making method and system based on an integrated computing and network industrial PLC controller, which effectively optimizes the dynamic allocation of control resources, significantly shortens the response time of control parameters and improves the overall system throughput.

[0005] In a first aspect, the present invention provides an intelligent decision-making method based on an integrated computing and network industrial PLC controller, the intelligent decision-making method based on an integrated computing and network industrial PLC controller comprising:

[0006] The control parameters of the water supply equipment are allocated in time slices and divided into decision cycles by the PLC controller to obtain the initial control parameter queue;

[0007] Determine the probability distribution of normal operation state, efficient control state and decision conflict state according to the initial control parameter queue, and obtain a water supply parameter conflict probability table;

[0008] Reallocating parameter control time resources based on the water supply parameter conflict probability table to obtain a water supply parameter optimization control sequence;

[0009] Converting the parameter groups that still have conflicts in the water supply parameter optimization control sequence into independent control variables, and performing optimization calculations on each independent control variable to obtain multi-parameter parallel control instructions for the water supply equipment;

[0010] Dynamic resource scheduling is performed on the multi-parameter parallel control instructions of the water supply equipment to generate an energy-saving control plan for the water supply equipment.

[0011] In a second aspect, the present invention provides an intelligent decision-making system based on an integrated computing and network industrial PLC controller, the intelligent decision-making system based on an integrated computing and network industrial PLC controller comprising:

[0012] The first allocation module is used to allocate time slices and divide the PLC controller decision cycle of the control parameters of the water supply equipment to obtain an initial control parameter queue;

[0013] An analysis module is used to determine the probability distribution of normal operation state, efficient control state and decision conflict state according to the initial control parameter queue, and obtain a water supply parameter conflict probability table;

[0014] A second allocation module is configured to reallocate parameter control time resources based on the water supply parameter conflict probability table to obtain a water supply parameter optimization control sequence;

[0015] An optimization calculation module is used to convert the parameter groups that still have conflicts in the water supply parameter optimization control sequence into independent control variables, and perform optimization calculations on each independent control variable to obtain multi-parameter parallel control instructions for the water supply equipment;

[0016] The resource scheduling module is used to dynamically schedule the multi-parameter parallel control instructions of the water supply equipment and generate an energy-saving control plan for the water supply equipment.

[0017] In the technical solution provided by the present invention, accurate allocation of time slices is performed through probability modeling based on Bayesian and Poisson distribution models, which effectively optimizes the dynamic allocation of control resources, significantly shortens the response time of control parameters and improves the overall throughput of the system; FastICA independent component analysis technology is used to convert conflicting parameters into independent control variables, and parallel optimization processing of multiple conflicting parameters in the same time slice is realized, which greatly reduces the parameter processing conflict rate and improves the resource utilization efficiency of the PLC controller; combined with the dynamic resource scheduling mechanism of the energy state of the water supply equipment, the multi-objective optimization of control performance and energy consumption is achieved through the Q-learning reinforcement learning algorithm. The total energy consumption of the system is reduced while ensuring control accuracy, and the service life of the water supply equipment is extended; the time resource reallocation strategy based on parameter correlation analysis and conflict sensitivity index enhances the system's adaptability to parameter request fluctuations, and can maintain stable operation even when the parameter request frequency suddenly increases; the task allocation mechanism corrected by the energy factor realizes computational load balancing that takes into account the remaining energy status of the equipment, effectively avoiding the problem of single-point device resource overload; the parameter grouping strategy based on the graph coloring algorithm and the multi-parameter parallel control architecture increase the number of water supply equipment that can be managed simultaneously by a single computing and network-integrated PLC controller, significantly enhancing the system's scalability.

[0018] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or understood by practicing the present invention. The purposes and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0019] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 A schematic diagram of an embodiment of an intelligent decision-making method based on an integrated computing and network industrial PLC controller in an embodiment of the present invention;

[0021] Figure 2 This is a schematic diagram of an embodiment of an intelligent decision-making system based on an integrated computing and network industrial PLC controller in an embodiment of the present invention. DETAILED DESCRIPTION

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0023] The terms "including," "having," and any variations thereof, as used in the embodiments of the present invention are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or device comprising a series of steps or units is not limited to the listed steps or units, but may optionally include other steps or units not listed, or may optionally include other steps or units inherent to the process, method, product, or device.

[0024] To facilitate understanding of this embodiment, first, an intelligent decision-making method based on an integrated computing and network industrial PLC controller disclosed in an embodiment of the present invention is introduced in detail. Figure 1 As shown, this method includes the following steps:

[0025] 101. Allocate time slices and divide the PLC controller decision cycle for the control parameters of the water supply equipment to obtain an initial control parameter queue;

[0026] It is understandable that the execution subject of the present invention can be an intelligent decision-making system based on an integrated computing and network industrial PLC controller, or a terminal or a server, which is not limited here. The embodiment of the present invention is described by taking the server as the execution subject as an example.

[0027] Specifically, the control parameters of the water supply system are collected. The initial collection process includes core parameters that affect system operation, such as water pressure, flow, water level, and water quality, to form an initial set of water supply parameters. Each parameter in this set is assigned a priority value based on its importance in system control. This priority value is quantified as an integer from 1 to 10, with higher values ​​indicating greater urgency or importance. Furthermore, a maximum allowable delay time is set for each parameter. This value, measured in milliseconds, constrains the maximum acceptable response time between parameter request and execution. To more accurately reflect the computational load of the PLC controller when processing different parameters, each parameter is assigned a computing resource utilization value, expressed as the percentage of CPU resources required during the control cycle. These three types of data together constitute a parameter attribute matrix, with each row representing a parameter and each column corresponding to the priority value, maximum allowable delay time, and computing resource utilization. Based on the parameter attribute matrix, the initial time slice allocation for each parameter is completed through a refined calculation formula. The calculation process weights the priority value and resource utilization, multiplies it by the length of the entire PLC control cycle, and normalizes it using the system adjustment coefficient. This ensures that time slice allocation reflects the parameter's importance while remaining consistent with the total scheduling cycle. Upon completion, the parameter time slice allocation results are obtained. The initial water supply parameter set is sorted by parameter priority, generating an initial control instruction set. Each instruction set entry contains a control parameter identifier, the corresponding time slice length, and its control logic command set. Simultaneously, a system state matrix is ​​established to record the operating parameter values, current energy status, and communication load information for each water supply device. This matrix is ​​structured as a two-dimensional array, with rows representing devices and columns representing device parameters such as operating frequency, motor power, and pump status. All elements are collected and updated in real time via the industrial fieldbus, ensuring the timeliness and accuracy of the matrix data. The parameter time slice allocation results, the initial control instruction set, and the device operating state matrix are stored in the PLC controller's shared memory area, resulting in the initial control parameter queue.

[0028] 102. Determine the probability distribution of normal operation state, efficient control state and decision conflict state according to the initial control parameter queue, and obtain a water supply parameter conflict probability table;

[0029] Specifically, based on the initial control parameter queue, the unique identifiers of the control parameters are matched with data recorded in the database over previous control cycles to extract a historical data set including normal operation, efficient control, and scheduling conflict states. This data reflects the true relationship between the system state of each parameter in actual operation and its corresponding control behavior. The historical data set is analyzed using conditional probability statistics to assess the probability of each control parameter occurring in different system states, revealing the correlation between the parameter and the system state. Furthermore, based on the request frequency of each parameter in the initial control parameter queue in each time slice, a probability distribution model for the number of requests per unit time is constructed. This model reflects the intensity and fluctuation trend of parameter scheduling requests, enabling the system to predict the likelihood of multiple parameters converging to request control resources in a specific time slice. Based on the conditional probability values ​​of the parameter states and the probability distribution of the parameter request frequency, the probability of decision conflicts occurring in each time slice is calculated. Based on the distribution characteristics of the current parameter queue and historical conflict cases, the time slices most likely to experience control conflicts between parameters are determined. When multiple high-priority or high-request frequency parameters exist simultaneously within a time slice, and these parameters have historically experienced resource competition, that time slice is included in the conflict time slice sequence and becomes a key scheduling monitoring target. To reveal the conflict relationships between parameters, a parameter conflict probability matrix is ​​constructed based on the resource allocation within the time slice. In this matrix, the element values ​​between each pair of parameters represent the likelihood of control conflict occurring between them within the same time slice. On this basis, a conflict sensitivity is calculated for each control parameter. This sensitivity reflects the degree to which a parameter is prone to conflict within the entire scheduling system. The conflict sensitivity value of each parameter is integrated with the conflict probability matrix to form a water supply parameter conflict probability table.

[0030] In this embodiment, the actual number of requests for each control parameter within different time slices is extracted from a historical dataset. Based on the extensive operational records accumulated during the long-term operation of the water supply control system, the time points at which each parameter was scheduled during each control cycle are analyzed to restore the request distribution characteristics within different time slices. This creates a preliminary parameter request frequency dataset, reflecting the real-time requirements and call density of the parameters. Pattern recognition is performed on these parameter request frequencies, clustering and grouping them by time slice. The changing trends observed across different time periods and operating scenarios are then extracted to determine the request frequency variation patterns. Based on the request frequency variation patterns, the Poisson distribution parameter λ is determined for each parameter in different operating scenarios. This parameter describes the average number of parameter requests per unit time. Using this λ value, the probability of k simultaneous requests arriving within each time slice is quantified, generating a parameter request quantity probability matrix. Each element in this matrix represents the probability of a specific parameter request reaching a certain frequency within a given time slice. Time slice and parameter combinations whose probability values ​​exceed a set threshold in the parameter request quantity probability matrix are marked and aggregated into a list of time slices with a high potential conflict risk in scheduling. This list serves as a key monitoring area for control system operation scheduling, reflecting critical time windows where problems such as request concentration, priority conflicts, or response delays may occur in the short term. Based on this list, control parameter priority information is introduced, and the intersection structure of different priority parameters in the same time slice is modeled to construct a time slice resource competition distribution map. This distribution map represents scheduling resource contention relationships using a graph structure, where nodes represent parameters and edge weights represent conflict probability or competition intensity. This helps identify core nodes and critical paths that are most prone to contention. The time slice resource competition distribution map is fused with the parameter request quantity probability matrix. By jointly evaluating the node priority weights in the graph and the request probabilities in the matrix, a parameter request frequency probability distribution is formed.

[0031] 103. Reallocate parameter control time resources based on the water supply parameter conflict probability table to obtain the optimized control sequence of water supply parameters;

[0032] Specifically, the parameter conflict matrix and parameter priority values ​​are extracted from the water supply parameter conflict probability table. By analyzing the conflict probability between each parameter in the conflict matrix and weighting the conflict values ​​with the corresponding parameter priority values, a conflict sensitivity index is calculated for each parameter, resulting in a set of conflict sensitivity indices for the entire control parameter set. The deviation rate from the average conflict sensitivity index is calculated based on the parameter conflict sensitivity index set to quantify the relative prominence of each parameter in terms of conflict risk. To ensure flexible resource adjustment, each deviation rate is multiplied by a preset adjustment intensity factor. This factor is set based on the control system's requirements for dynamic adjustment sensitivity and fluctuates within a reasonable range. The product is the time slice adjustment coefficient for that parameter, indicating the ratio by which its current time slice should be enlarged or reduced. The time slice adjustment coefficient is used to modify the time slice values ​​initially allocated based on priority and resource demand, resulting in adjusted time slice allocations for each parameter. This adjustment process ensures that highly conflict-sensitive parameters receive more control resources to avoid scheduling congestion and ensures that the total time slice allocation does not exceed the upper limit of the decision cycle, maintaining system scheduling balance. The correction result establishes a dynamic mapping relationship between time slice resources and parameter conflict risks. Based on the data in the water supply parameter conflict probability table, a parameter relationship topology diagram is constructed. Each control parameter is used as a node in the diagram. If there is a conflict relationship between any two parameters, a connecting edge is established between the nodes, and the weight of the edge is set to the corresponding conflict probability value. The topology diagram is processed for conflict avoidance. According to the edge weight, the parameter pairs with a conflict probability higher than the preset threshold are identified, and a forced grouping strategy is used to assign these parameters to different time slice groups to ensure that they do not request resources at the same time in the same time slice, thereby reducing the probability of scheduling conflicts. Through the dual mechanisms of time slice correction and parameter topology reconstruction, the water supply parameter optimization control timing is output.

[0033] 104. Convert the parameter groups that still have conflicts in the water supply parameter optimization control sequence into independent control variables, and perform optimization calculations on each independent control variable to obtain multi-parameter parallel control instructions for the water supply equipment;

[0034] Specifically, from the water supply parameter optimization control time series, parameter groups with significant conflict probabilities within the same time slice were identified. These parameter groups, previously decoupled due to their complex coupling and high-frequency request behavior, were labeled as conflicting parameter sets, and correlation data between the parameters was extracted. A parameter correlation coefficient matrix was constructed based on historical operation records and actual sampling results. This matrix quantifies the strength and direction of coordinated changes between parameters and serves as an important basis for determining whether potential control interference exists. Historical observations of the conflicting parameter groups were rearranged in a format with parameters as rows and time points as columns to form an original observation data matrix. This matrix captures the time-varying behavior patterns of each parameter over multiple control cycles. Based on this, an independent component analysis algorithm was introduced to process the matrix, extracting a set of independent control variables from the mixed observation data. By converting the original highly correlated parameter set into a set of statistically independent control factors, the coupling structure between the parameters was broken. A weighted objective function, including control delay and energy consumption terms, was established for each independent control variable. The control delay term reflects the time required from parameter request to execution completion, and the energy consumption term quantifies the energy load caused by the control operation. The weights of the two are set according to the sensitivity to response speed and energy consumption in actual applications. In order to effectively solve the objective function, the gradient descent method is used to iteratively optimize each independent control variable. By continuously adjusting the value direction and step size of the control variable, it gradually converges to the optimal solution of the objective function and obtains the optimized independent control variable value. The optimized independent control variable value is converted back to the parameter space and inversely transformed through the mixing matrix established by the previous independent component analysis to ensure that the control behavior of each independent control factor on the original parameter is restored to the form of executable commands. Based on the restored parameter control quantity, multi-parameter control instructions for water supply equipment that can be executed in parallel within the same time slice are generated.

[0035] 105. Dynamically schedule resources for multi-parameter parallel control instructions of water supply equipment and generate energy-saving control plans for water supply equipment.

[0036] Specifically, a system state space is constructed to describe the current control environment. This state space includes the set of currently active parameters, the priority values ​​of each parameter, the specific allocation of time slices, the energy status of each water supply device, and the communication load of each node in the system. This state space construction provides a comprehensive understanding of the dynamic changes in resources involved in the control process. Based on this, an action space is defined, which includes a series of optional resource scheduling operations, such as dynamic adjustment of resource allocation ratios, optimization of communication channels, and temporary changes in control parameter priorities. A system state-action mapping table is formed by mapping system states to action contents. The current energy reserves of each water supply device are monitored in real time, and the energy factor is defined as the ratio of its current remaining energy value to its fully charged state value. The introduction of the energy factor enables the system to dynamically adjust the control computation capacity of each device based on its available energy. Therefore, the original basic computational load of each device is modified based on its energy factor, resulting in a computational task allocation that better reflects the device's actual state. A reward function is constructed based on the average response delay, total energy consumption, and control accuracy of the water supply devices, and a comprehensive performance evaluation index is calculated using the reward function. A Q-learning reinforcement learning algorithm is implemented based on the system state-action mapping table, device computational task allocation, and comprehensive performance evaluation indicators. During training, the system continuously establishes the optimal mapping relationship between state and action through trial and error, gradually optimizing the Q-value table and, through learning from historical experience and environmental feedback, obtaining the optimal control strategy. Based on the optimal control strategy, the execution devices and communication channels for each control task in the multi-parameter parallel control instructions for the water supply equipment are assigned, forming a structured task allocation scheme. This task allocation scheme is then integrated with the original multi-parameter parallel control instructions to form an energy-saving control scheme for the water supply equipment.

[0037] In this embodiment, the continuous system state space and action space are discretized. State parameters, such as device energy levels, communication load, and control response delay, which originally consisted of numerous floating-point or continuous values, are divided into finite levels or intervals, such as high, medium, and low, making the state variables easier to encode and index. Similarly, all optional actions are discretized, for example, by setting resource allocation ratios to a few predefined levels or limiting communication channel selection to a finite number of paths. This creates a two-dimensional table structure consisting of states and actions. Based on this structure, a Q-value table, or Q matrix, is initialized. Each row represents a system state and each column represents a possible action. Each element in the table is initially set to the same value and is used to record the expected benefits of taking specific actions under different states. When the system receives a control instruction generated by the multi-parameter parallel control module, it identifies the corresponding system state based on the current control context, device state, and time-slice resource allocation. It then calculates an immediate reward value based on the execution results of the control instruction and a pre-established comprehensive performance evaluation index system. This reward value comprehensively considers the system's average response time, energy consumption, and control accuracy, reflecting the actual benefits of the selected action in the current state and serving as an important basis for updating the Q-value. The system locates the corresponding state-action pair in the Q-value table corresponding to the current state and assigns the immediate reward to it. The action selection strategy employs an ε-greedy approach, selecting the optimal action with the maximum Q-value in the current state with high probability in some cases and randomly selecting with lower probability in other cases, thus achieving a balance between exploration and exploitation. By dynamically switching between maximum selection and random exploration, the system fully leverages existing learning results while avoiding getting stuck in local optima and missing out on better solutions. After selecting an action, the system executes it during actual control and calculates an update to the Q-value based on feedback, such as the transition to the new state and the new reward. This update is determined by combining the old Q-value in the current state, the immediate reward, and the maximum Q-value of all actions in the new state. The update is then weighted averaged using the learning rate and discount factor to replace the corresponding position in the old Q-value table. The entire reinforcement learning process consists of four steps: action selection, state observation, immediate reward evaluation, and iterative Q-value updates. This cycle continues throughout the system's operation. As experience accumulates, the values ​​in the Q-value table gradually converge, approaching the long-term expected reward in the real environment. After the Q value converges, the action with the maximum Q value is extracted from the Q value table for each discretized system state as the optimal operation choice for that state. The set of optimal actions in all states constitutes the optimal control strategy output.

[0038] In the embodiment of the present invention, the time slice is accurately allocated by probabilistic modeling based on Bayesian and Poisson distribution models, which effectively optimizes the dynamic allocation of control resources, significantly shortens the response time of control parameters and improves the overall throughput of the system; the FastICA independent component analysis technology is used to convert conflicting parameters into independent control variables, and the parallel optimization processing of multiple conflicting parameters in the same time slice is realized, which greatly reduces the parameter processing conflict rate and improves the resource utilization efficiency of the PLC controller; combined with the dynamic resource scheduling mechanism of the energy state of the water supply equipment, the Q-learning reinforcement learning algorithm is used to achieve multi-objective optimization of control performance and energy consumption, while ensuring The total energy consumption of the system is reduced while maintaining control accuracy, thereby extending the service life of the water supply equipment. The time resource reallocation strategy based on parameter correlation analysis and conflict sensitivity index enhances the system's adaptability to parameter request fluctuations, and can maintain stable operation even when the frequency of parameter requests suddenly increases. The task allocation mechanism corrected by the energy factor achieves computational load balancing that takes into account the remaining energy state of the equipment, effectively avoiding the problem of single-point device resource overload. The parameter grouping strategy based on the graph coloring algorithm and the multi-parameter parallel control architecture increase the number of water supply equipment that can be managed simultaneously by a single computing-network integrated PLC controller, significantly enhancing the system's scalability.

[0039] In a specific embodiment, the process of executing step 101 may specifically include the following steps:

[0040] The control parameters in the water supply system are collected to obtain an initial water supply parameter set, and each parameter in the initial water supply parameter set is assigned a parameter priority value, a maximum allowable delay time value, and a computing resource occupancy value to obtain a parameter attribute matrix;

[0041] Calculate the initial time slice allocation of each parameter according to the parameter attribute matrix to obtain the parameter time slice allocation result;

[0042] Sorting the initial water supply parameter set according to the parameter priority value to obtain the initial control instruction set;

[0043] Establish a system status matrix containing the operating parameter values, energy status values ​​and communication load values ​​of each water supply equipment, and collect and update the values ​​of each matrix element in real time through the industrial field bus to obtain the equipment operating status information;

[0044] The parameter time slice allocation result, the initial control instruction set and the equipment operation status information are stored in the shared memory area of ​​the PLC controller to obtain the initial control parameter queue.

[0045] Specifically, the control parameters of the water supply system are collected to form an initial water supply parameter set. Through data communication with various devices at all levels of the water supply system and utilizing various sensors deployed at key locations, fine-grained monitoring of the system's operating status is performed. Collected parameters include water pressure, water level, instantaneous flow, cumulative flow, pump speed, valve opening, water quality indicators, as well as key control variables affecting water supply stability and energy efficiency, such as ambient temperature and humidity, motor temperature rise, and power consumption. All collected raw data is formatted, denoised, and de-anomalyed in a preprocessing module to form the initial water supply parameter set. Each control parameter in the initial water supply parameter set is assigned three attribute values: parameter priority, maximum allowable delay, and computing resource utilization. Parameter priority values ​​are set based on the importance of the control task or automatically learned from historical control behavior. They are graded using integers, with higher values ​​indicating a more urgent control need. The maximum allowable delay reflects the maximum response time allowed between the trigger request and execution of the parameter, determining its suitability for parallel scheduling with other parameters. The calculated resource utilization value represents the percentage of processing resources required by the PLC controller to process the parameter control task, taking into account factors such as CPU time, memory access load, and communication bandwidth. By organizing the three attribute values ​​into a structured data table, a parameter attribute matrix is ​​constructed, with each row corresponding to a parameter and each column corresponding to an attribute item. Based on the parameter attribute matrix, the time slice allocation for all control parameters is calculated. A time slice is the computation and response time window allocated by the PLC controller to a parameter within each control cycle. Its length determines the priority of parameter scheduling and the density of resource allocation. During the time slice allocation process, a weighted ratio of priority and resource utilization is introduced for normalization. This ensures that high-priority parameters with moderate resource utilization receive more control time, while low-priority or high-resource utilization parameters have their time slices appropriately compressed, achieving overall scheduling load balancing. This process dynamically allocates time slice lengths for each parameter while ensuring that the total control cycle time remains unchanged, generating a parameter time slice allocation result. This result is used to guide the periodic scheduling frequency and specific response time of the parameter. The initial set of water supply parameters is sorted by parameter priority, with parameters with higher priority values ​​placed at the front of the list. This prioritizes control tasks for highly urgent parameters during the scheduling process. The sorting results serve as the basis for the initial set of control instructions. Each control instruction includes parameter identification, priority information, allocated time slice length, and a preset control strategy template, forming the core unit of the scheduling queue. This instruction set serves as the primary basis for task distribution within each control cycle, with the PLC controller reading and executing the corresponding control commands one by one. At the same time, a system status matrix is ​​established to describe the current operating status of the water supply equipment. The construction of this matrix relies on data collected by the industrial field bus and the internal sensor units of each device.The rows of the matrix correspond to specific devices or functional units in the water supply system, such as pumps, motors, reservoirs, and valve control modules. The columns represent various operating parameters, such as voltage, current, speed, temperature, vibration, communication load, and remaining battery power. Each matrix cell represents the real-time value of a specific operating indicator for a device. The matrix is ​​continuously updated via a high-speed communication interface with the PLC controller to ensure that the scheduling algorithm receives the latest operating information. The time-slice allocation results, the initial control instruction set, and the device operating status information from the system state matrix are stored in the PLC controller's shared memory area. This shared memory serves as a common data interface between the multi-threaded scheduling module, the system state monitoring module, and the control execution module. Its structure is designed for high-speed, low-latency access and includes a synchronization lock mechanism to prevent data read and write conflicts. The time-slice allocation results serve as the basic time structure for the parameter scheduling process, providing a time window reference. The initial control instruction set provides the scheduling order and specific control action instructions. The dynamic data in the system state matrix assists in determining whether certain parameters need to be delayed, skipped, or weighted. The initial control parameter queue formed by these three elements is executed step by step by the scheduling core module according to a time-triggered method.

[0046] In a specific embodiment, the process of executing step 102 may specifically include the following steps:

[0047] According to the initial control parameter queue, a historical data set including normal operation status, efficient control status and decision conflict status is extracted from the historical database of the integrated computing and network industrial PLC controller;

[0048] Perform conditional probability distribution calculation on historical data sets to obtain parameter state conditional probability values;

[0049] According to the parameter request frequency in each time slice, the probability of occurrence of different request quantities per unit time is calculated to obtain the parameter request frequency probability distribution;

[0050] Based on the parameter state conditional probability value and the parameter request frequency probability distribution, the probability value of decision conflict in each time slice is calculated to obtain the conflict time slice sequence;

[0051] Construct a parameter conflict probability matrix, calculate the probability value of decision conflict for each pair of parameters in the same time slice, and obtain the parameter conflict relationship within the time slice;

[0052] The conflict sensitivity value of each parameter in the conflict time slice sequence is calculated to obtain the parameter conflict sensitivity sequence, and the parameter conflict sensitivity sequence and the parameter conflict relationship within the time slice are combined into a water supply parameter conflict probability table.

[0053] Specifically, all parameter identifiers contained in the control parameter queue are used to extract relevant operation log data from the historical database embedded in the integrated computing and network industrial PLC controller. The extracted data is then classified into different states based on the operation result tags in the logs. This database accumulates a large amount of state data from actual operation processes and archives the system response results for each control cycle. These three categories are classified into normal operation, efficient control, and decision conflict. These correspond to situations where the system successfully completes control according to plan, achieves efficient resource allocation under optimization strategies, and experiences abnormal responses due to parameter concurrency conflicts or control resource competition. By filtering and matching these different state data by parameters, a multi-state historical dataset covering all key parameters is generated. The distribution of each parameter in the dataset under different system states is analyzed, and the state probability of each parameter under specific state conditions is calculated using conditional probability calculation methods. By analyzing the relationship between the parameter's historical state and its operating environment, the probability distribution of the parameter being in a normal, efficient, or conflict state is determined. This analysis uses parameters as known conditions and the system state as the outcome event, examining the frequency of the system's occurrence of different operating states under given parameter values ​​or scheduling conditions. The resulting set of conditional probability values ​​reflects the degree of impact of each parameter on system stability. Furthermore, based on the request frequency of each parameter within each time slice, the probability of the number of requests occurring per unit time is derived. Data recorded by the PLC controller regarding parameter request scheduling during each control cycle is used to count the number of parameters simultaneously requesting control resources within each discrete time slice, and the frequency distribution of these requests is recorded. Statistical analysis provides a comprehensive picture of scheduling density within a time slice, characterizing the probability trends of concurrent requests under different control load conditions. These statistical results are converted into a parameter request frequency probability distribution function, reflecting the relationship between the number of concurrent requests per unit time and their probability of occurrence. The conditional probability values ​​of the parameter states are combined with the probability distribution of the parameter request frequency to calculate the overall probability of potential decision conflicts within each time slice. Taking the time slice as a unit, parameter state characteristics are introduced based on the analysis of the parameter request frequency within that time period to determine the likelihood of their combined occurrence leading to a conflict. When high-frequency requests occur simultaneously with a highly conflict-sensitive state, the time slice is considered a critical scheduling window with a high conflict risk. This mechanism generates a sequence of conflict time slices, listing all time slice indices with high conflict probabilities and labeling the main parameters involved in the conflict within each time slice. To quantify the probability of scheduling conflicts between parameters within a specific time slice, a parameter conflict probability matrix is ​​constructed based on the data from the conflict time slices. In this matrix, the relationship between each pair of parameters is quantified as the historical probability of their co-initiating control conflict within the same time slice. Higher values ​​indicate a higher frequency of mutually exclusive or competitive scheduling behavior for that parameter pair.This matrix has the characteristics of symmetry and sparsity, and based on this, a conflict map between parameters is established for subsequent graph coloring optimization and parallel scheduling strategy design. The conflict sensitivity value is calculated for each participating parameter in the conflict time slice sequence. This value is a weighted summation of the importance of the parameter itself, that is, the priority factor, while statistically analyzing the probability of conflict with other parameters. The result reflects the centrality of the parameter in the competition for control resources. The higher the conflict sensitivity value, the more likely the parameter is to conflict with other parameters, and the higher its priority in the scheduling queue. Therefore, it needs to be given higher attention and scheduling priority during the resource scheduling process. By calculating the sensitivity values ​​of all parameters, a set of ordered parameter conflict sensitivity sequences is obtained. The conflict sensitivity sequence is merged with the parameter conflict probability matrix to form a water supply parameter conflict probability table.

[0054] In a specific embodiment, the step of calculating the probability of occurrence of different numbers of requests per unit time based on the parameter request frequency in each time slice to obtain the parameter request frequency probability distribution may specifically include the following steps:

[0055] Count the number of requests for each control parameter in different time slices from the historical data set to obtain the parameter request frequency;

[0056] Group the parameter request frequencies to obtain the request frequency change pattern;

[0057] Based on the request frequency change pattern, the Poisson distribution parameter λ value of each parameter in different operating scenarios is determined. The Poisson distribution parameter λ value is substituted into the Poisson distribution formula to calculate the probability distribution of k requests arriving at the same time, and the parameter request quantity probability matrix is ​​obtained.

[0058] Mark the time slices and parameter combinations whose probability values ​​in the parameter request quantity probability matrix exceed the set threshold, and obtain a list of time slices with high conflict risks;

[0059] Calculate the time slice resource competition distribution map using the high-conflict risk time slice list and parameter priority values;

[0060] Combine the time slice resource contention distribution graph with the parameter request quantity probability matrix to output the parameter request frequency probability distribution.

[0061] Specifically, by traversing the scheduling records of each time slice within a historical control cycle, information on whether each parameter initiated a control request within each time slice is extracted and archived by time slice, forming a three-dimensional data set containing control parameters, time slice indexes, and request counts. The statistical process considers the scheduling logs of the control system under different operating loads, external supply and demand conditions, and device states. Therefore, the request frequencies of the same parameter within different time periods are accumulated and normalized to obtain the request frequency of each parameter within different time slices, that is, the average number of scheduling requests initiated by the parameter per time slice. Parameter request frequencies are grouped to identify patterns in parameter request behavior. This grouping process classifies parameters based on request frequency, time slice index, and operating environment status, forming several representative request frequency pattern clusters. Using a clustering algorithm or threshold segmentation method, parameters with significant variation characteristics are grouped into the same request frequency variation pattern. Each pattern represents the typical behavior characteristics of a parameter type under a specific operating scenario. Based on the statistical characteristics of each frequency variation pattern, the value of the Poisson distribution parameter λ is calculated for each operating scenario. The Poisson distribution, a statistical tool for describing the probability of discrete events occurring per unit time, is suitable for modeling the behavior of control parameter requests within a time slice. The mean of the time slice request data contained in each frequency variation pattern is calculated and used as the basis for estimating the λ value, yielding the expected number of requests each parameter is likely to initiate in each time slice under a specific operating state. These λ values ​​are substituted into the standard Poisson distribution expression and solved for the probability of occurrence of different request numbers k. This generates a parameter request quantity probability matrix, which plots the time slice on the horizontal axis and the number of requests k on the vertical axis. The value within each matrix cell represents the estimated probability of reaching k parameter requests within that time slice. Time slices and parameter combinations whose probability values ​​exceed a set threshold in the parameter request quantity probability matrix are marked. When the request quantity probability value corresponding to a time slice in the matrix exceeds the threshold, it indicates that multiple parameters are collectively initiating scheduling requests within that time slice, exacerbating resource contention and causing scheduling delays or response failures. The system labels all time slices and their associated parameters that meet this condition and compiles them into a list of time slices with high conflict risk. Using this list of time slices and parameter priority values, a scheduling contention relationship diagram is constructed. The participating parameters in each high-risk time slice are extracted and weighted according to their priority levels. Based on this, a time slice resource contention distribution graph is constructed. In this graph, each node represents a control parameter, and edges between nodes represent conflicts between parameters that may simultaneously request the same scheduling resource. Edge weights are calculated based on the degree of request overlap and the difference in priority levels. This graph shows which parameters in a given time slice have the most intense resource contention and which are the key control factors for system response bottlenecks.The time slice resource competition distribution map is fused with the parameter request quantity probability matrix. The original probability matrix is ​​modified by weighting the correlation between the probability estimate of the request quantity in each time slice and the resource competition intensity. This matrix not only reflects the frequency of scheduling requests, but also expresses the conflict relationship between requests and the severity of resource contention, thus obtaining the parameter request frequency probability distribution model.

[0062] In a specific embodiment, the process of executing step 103 may specifically include the following steps:

[0063] According to the water supply parameter conflict probability table, the parameter conflict matrix and parameter priority values ​​are extracted to obtain the parameter conflict sensitivity index set;

[0064] Calculating the deviation rate from the average conflict sensitivity index based on the parameter conflict sensitivity index set, and multiplying the deviation rate by a preset adjustment intensity factor to obtain a time slice adjustment coefficient;

[0065] Correct the initially allocated time slice value according to the time slice adjustment coefficient to obtain the adjusted time slice allocation value;

[0066] A parameter relationship topology diagram is constructed according to the water supply parameter conflict probability table. Parameters with conflict probabilities higher than a preset threshold are forcibly assigned to different time slice groups based on the parameter relationship topology diagram to obtain the water supply parameter optimization control timing.

[0067] Specifically, the parameter conflict matrix and the priority values ​​of each control parameter are extracted from the conflict probability table. The parameter conflict matrix is ​​a symmetric structure whose elements represent the probability of scheduling conflict between any two control parameters within the same time slice. The priority value, a pre-assigned control importance weight for each parameter, serves as a key basis for system scheduling decisions and resource allocation. These two pieces of information are combined to calculate a conflict sensitivity index for each parameter. This index considers not only the frequency and intensity of conflicts between the parameter and other parameters, but also the control priority of the conflicting parameters within the system. By weightedly accumulating each conflicting edge with the priority of the other parameter, the conflict coreness of the parameter in the scheduling structure, or conflict sensitivity index, is derived. Based on the set of parameter conflict sensitivity indices, the deviation rate from the average conflict sensitivity index is calculated to measure the degree of deviation of each parameter from the average conflict sensitivity. The conflict sensitivity values ​​of all parameters are averaged, and the relative deviation rate of each parameter from the average is calculated to reflect the intensity of conflict and its relative position in resource contention within the overall scheduling structure. To ensure that the deviation rate plays a moderate role in the time-slice adjustment process, a preset adjustment intensity factor is introduced. This factor is set based on the actual control system's requirements for resource allocation sensitivity. A larger value indicates a greater tendency for the system to make significant time-slice adjustments based on sensitivity differences; conversely, a smaller value indicates a system that seeks a relatively stable scheduling structure. The adjustment intensity factor is multiplied by the deviation rate to obtain the time-slice adjustment coefficient. This coefficient provides a specific time-slice adjustment ratio for each parameter, determining the control window to be expanded or reduced in the next round of resource allocation. The initial time-slice allocation is then revised based on the time-slice adjustment coefficient. While maintaining the total control cycle, time-slice quotas are appropriately increased for highly sensitive parameters with high conflict risks to alleviate their scheduling pressure. Time-slices for less sensitive parameters or those in non-core conflict chains are appropriately reduced to free up scheduling resources for critical parameters. During this revision process, linear normalization or proportional balancing algorithms are used to ensure that the sum of all adjusted time-slices for all parameters remains within the set upper limit of the control cycle T, thereby avoiding system beat imbalance or scheduling overload caused by excessive adjustments. A parameter relationship topology map is constructed based on the water supply parameter conflict probability table. In this topological graph, each node represents a control parameter. Any two parameters with scheduling conflicts are connected by an edge, with the weight of the edge set to the conflict probability between them. The overall graph structure reflects the conflict propagation paths, scheduling coupling relationships, and the intensity of resource competition between parameters in the water supply control system. This graph can be used to identify conflict hotspots, conflict clusters, and scheduling bottleneck paths.To achieve more thorough conflict isolation, the topology graph is structured and colored. All parameter pairs in the graph with conflict probabilities exceeding a preset threshold are forcibly assigned to different time slice groups. This physically isolates them and ensures that they do not simultaneously request control resources within the same time slice during scheduling. This operation is equivalent to a round of conflict probability-based grouping of the control parameter set. Parameters within a group are relatively independent, and inter-group parameter conflicts are minimized, thus forming a group scheduling structure with clear hierarchies and non-interfering tasks within the time slice structure. The corrected time slice allocation values ​​are integrated with the parameter grouping information obtained from the topology coloring to form the optimized control sequence for water supply parameters.

[0068] In a specific embodiment, the process of executing step 104 may specifically include the following steps:

[0069] Extract the parameter groups marked as conflicting from the water supply parameter optimization control time series and obtain the parameter correlation coefficient matrix;

[0070] Arrange the historical observation values ​​of the conflict parameter groups corresponding to the parameter correlation coefficient matrix by row parameters and column time points to construct the original observation data matrix;

[0071] Independent component analysis is performed on the original observation data matrix to obtain mutually independent control variables, and a weighted objective function including control delay term and energy consumption term is established based on each independent control variable;

[0072] Perform gradient descent optimization calculation on each weighted objective function to obtain the optimized value of the independent control variable;

[0073] The optimized independent control variable values ​​are converted back into the parameter space to generate multi-parameter parallel control instructions for water supply equipment that can be executed in parallel within the same time slice.

[0074] Specifically, parameter groups that still exhibit resource contention or control interference during actual operation are identified from the optimized control sequence. These parameters are core competitors in the resource scheduling structure, have high request frequencies, similar priorities, and certain functional dependencies between control objectives. These parameters are clustered using a conflict probability marking mechanism, and conflicting parameter groups are extracted as targets for subsequent parallel decomposition. After identifying conflicting parameter groups, the degree of linear correlation between them is quantified to determine whether the mathematical basis for independence decomposition exists. A correlation coefficient matrix is ​​constructed based on multiple runs of this parameter group during historical control cycles. Each element in this matrix represents the degree of consistency between the changes between two parameters during operation. The value ranges from -1 to 1, with positive values ​​indicating positive correlation and negative values ​​indicating negative correlation, and larger absolute values ​​indicating stronger correlation. High correlation between conflicting parameter groups indicates information overlap, duplication of control logic, or intersecting resource allocation paths during scheduling execution, making it impossible to completely decouple the conflicts by directly modifying time slices. Therefore, a spatial transformation is performed on highly correlated parameters, re-expressing the multi-dimensionally coupled parameters in the original parameter space as a set of statistically independent control variables to fundamentally improve parallel scheduling capabilities. To achieve this transformation, the original historical observation data is rearranged to construct a standardized observation data matrix. The historical observation values ​​of each parameter in the conflicting parameter group are arranged in rows, with each row corresponding to a parameter and each column representing a different sampling time point. This forms a two-dimensional observation matrix with parameters as rows and time points as columns. In this matrix, each data element represents the sampled value of a control parameter at a specific time point. These values ​​constitute the dynamic observation trajectory of the parameter's behavior during the system's past operation. This matrix is ​​then processed using an independent component analysis algorithm, which extracts statistically independent latent factors hidden behind the observation data from multivariate statistical data. This algorithm decomposes the original parameter matrix into two sub-matrices: a mixing matrix that describes how the independent variables are linearly combined to generate the original parameters, and a matrix of independent variables, representing the decoupled control factors. Based on each independent control variable, a weighted objective function containing a control delay term and an energy consumption term is established. The control delay term is used to characterize the response time offset caused by the scheduling behavior corresponding to the variable during execution, and the energy consumption term reflects the level of consumption of equipment power or communication resources during the execution of the control behavior. According to the operating scenario and priority strategy, reasonable weights are set for the two indicators to construct a weighted objective function. Gradient descent optimization calculation is performed on each weighted objective function. During the optimization process, starting from the initial estimate of the current variable, it is iteratively updated along the direction of the gradient descent of the objective function. At each step, the variable is finely adjusted according to the current gradient value, and the optimal solution of the objective function is gradually approached. The optimization process is terminated according to the set convergence condition or the maximum number of iterations.For each independent control variable, its optimized numerical expression is obtained, which represents the optimal behavior or scheduling response strategy that the control variable should take in the current control cycle. The optimized independent control variable values ​​are converted back to the parameter space to generate specific executable parallel control instructions. An inverse mapping operation is performed through the mixing matrix recorded in the independent component analysis, and each optimized independent variable is reconverted into the individual parameter values ​​in the original parameter set according to its weight in the linear combination, and the specific control quantity of each original control parameter in the current scheduling cycle is restored. The reconstructed control parameters are integrated into the same time slice to generate multi-parameter parallel control instructions for water supply equipment with a reasonable structure, independent instructions, and non-interference in resources.

[0075] In a specific embodiment, the process of executing step 105 may specifically include the following steps:

[0076] Define a system state space that includes the current active parameter set, parameter priority, time slice allocation, device energy status, and communication load. Construct an action space that includes resource allocation ratio adjustment, communication channel selection, and temporary modification of parameter priority to obtain a system state-action mapping table.

[0077] Calculate the ratio of the current energy value of each water supply device to the full power as the energy factor, and correct the basic calculation load according to the energy factor to obtain the equipment calculation task allocation;

[0078] A reward function is constructed based on the average response delay, total energy consumption, and control accuracy of the water supply equipment, and a comprehensive performance evaluation index is calculated using the reward function.

[0079] Based on the system state-action mapping table, the equipment computing task allocation and the comprehensive performance evaluation index, the Q-learning reinforcement learning algorithm is executed to obtain the optimal control strategy;

[0080] Allocate execution equipment and communication channels for each control task in the multi-parameter parallel control instructions of the water supply equipment according to the optimal control strategy to obtain a task allocation plan;

[0081] The task allocation scheme is integrated with the multi-parameter parallel control instructions of the water supply equipment to generate an energy-saving control scheme for the water supply equipment.

[0082] Specifically, a system state space is defined to describe the current operating state of the water supply control system. This includes the set of currently active control parameters—the indexed set of parameters participating in the control task during the current scheduling cycle; the control priority values ​​of the parameters, which rank their importance within the control target sequence; the time slice allocation corresponding to the parameters, which represents the proportion of computing resources they receive during the control cycle; the current energy state of each control device in the water supply system—the remaining power or battery capacity; and the network load level of the communication module during the current scheduling cycle, which reflects the bandwidth and latency pressures on the system's information transmission. These state quantities collectively form a high-dimensional state vector, with each state combination representing the specific control and resource allocation environment of the water supply system at a given moment. An action space is then constructed, encompassing all available resource scheduling operations. In this control system, actions include adjusting control resource allocation ratios—dynamically increasing or decreasing parameter time-slice quotas within a given cycle based on current parameter priorities and scheduling requirements; selecting communication channels—selecting the optimal transmission path for data exchange between different devices based on network load and communication energy consumption; and temporarily modifying control parameter priorities. In specific operating scenarios, based on abnormal conditions or policy feedback, the priority weights of certain parameters are increased or decreased to influence their scheduling order. These actions constitute the set of operations performed by the system in each state. The Cartesian product of the action space and the state space forms a state-action mapping table. This mapping table is used in the reinforcement learning process to record the expected benefits of each action under the current policy in a specific state, thereby guiding the agent to learn the optimal control path. After completing the state-action modeling, an energy consumption modeling mechanism is introduced to dynamically adjust the load capacity of the devices. The energy factor is calculated by comparing the current energy state of each water supply device to its rated full power state, representing the remaining operating capacity of the device. Based on the energy factor, the device's base computational load in task allocation is adjusted. Devices with more energy are assigned a higher load weight, while devices with less energy are assigned a lower priority in the scheduling process. This extends the system's overall operating time and prevents certain devices from being interrupted due to power depletion. Furthermore, to guide the convergence of the reinforcement learning algorithm, a reward function is designed to evaluate the performance of scheduling strategies. This function consists of three core metrics: average response latency, which measures the average time from initiation to completion of a control task and reflects the system's real-time responsiveness; total energy consumption, which encompasses both computational and communication energy consumption during device operation and is a key objective for system energy conservation; and control accuracy, which measures the error between the control result and the target parameter, as a measure of the reliability of the control effect. Weighting coefficients for these three metrics are set based on the operating environment's requirements, combining them into a unified, comprehensive performance evaluation metric. This metric is calculated in real time after each state-action pair is executed and used to update the Q-value table, serving as the core feedback source for the Q-learning process.A Q-learning reinforcement learning algorithm is used to train the policy based on the system state-action mapping table, the device-calculated task allocation, and comprehensive performance evaluation metrics. During each control cycle, the algorithm selects the action with the highest expected reward from the Q-value table based on the current state vector, or performs random exploration with a certain probability using the ε-greedy strategy. After the action is executed, the actual reward is calculated based on the control results, and the Q-value of the corresponding state-action pair is iteratively updated according to the Q-learning update formula. By repeatedly selecting action, obtaining feedback, and updating Q-values, the system gradually learns the optimal action combination for each state, forming an optimal control policy set. When the Q-value table converges stably, the system extracts the policy output as the actual execution basis for the current scheduling cycle. During the actual scheduling execution, the optimal policy output is used to assign each control task in the multi-parameter parallel control instruction. The device that best suits its energy consumption level and scheduling capability is selected to perform the corresponding task. The optimal communication path is selected based on the current communication network load and channel energy efficiency for data exchange. The task device assignment information is integrated with the original multi-parameter control instruction to form a task allocation plan. Integrate the task allocation plan with control instructions to build an energy-saving control plan for water supply equipment.

[0083] In a specific embodiment, the execution step executes a Q-learning reinforcement learning algorithm based on the system state-action mapping table, the device computing task allocation, and the comprehensive performance evaluation index to obtain the optimal control strategy, which may specifically include the following steps:

[0084] Discretize the system state space and action space to obtain the initial Q value table;

[0085] Calculate the instant reward value based on the multi-parameter parallel control instructions and comprehensive performance evaluation indicators of the water supply equipment to obtain the reward value of the current state-action pair;

[0086] According to the system state-action mapping table, switch between greedy selection of the maximum Q value action and random exploration action to obtain the action selection strategy;

[0087] Based on the action selection strategy, the selected action is executed and the Q-value update amount is calculated. The Q-value update amount is iteratively updated, and the reinforcement learning process of action selection, observation feedback and Q-value update is repeated to obtain the converged Q-value distribution.

[0088] Based on the converged Q-value distribution, the corresponding maximum Q-value action in each system state is extracted and the optimal control strategy is output.

[0089] Specifically, the system state space and action space are discretized to construct a finite set of state-action pairs suitable for Q-value learning. The system state space consists of multiple continuous variables, such as control parameter priority, device energy status, communication network load, and time-slice resource allocation ratio. These continuous variables are discretized into discrete levels, such as low, medium, and high priority; full, half, and low energy status; and light, medium, and heavy communication load. This method compresses the state space into a set of finite state vectors, each representing a typical scheduling scenario for the system. Similarly, the resource scheduling operations involved in the action space contain multiple variables, such as the direction and magnitude of resource allocation ratio adjustments, communication channel selection, and fine-tuning of control parameter priorities. The parameter range for each action is divided, and the communication channels are numbered to limit the range of available options, forming a discrete set of actions. After discretization of the state and action spaces, an initial Q-value table, or Q matrix, is constructed. The number of rows in this Q matrix equals the total number of states in the discretized state space, and the number of columns equals the number of available actions. Each cell in the matrix represents the expected reward of executing a specific action in a specific state. Initially, all Q-values ​​are set to zero or a small constant, indicating that the system has not yet learned the effects of each state-action pair. As training progresses, the system continuously adjusts the Q-values, gradually approximating the long-term reward of the optimal strategy. When the system receives a multi-parameter parallel control instruction for a water supply device, it instantly evaluates the execution performance within the current control cycle. A comprehensive performance evaluation metric is used, consisting of three components: the average delay time of the control response, which measures the real-time execution of the instruction; the total energy consumption during the control process, including device computing and communication energy consumption; and control accuracy, which is the error between the control result and the desired target. These three indicators are weighted according to pre-defined weights to create a composite immediate reward value. A higher reward value indicates a better control effect for the current state-action pair, and the system is more inclined to repeat the action in similar states; conversely, it is more inclined to try other actions. The calculation of immediate rewards provides quantitative feedback for each action decision, enabling experience-based, step-by-step optimization. During reinforcement learning, an ε-greedy strategy is used for action selection. This means that for each state judgment, the system does not always select the action with the highest current Q-value. Instead, it executes a random action with a certain probability. A value of ε is set, representing the exploration probability. In each decision, the action with the highest Q-value is selected with a probability of 1-ε, and other actions are randomly selected with a probability of ε. This strategy effectively avoids the risk of falling into local optima, allowing the system to discover potentially better scheduling paths through continuous trial and error. As training progresses, the system gradually reduces the value of ε, shifting from an exploration-oriented approach to an exploitation-oriented approach, allowing the control strategy to stabilize and converge to the optimal solution.After selecting an action, the system executes the action—allocating resources, adjusting priorities, or switching communication paths within the current control state—and observes feedback from the control system. Rewards are recalculated based on this feedback, and the Q-value for the current state-action pair is corrected according to the Q-learning update formula. The Q-value update considers a weighted average between the current immediate reward and the maximum potential future reward, allowing the Q-value to gradually approach the true reward after each decision. The system uses a learning rate to determine the magnitude of Q-value updates, and a discount factor to weight future rewards. These two parameters jointly influence the speed and stability of Q-table convergence. By repeatedly repeating the process of state identification, action selection, execution feedback, and Q-value updates, the system gradually accumulates experience and develops optimal response strategies for different control scenarios. After multiple rounds of iterative training, the system's Q-value table converges, indicating that the system has mastered the optimal action for each state. At this point, the system directly extracts from the Q-table the action corresponding to the maximum Q-value for each state, which is the optimal control policy for that state. The set of optimal actions for all states together constitutes a comprehensive policy mapping table.

[0090] The above describes the intelligent decision-making method based on the integrated industrial PLC controller of the computing network in the embodiment of the present invention. The following describes the intelligent decision-making system based on the integrated industrial PLC controller of the computing network in the embodiment of the present invention. Figure 2 In one embodiment of the present invention, an intelligent decision-making system based on an integrated computing and network industrial PLC controller includes:

[0091] The first allocation module 201 is used to allocate time slices and divide the PLC controller decision cycle of the control parameters of the water supply equipment to obtain an initial control parameter queue;

[0092] An analysis module 202 is used to determine the probability distribution of normal operation state, efficient control state and decision conflict state according to the initial control parameter queue, and obtain a water supply parameter conflict probability table;

[0093] The second allocation module 203 is used to reallocate the parameter control time resources based on the water supply parameter conflict probability table to obtain the water supply parameter optimization control time sequence;

[0094] The optimization calculation module 204 is used to convert the parameter groups that still have conflicts in the water supply parameter optimization control sequence into independent control variables, and perform optimization calculations on each independent control variable to obtain multi-parameter parallel control instructions for the water supply equipment;

[0095] The resource scheduling module 205 is used to perform dynamic resource scheduling on the multi-parameter parallel control instructions of the water supply equipment and generate an energy-saving control plan for the water supply equipment.

[0096] Through the collaborative cooperation of the above components, accurate time slice allocation is performed through probabilistic modeling based on Bayesian and Poisson distribution models, which effectively optimizes the dynamic allocation of control resources, significantly shortens the response time of control parameters and improves the overall system throughput; FastICA independent component analysis technology is used to convert conflicting parameters into independent control variables, realizing parallel optimization processing of multiple conflicting parameters in the same time slice, significantly reducing the parameter processing conflict rate and improving the resource utilization efficiency of the PLC controller; combined with the dynamic resource scheduling mechanism of the energy state of the water supply equipment, the Q-learning reinforcement learning algorithm is used to achieve multi-objective optimization of control performance and energy consumption. , while ensuring control accuracy, the total energy consumption of the system is reduced, and the service life of the water supply equipment is extended; the time resource reallocation strategy based on parameter correlation analysis and conflict sensitivity index enhances the system's adaptability to parameter request fluctuations, and can maintain stable operation even when the parameter request frequency suddenly increases; through the task allocation mechanism corrected by the energy factor, computing load balancing considering the remaining energy state of the equipment is achieved, effectively avoiding the problem of single-point device resource overload; the parameter grouping strategy based on the graph coloring algorithm and the multi-parameter parallel control architecture increase the number of water supply equipment that can be managed simultaneously by a single computing-network integrated PLC controller, significantly enhancing the system's scalability.

[0097] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0098] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0099] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent decision-making method based on an integrated computing and network industrial PLC controller, characterized in that: include: The control parameters of the water supply equipment are allocated in time slices and divided into decision cycles by the PLC controller to obtain the initial control parameter queue; Determine the probability distribution of normal operation state, efficient control state and decision conflict state according to the initial control parameter queue, and obtain a water supply parameter conflict probability table; specifically include: according to the initial control parameter queue, extract a historical data set including normal operation state, efficient control state and decision conflict state from the historical database of the industrial PLC controller with integrated computing and network; perform conditional probability distribution calculation on the historical data set to obtain a parameter state conditional probability value; according to the parameter request frequency in each time slice, calculate the probability of occurrence of different request numbers per unit time, and obtain the parameter request frequency probability distribution; based on the parameter state conditional probability value and the parameter request frequency probability distribution, calculate the probability value of decision conflict in each time slice, and obtain a conflict time slice sequence; construct a parameter conflict probability matrix, calculate the probability value of decision conflict for each pair of parameters in the same time slice, and obtain the parameter conflict relationship within the time slice; calculate the conflict sensitivity value of each parameter in the conflict time slice sequence to obtain a parameter conflict sensitivity sequence, and merge the parameter conflict sensitivity sequence and the parameter conflict relationship within the time slice into a water supply parameter conflict probability table; Reallocating parameter control time resources based on the water supply parameter conflict probability table to obtain a water supply parameter optimization control sequence; Converting the parameter groups that still have conflicts in the water supply parameter optimization control sequence into independent control variables, and performing optimization calculations on each independent control variable to obtain multi-parameter parallel control instructions for the water supply equipment; Dynamic resource scheduling is performed on the multi-parameter parallel control instructions of the water supply equipment to generate an energy-saving control plan for the water supply equipment.

2. The intelligent decision-making method based on the industrial PLC controller integrated with computing and network according to claim 1 is characterized in that: The control parameters of the water supply equipment are allocated in time slices and divided into decision cycles by the PLC controller to obtain an initial control parameter queue, including: Collecting control parameters in the water supply system to obtain an initial water supply parameter set, and assigning a parameter priority value, a maximum allowable delay time value, and a computing resource occupancy value to each parameter in the initial water supply parameter set to obtain a parameter attribute matrix; Calculating the initial time slice allocation amount of each parameter according to the parameter attribute matrix to obtain the parameter time slice allocation result; Sorting the initial water supply parameter set according to the parameter priority values ​​to obtain an initial control instruction set; Establish a system status matrix containing the operating parameter values, energy status values ​​and communication load values ​​of each water supply equipment, and collect and update the values ​​of each matrix element in real time through the industrial field bus to obtain the equipment operating status information; The parameter time slice allocation result, the initial control instruction set and the device operation status information are stored in a shared memory area of ​​a PLC controller to obtain an initial control parameter queue.

3. The intelligent decision-making method based on the industrial PLC controller integrated with computing and network according to claim 1 is characterized in that: The method of calculating the occurrence probability of different numbers of requests per unit time based on the parameter request frequency in each time slice to obtain the parameter request frequency probability distribution includes: Counting the number of requests for each control parameter in different time slices from the historical data set to obtain the parameter request frequency; Grouping the parameter request frequencies to obtain a request frequency change pattern; Determine the Poisson distribution parameter λ value of each parameter in different operation scenarios based on the request frequency change pattern, and substitute the Poisson distribution parameter λ value into the Poisson distribution formula to calculate the probability distribution of k requests arriving at the same time, to obtain a parameter request quantity probability matrix; Mark the time slices and parameter combinations whose probability values ​​in the parameter request quantity probability matrix exceed a set threshold, and obtain a list of time slices with high conflict risks; Calculating a time slice resource competition distribution map using the high conflict risk time slice list and parameter priority values; The time slice resource contention distribution graph is combined with the parameter request quantity probability matrix to output a parameter request frequency probability distribution.

4. The intelligent decision-making method based on the industrial PLC controller integrated with computing and network according to claim 1 is characterized in that: The reallocation of parameter control time resources based on the water supply parameter conflict probability table to obtain a water supply parameter optimization control sequence includes: Extracting a parameter conflict matrix and a parameter priority value according to the water supply parameter conflict probability table to obtain a parameter conflict sensitivity index set; Calculating a deviation rate from an average conflict sensitivity index based on the parameter conflict sensitivity index set, and multiplying the deviation rate by a preset adjustment strength factor to obtain a time slice adjustment coefficient; Correcting the initially allocated time slice value according to the time slice adjustment coefficient to obtain an adjusted time slice allocation value; A parameter relationship topology diagram is constructed according to the water supply parameter conflict probability table, and parameters with conflict probabilities higher than a preset threshold are forcibly allocated to different time slice groups according to the parameter relationship topology diagram to obtain a water supply parameter optimization control sequence.

5. The intelligent decision-making method based on the integrated computing and network industrial PLC controller according to claim 1 is characterized in that: The method of converting the parameter groups that still have conflicts in the water supply parameter optimization control sequence into independent control variables, and performing optimization calculations on each independent control variable to obtain multi-parameter parallel control instructions for the water supply equipment includes: Extracting the parameter groups marked as conflicting from the water supply parameter optimization control time series to obtain a parameter correlation coefficient matrix; Arrange the historical observation values ​​of the conflict parameter group corresponding to the parameter correlation coefficient matrix by row parameters and column time points to construct an original observation data matrix; Performing independent component analysis on the original observation data matrix to obtain mutually independent control variables, and establishing weighted objective functions including control delay terms and energy consumption terms based on each independent control variable; Performing a gradient descent optimization calculation on each of the weighted objective functions to obtain an optimized independent control variable value; The optimized independent control variable values ​​are converted back into the parameter space to generate multi-parameter parallel control instructions for the water supply equipment that can be executed in parallel within the same time slice.

6. The intelligent decision-making method based on the industrial PLC controller integrated with computing and network according to claim 1 is characterized in that: The dynamic resource scheduling of the multi-parameter parallel control instructions of the water supply equipment to generate an energy-saving control plan for the water supply equipment includes: Define a system state space that includes the current active parameter set, parameter priority, time slice allocation, device energy status, and communication load. Construct an action space that includes resource allocation ratio adjustment, communication channel selection, and temporary modification of parameter priority to obtain a system state-action mapping table. Calculate the ratio of the current energy value of each water supply device to the full power as the energy factor, and correct the basic calculation load according to the energy factor to obtain the device calculation task allocation; Constructing a reward function based on the average response delay, total energy consumption, and control accuracy of the water supply equipment, and calculating a comprehensive performance evaluation index using the reward function; Executing a Q-learning reinforcement learning algorithm based on the system state-action mapping table, the device computing task allocation, and the comprehensive performance evaluation index to obtain an optimal control strategy; Allocate execution equipment and communication channels for each control task in the multi-parameter parallel control instruction of the water supply equipment according to the optimal control strategy to obtain a task allocation plan; The task allocation plan is integrated with the multi-parameter parallel control instructions of the water supply equipment to generate an energy-saving control plan for the water supply equipment.

7. The intelligent decision-making method based on the integrated computing and network industrial PLC controller according to claim 6 is characterized in that: The performing of a Q-learning reinforcement learning algorithm based on the system state-action mapping table, the device computing task allocation, and the comprehensive performance evaluation index to obtain an optimal control strategy includes: Discretizing the system state space and the action space to obtain an initial Q-value table; Calculate the instant reward value according to the multi-parameter parallel control instruction of the water supply equipment and the comprehensive performance evaluation index to obtain the reward value of the current state-action pair; According to the system state-action mapping table, switching between greedy selection of the maximum Q value action and random exploration action to obtain an action selection strategy; Executing the selected action based on the action selection strategy and calculating the Q-value update amount, iteratively updating the Q-value update amount, and repeating the reinforcement learning process of action selection, observation feedback, and Q-value update to obtain a converged Q-value distribution; The maximum Q-value action corresponding to each system state is extracted based on the converged Q-value distribution, and the optimal control strategy is output.

8. An intelligent decision-making system based on an integrated computing and network industrial PLC controller, characterized in that: The method for executing the intelligent decision-making method based on the integrated computing and network industrial PLC controller according to any one of claims 1 to 7, wherein the intelligent decision-making system based on the integrated computing and network industrial PLC controller comprises: The first allocation module is used to allocate time slices and divide the PLC controller decision cycle of the control parameters of the water supply equipment to obtain an initial control parameter queue; An analysis module is used to determine the probability distribution of normal operation state, efficient control state and decision conflict state according to the initial control parameter queue, and obtain a water supply parameter conflict probability table; A second allocation module is configured to reallocate parameter control time resources based on the water supply parameter conflict probability table to obtain a water supply parameter optimization control sequence; An optimization calculation module is used to convert the parameter groups that still have conflicts in the water supply parameter optimization control sequence into independent control variables, and perform optimization calculations on each independent control variable to obtain multi-parameter parallel control instructions for the water supply equipment; The resource scheduling module is used to dynamically schedule the multi-parameter parallel control instructions of the water supply equipment and generate an energy-saving control plan for the water supply equipment.

Images