Intelligent dispatching control method, device and equipment for water supply pump station and medium

Abstract

The invention relates to the technical field of water supply dispatching. The intelligent scheduling control method, device and equipment for the water supply pump station and the medium are provided, and the method comprises the steps that multi-dimensional operation data of a water supply pipe network is collected and processed in real time, and a dynamic sensing data set is generated; based on the dynamic perception data set, scheduling decision processing is conducted through a double-path inference engine, decision output is generated, and the decision output comprises a water pump control instruction and an explainable basis of the water pump control instruction; performing dynamic credibility evaluation processing on the decision output to generate an instruction credibility score; performing instruction execution decision processing based on a comparison result of the instruction credibility score and a preset threshold, and generating an execution strategy instruction; and based on control effect feedback data of the execution strategy instruction, performing iterative optimization processing on the dual-path inference engine through a man-machine cooperation verification closed loop, so as to achieve the technical effects of improving the transparency of a decision process, optimizing the risk dynamic adaptive capacity and enhancing the man-machine cooperation evolution efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of water supply scheduling, and in particular to an intelligent scheduling control method, device, equipment and medium for a water supply pump station. Background Art

[0002] With the in-depth application of artificial intelligence technology in the field of industrial control, the intelligent dispatching system of water supply pumping stations has gradually replaced traditional manual decision-making and become the core means to ensure the safety of urban water supply.

[0003] However, deep learning models, due to the unexplainable nature of their underlying logic, can lead to uncontrollable high-risk instructions in water supply scheduling decisions. Static risk interception mechanisms, lacking the ability to adapt to operating conditions, create an imbalance between safety and operational stability. The knowledge transfer barrier between human experience and machine decision-making hinders the system's iterative optimization capabilities. Summary of the Invention

[0004] Based on this, it is necessary to provide intelligent scheduling and control methods, devices, equipment and media for water supply pump stations to address the above technical problems, so as to achieve the technical effects of improving the transparency of the decision-making process, optimizing the dynamic adaptability of risks and enhancing the efficiency of human-machine collaborative evolution.

[0005] In a first aspect, the present application provides a method for intelligent dispatching and controlling a water supply pump station, the method comprising:

[0006] Collect and process multi-dimensional operation data of the water supply network in real time to generate dynamic perception data sets;

[0007] Based on the dynamic perception data set, the dual-path reasoning engine is used to process scheduling decisions and generate decision outputs, which include pump control instructions and their interpretable basis.

[0008] Perform dynamic credibility assessment on decision outputs to generate instruction credibility scores;

[0009] Based on the comparison result of the instruction credibility score and the preset threshold, the instruction execution decision processing is carried out to generate the execution strategy instruction;

[0010] Based on the control effect feedback data of the execution strategy instructions, the dual-path reasoning engine is iteratively optimized through a human-machine collaborative verification closed loop.

[0011] Furthermore, based on the dynamic perception data set, a dual-path inference engine is used to process scheduling decisions and generate decision outputs, including:

[0012] Perform deep feature extraction on the dynamic perception dataset through the main decision path to generate the pump operation status feature vector;

[0013] Perform control strategy deduction based on the pump operation state feature vector to generate pump control instructions;

[0014] Deconstruct the decision features of the dynamic perception dataset through the interpretation path to generate the key decision feature set;

[0015] Conduct rule mapping based on key decision feature sets to generate interpretable evidence;

[0016] The pump control instructions and interpretable basis are collaboratively packaged to generate decision output.

[0017] Furthermore, a control strategy is deduced based on the pump operating state feature vector to generate a pump control instruction, including:

[0018] Perform multi-objective collaborative analysis on the pump operating state feature vector to generate a set of candidate control strategies;

[0019] Perform dynamic constraint verification on the candidate control strategy set based on the pre-stored device constraint condition set to generate a feasible control strategy set;

[0020] Through the real-time performance evaluation mechanism, the feasible control strategy set is ranked by comprehensive performance to generate the target control strategy;

[0021] The target control strategy is subjected to instruction conversion and encoding processing to generate water pump control instructions.

[0022] Furthermore, the decision feature deconstruction process of the dynamic perception dataset is performed through the interpretation path to generate a key decision feature set, including:

[0023] Perform causal correlation analysis on dynamic perception data sets to generate feature causal correlation graphs;

[0024] Extract decision-making key paths based on the characteristic causal relationship graph and generate a set of decision-making influencing factors;

[0025] Through feature orthogonal decoupling, the decision-making factor set is reduced in dimension to generate the core decision feature set;

[0026] The core decision feature set is semantically structured and encapsulated to generate the key decision feature set.

[0027] Furthermore, rule mapping is performed based on the key decision feature set to generate interpretable evidence, including:

[0028] Perform semantic unit parsing on the key decision feature set to generate a semantic feature tuple set;

[0029] Based on the pre-stored engineering rule knowledge base, the semantic feature tuple set is matched and a matching rule sequence is generated;

[0030] Through the dynamic situation adaptation mechanism, the matching rule sequence is logically reconstructed to generate a causal logic chain;

[0031] Perform readable semantic compilation on the causal logic chain to generate interpretable evidence.

[0032] Furthermore, based on the control effect feedback data of the execution strategy instructions, the dual-path reasoning engine is iteratively optimized through a human-machine collaborative verification closed loop, including:

[0033] Conduct multi-dimensional performance analysis on the control effect feedback data of the execution strategy instructions and generate a control effectiveness evaluation report;

[0034] In response to the manual correction instruction, performing feature inversion processing on the manual correction instruction to generate a manual correction feature set;

[0035] Compare the control effectiveness evaluation report with the manually corrected feature set to generate a model optimization parameter set;

[0036] Based on the model optimization parameter set, the mapping relationship of the interpretation path of the dual-path reasoning engine is updated to generate an optimized dual-path reasoning engine.

[0037] Furthermore, the control effectiveness evaluation report and the manually corrected feature set are compared and processed to generate a model optimization parameter set, including:

[0038] Extract and process the feature weight distribution of the control effectiveness evaluation report to generate the model decision feature weight set;

[0039] Perform manual decision logic deconstruction on the manually corrected feature set to generate a manual decision feature weight set;

[0040] The three-dimensional difference quantification model is used to compare the model decision feature weight set with the manual decision feature weight set to generate feature weight difference, decision logic difference and risk response difference.

[0041] Based on the preset parameter generation rule library, the feature weight difference, decision logic difference and risk response difference are optimized and synthesized to generate a model optimization parameter set.

[0042] In a second aspect, the present application also provides an intelligent dispatching control device for a water supply pump station, the device comprising:

[0043] The data perception module is used to collect and process the multi-dimensional operation data of the water supply network in real time and generate a dynamic perception data set;

[0044] The dual-path decision module is used to make scheduling decisions based on the dynamic perception data set through the dual-path reasoning engine and generate decision outputs. The decision outputs include water pump control instructions and their interpretable basis.

[0045] The trustworthy evaluation module is used to perform dynamic trustworthiness evaluation on the decision output and generate the instruction trustworthiness score;

[0046] An execution decision module is used to make an instruction execution decision based on the comparison result of the instruction credibility score and the preset threshold, and generate an execution strategy instruction;

[0047] The closed-loop optimization module is used to iteratively optimize the dual-path reasoning engine based on the control effect feedback data of the execution strategy instructions through human-machine collaborative verification closed loop.

[0048] In a third aspect, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of any method in the first aspect of the present application are implemented.

[0049] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of any method in the first aspect of the present application when the computer program is executed by a processor.

[0050] The present application provides an intelligent scheduling and control method, device, equipment and medium for a water supply pump station. The method includes: real-time collection and processing of multi-dimensional operating data of a water supply network to generate a dynamic perception data set; based on the dynamic perception data set, scheduling decision processing is performed through a dual-path reasoning engine to generate a decision output, and the decision output includes a water pump control instruction and its interpretable basis; dynamic credibility evaluation processing is performed on the decision output to generate an instruction credibility score; instruction execution decision processing is performed based on the comparison result of the instruction credibility score and a preset threshold to generate an execution strategy instruction; based on the control effect feedback data of the execution strategy instruction, the dual-path reasoning engine is iteratively optimized through a human-computer collaborative verification closed loop to achieve the technical effects of improving the transparency of the decision-making process, optimizing the dynamic adaptability of risks and enhancing the effectiveness of human-computer collaborative evolution. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0052] Figure 1Flowchart of an intelligent dispatching control method for a water supply pump station in one embodiment of the present invention;

[0053] Figure 2 A flowchart of performing decision feature deconstruction processing on a dynamic perception data set through an interpretation path to generate a key decision feature set in one embodiment of the present invention;

[0054] Figure 3 This is a structural diagram of an intelligent dispatching control device for a water supply pump station in one embodiment of the present invention. DETAILED DESCRIPTION

[0055] In order to make the above-mentioned purposes, features and advantages of the present application more clearly understood, the specific implementation methods of the present application are described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to fully understand the present application. However, the present application can be implemented in many other ways than those described herein, and those skilled in the art can make similar improvements without violating the connotation of the application. Therefore, the present application is not limited to the specific embodiments disclosed below.

[0056] like Figure 1 As shown, the present application provides a water supply pump station intelligent scheduling control method, the method comprising:

[0057] S101: Collect and process multi-dimensional operation data of the water supply network in real time to generate a dynamic perception data set.

[0058] Specifically, various sensors deployed throughout the water supply network, such as pressure sensors, flow sensors, and water quality sensors, collect real-time operational data. These sensors, located at key nodes and locations, provide real-time awareness of the network's operational status. The collected data is then transmitted to a data processing system for preprocessing, including noise removal, missing value filling, and data standardization, to ensure accuracy and usability.

[0059] The preprocessed data is integrated, correlating and fusing data from different sensors and dimensions to construct a comprehensive dataset containing multi-dimensional information such as pressure, flow, and water quality. Using data mining and feature extraction techniques, key features that reflect the real-time operating status of the water supply network are extracted from the comprehensive dataset, generating a dynamic perception dataset that provides a real-time, more accurate data foundation for subsequent intelligent scheduling and control of water supply pumping stations.

[0060] S102: Based on the dynamic perception data set, a dual-path reasoning engine is used to perform scheduling decision processing and generate a decision output. The decision output includes a water pump control instruction and an interpretable basis thereof.

[0061] Specifically, the dynamic perception dataset is fed into a dual-path inference engine, consisting of a main decision path and an interpretation path. The main decision path performs deep feature extraction on the dynamic perception dataset to generate a pump operating state feature vector. This feature vector is then used to deduce the control strategy and generate pump control instructions. Simultaneously, the interpretation path deconstructs the decision features of the dynamic perception dataset to generate a key decision feature set. This key decision feature set is then matched with rules in a pre-stored engineering rule knowledge base through rule mapping to generate interpretable evidence. The pump control instructions and interpretable evidence are then collaboratively packaged to generate a complete decision output.

[0062] S103: Perform dynamic credibility evaluation on the decision output to generate an instruction credibility score.

[0063] Specifically, key features are extracted from the decision output, including multi-dimensional characteristics such as decision consistency, historical accuracy, and system stability. These features are then matched with a pre-defined credibility assessment model, and each feature is weighted using the model's weight coefficients. The preliminary calculation results are verified and adjusted through a multi-layered logic judgment and verification mechanism to ensure the reliability of the assessment results. The adjusted results are then used to generate an instruction credibility score.

[0064] S104: Perform instruction execution decision processing based on the comparison result of the instruction credibility score and the preset threshold value, and generate an execution policy instruction.

[0065] Specifically, a preset execution decision rule set is obtained, which includes different threshold intervals and their corresponding instruction execution strategies. The calculated instruction credibility score is compared with the preset thresholds one by one. Based on the threshold interval in which the score falls within the rule set, the corresponding instruction execution strategy is matched. For example, if the score is above the upper threshold, the raw water pump control instruction is directly executed; if the score is in the middle range, a partial execution or adjusted execution strategy is triggered; if the score is below the lower threshold, instruction execution is terminated, an early warning mechanism is triggered, and a manual intervention request is generated. The execution strategy instruction is then generated based on the matching results.

[0066] S105: Based on the control effect feedback data of the execution strategy instructions, the dual-path reasoning engine is iteratively optimized through a human-machine collaborative verification closed loop.

[0067] Specifically, operational data following the execution of strategic instructions is collected, including changes in indicators such as pressure, flow, and water quality in the water supply network, as well as the operating status and energy consumption of water pumps. This feedback data is subjected to a multi-dimensional performance analysis to generate a control effectiveness evaluation report, encompassing aspects such as task completion, resource utilization efficiency, and system stability. Furthermore, the system monitors for manual adjustments or interventions, responds to manual corrections, performs feature inversion, extracts key elements and priorities in the manual decision-making process, and generates a set of manually corrected features.

[0068] Next, the system decision features extracted from the control effectiveness evaluation report are compared with the manual decision features in the manually corrected feature set to identify differences in feature weights, decision logic, and risk response. This generates a comparison result that includes the degree of difference in feature weights, decision logic, and risk response. A rule library is generated based on preset parameters, and the comparison results are converted into a model optimization parameter set to adjust weight allocation, rule matching, and decision logic during the inference process. The model optimization parameter set is then applied to the interpretation path of the dual-path inference engine, updating its internal mapping relationships. This allows the inference engine to incorporate manual experience in subsequent decisions, improving decision accuracy and adaptability, thereby completing the iterative optimization process of the dual-path inference engine.

[0069] An embodiment of the present application provides an intelligent scheduling and control method for a water supply pump station, including: real-time collection and processing of multi-dimensional operating data of a water supply network to generate a dynamic perception data set; based on the dynamic perception data set, scheduling decision processing is performed through a dual-path reasoning engine to generate a decision output, and the decision output includes a water pump control instruction and an interpretable basis thereof; dynamic credibility evaluation processing is performed on the decision output to generate an instruction credibility score; instruction execution decision processing is performed based on a comparison result between the instruction credibility score and a preset threshold to generate an execution strategy instruction; based on control effect feedback data of the execution strategy instruction, the dual-path reasoning engine is iteratively optimized through a human-machine collaborative verification closed loop to achieve the technical effects of improving the transparency of the decision-making process, optimizing the dynamic adaptability of risks, and enhancing the effectiveness of human-machine collaborative evolution.

[0070] Furthermore, based on the dynamic perception data set, a dual-path inference engine is used to perform scheduling decision processing and generate decision outputs, including:

[0071] Perform deep feature extraction on the dynamic perception dataset through the main decision path to generate the pump operation status feature vector;

[0072] Perform control strategy deduction based on the pump operation state feature vector to generate pump control instructions;

[0073] Deconstruct the decision features of the dynamic perception dataset through the interpretation path to generate the key decision feature set;

[0074] Conduct rule mapping based on key decision feature sets to generate interpretable evidence;

[0075] The pump control instructions and interpretable basis are collaboratively packaged to generate decision output.

[0076] Specifically, the main decision path is used to perform deep feature extraction on the dynamic perception dataset. This involves analyzing the multidimensional information in the dynamic perception dataset, identifying and extracting features closely related to the pump's operating status, such as pressure change trends, flow fluctuation patterns, and equipment operating parameters. These features are then integrated to generate a multidimensional data structure that more comprehensively reflects the pump's current operating status, namely the pump's operating status feature vector. Based on this feature vector, a preset control strategy model is used to perform control strategy deduction. By simulating the impact of different control instructions on the pump's operating status, the system response after executing specific instructions is predicted, and the instruction combination that enables the water supply network to achieve the expected operating goals is selected. This generates more accurate pump control instructions that clearly include specific operating parameters such as the pump's start and stop, speed adjustment, and flow control, ensuring that the pump operates according to the predetermined strategy to maintain a stable water supply in the water supply network.

[0077] At the same time, the dynamic perception dataset is deconstructed through an interpretive path. Using causal analysis algorithms and feature association models, the system deeply explores the data's inherent logic and key influencing factors, identifying features that have a significant impact on decision-making, such as pressure changes at key nodes and fluctuations in flow demand in specific areas. These key features are systematically organized to generate a key decision feature set, providing a foundation for subsequent rule mapping. Next, rule mapping is performed based on the key decision feature set. A pre-existing engineering rule knowledge base is used to match key decision features with corresponding rules. Using a logical reasoning mechanism, the matched rules are combined and reasoned to generate interpretable evidence with causal relationships and logical coherence. This evidence is presented in natural language, detailing the decision's rationale, causes, and expected effects. For example, it explains that the decision to start or stop a water pump is due to the monitored pressure in a certain area falling below a set threshold and an increasing flow demand. To ensure stable water supply, the pump's operating status needs to be adjusted promptly. This allows operators to clearly understand the rationale and necessity of the system's decisions.

[0078] The generated water pump control instructions and interpretable basis are collaboratively packaged and processed. Through the data encapsulation mechanism, the two are integrated into an organic whole to generate a complete decision output. The decision output contains both specific control instructions and relatively detailed explanations to ensure that while executing the instructions, the operator can have a relatively comprehensive understanding of the logic and basis behind the decision, thereby enhancing the trust in the system decision and the accuracy of execution, and providing strong support for the intelligent scheduling of water supply pump stations.

[0079] Furthermore, a control strategy is deduced based on the pump operating state feature vector to generate a pump control instruction, including:

[0080] Perform multi-objective collaborative analysis on the pump operating state feature vector to generate a set of candidate control strategies;

[0081] Perform dynamic constraint verification on the candidate control strategy set based on the pre-stored device constraint condition set to generate a feasible control strategy set;

[0082] Through the real-time performance evaluation mechanism, the feasible control strategy set is ranked by comprehensive performance to generate the target control strategy;

[0083] The target control strategy is subjected to instruction conversion and encoding processing to generate water pump control instructions.

[0084] Specifically, a multi-objective collaborative analysis algorithm is used to process the characteristic vector of the water pump operating status, identify the relationships and influences between multiple key objectives such as water supply stability, energy efficiency and equipment wear, and generate a set of candidate control strategies that can meet different objective combinations through the objective trade-off model. Each strategy includes operating parameters such as water pump start and stop, speed regulation and flow distribution.

[0085] Afterwards, the pre-stored device constraint condition set is called, including the power upper limit, flow range and pressure limit of the water pump, and dynamic constraint verification processing is performed on the candidate control strategy set to screen out feasible control strategy sets that meet the equipment safety and operation constraints, ensuring that each strategy will not exceed the physical and safety limitations of the equipment during actual execution.

[0086] Through the real-time performance evaluation mechanism, the performance evaluation model is used to calculate the comprehensive performance of each strategy in the set of feasible control strategies. Based on the comprehensive benefits of the strategies in terms of water supply guarantee, energy saving and equipment life extension, they are ranked according to their performance level to generate the optimal target control strategy. This strategy can achieve the best comprehensive operating performance while satisfying all constraints as much as possible.

[0087] The target control strategy is subjected to instruction conversion and encoding processing. An instruction encoding conversion algorithm is used to convert the abstract control strategy into specific water pump control instructions that comply with the water pump control system communication protocol and format requirements, ensuring that the instructions can be received and executed by the water pump equipment. The generated water pump control instructions specify the operating parameters and operating actions of the water pump in more detail.

[0088] like Figure 2 As shown in the figure, the decision feature deconstruction process of the dynamic perception dataset is carried out through the interpretation path to generate a key decision feature set, including:

[0089] S201: Perform causal correlation analysis on the dynamic perception data set to generate a feature causal correlation graph;

[0090] S202: Extracting decision-making key paths based on the feature causal relationship graph to generate a set of decision-making influencing factors;

[0091] S203: Performing dimensionality reduction processing on the decision-influencing factor set through feature orthogonal decoupling processing to generate a core decision feature set;

[0092] S204: Perform semantic structured encapsulation processing on the core decision feature set to generate a key decision feature set.

[0093] Specifically, in step S201, a causal association analysis algorithm is used to process the dynamic perception data set, identify the causal relationship between data elements, and construct a feature causal association graph, which displays the causal influence between different features in the form of nodes and directed edges. Among them, the nodes represent features, and the directed edges represent the direction and strength of the causal relationship. In step S202, based on the feature causal association graph, a key path extraction algorithm is used to identify the paths that have a significant impact on the decision results, extract the key features on the above paths, and generate a set of decision influencing factors. This set includes important features that directly and indirectly affect the decision results, covering key factors such as pressure changes, flow fluctuations, and water pump operating status of the water supply network.

[0094] In step S203, the decision-influencing factor set is subjected to feature orthogonal decoupling processing, and the orthogonalization algorithm is used to remove the correlation and redundancy between features, reduce the dimension of the feature set, extract independent and core features, and generate a core decision feature set. This feature set retains the features that have the greatest impact on the decision, removes redundant information, and improves the efficiency and accuracy of subsequent processing. In step S204, the core decision feature set is subjected to semantic structured encapsulation processing, and the semantic encapsulation algorithm is used to associate the core features with predefined semantic labels and structures to generate a key decision feature set. This feature set is presented in a structured manner, and each feature has a clear semantic meaning and data type to facilitate subsequent rule mapping and interpretation processing, and can more clearly demonstrate the role and significance of features in the decision-making process.

[0095] Furthermore, rule mapping is performed based on the key decision feature set to generate interpretable evidence, including:

[0096] Perform semantic unit parsing on the key decision feature set to generate a semantic feature tuple set;

[0097] Based on the pre-stored engineering rule knowledge base, the semantic feature tuple set is matched and a matching rule sequence is generated;

[0098] Through the dynamic situation adaptation mechanism, the matching rule sequence is logically reconstructed to generate a causal logic chain;

[0099] Perform readable semantic compilation on the causal logic chain to generate interpretable evidence.

[0100] Specifically, the key decision feature set is parsed into semantic units, and a semantic analysis algorithm is used to identify the semantic units in the features, such as feature type, feature value range, and feature correlation, to generate a set of semantic feature tuples. Each tuple contains key semantic information such as the feature name, type, current status, and change trend.

[0101] Based on a pre-existing engineering rule knowledge base, the rule matching engine is invoked to match semantic feature tuples against the rules in the knowledge base one by one. Through pattern recognition and logical reasoning, the engine selects rules that match the current feature state and change trends, generating a matching rule sequence. This sequence contains a series of rules, each of which defines the decision logic and interpretation to be taken under specific conditions.

[0102] Through a dynamic contextual adaptation mechanism, the system analyzes the current water supply system's contextual characteristics, such as peak water supply periods, equipment maintenance periods, or sudden failures, and reconstructs the logical relationships within the matching rule sequence. Based on these contextual characteristics, the rule execution order and logical connections are adjusted to generate a causal logic chain. This ensures that the logical relationships between rules align with actual water supply operations, forming a coherent decision-making logic.

[0103] The causal logic chain is semantically compiled and processed, using a natural language generation algorithm to convert the rules and logical relationships in the logic chain into easily understandable natural language descriptions, generating interpretable evidence. This evidence is presented in text form, detailing the basis, causes, and expected effects of the decision, enabling operators to more clearly understand the rationality and necessity of the system's decisions.

[0104] Furthermore, based on the control effect feedback data of the execution strategy instructions, the dual-path reasoning engine is iteratively optimized through a human-machine collaborative verification closed loop, including:

[0105] Conduct multi-dimensional performance analysis on the control effect feedback data of the execution strategy instructions and generate a control effectiveness evaluation report;

[0106] In response to the manual correction instruction, performing feature inversion processing on the manual correction instruction to generate a manual correction feature set;

[0107] Compare the control effectiveness evaluation report with the manually corrected feature set to generate a model optimization parameter set;

[0108] Based on the model optimization parameter set, the mapping relationship of the interpretation path of the dual-path reasoning engine is updated to generate an optimized dual-path reasoning engine.

[0109] Specifically, multi-dimensional performance analysis is performed on feedback data from the control effectiveness of executing strategic instructions. This process involves the calculation and evaluation of multiple performance indicators, including water supply stability, energy efficiency, equipment wear, and water supply coverage. Through the performance evaluation model, this feedback data is converted into specific performance indicators, generating a comprehensive control effectiveness evaluation report. This report not only includes the quantitative results of each performance indicator, but also provides a comprehensive evaluation of the overall control effectiveness and improvement suggestions.

[0110] Then, in response to the manual correction instructions, feature inversion processing is performed on them. This involves using a command parsing algorithm to decompose the manual correction instructions into multiple feature elements, extracting key features of the correction action, such as the target parameter, correction amplitude, and trigger conditions for the correction, and generating a manual correction feature set. This feature set, as a digital representation of manual experience, provides an important reference for subsequent model optimization.

[0111] Compare the control effectiveness assessment report with the manually modified feature set. This involves using a feature comparison algorithm to identify discrepancies between the two, including gaps between effectiveness indicators and human expectations, inconsistencies between decision logic and human experience, and differences in risk response strategies. This generates a model optimization parameter set that specifies which aspects of the inference engine require adjustment and optimization, such as reallocation of feature weights, revision of decision rules, and updates to the risk assessment model.

[0112] Based on the model optimization parameter set, the dual-path reasoning engine's interpretation path is updated with mapping relationships. This involves applying the optimized parameters to the inference engine's interpretation path through a parameter adjustment algorithm, updating its internal feature mapping relationships, rule matching logic, and causal logic chain. This allows the inference engine to better integrate human experience, improve decision-making accuracy and adaptability, and ultimately generate an optimized dual-path reasoning engine.

[0113] Furthermore, the control effectiveness evaluation report and the manually corrected feature set are compared and processed to generate a model optimization parameter set, including:

[0114] Extract and process the feature weight distribution of the control effectiveness evaluation report to generate the model decision feature weight set;

[0115] Perform manual decision logic deconstruction on the manually corrected feature set to generate a manual decision feature weight set;

[0116] The three-dimensional difference quantification model is used to compare the model decision feature weight set with the manual decision feature weight set to generate feature weight difference, decision logic difference and risk response difference.

[0117] Based on the preset parameter generation rule library, the feature weight difference, decision logic difference and risk response difference are optimized and synthesized to generate a model optimization parameter set.

[0118] Specifically, the control effectiveness evaluation report is processed for feature weight distribution extraction. This involves using a weight analysis algorithm to identify the weight distribution of each effectiveness indicator in the report, extracting the key features that influence decision-making and their weights, and generating a model decision feature weight set. This weight set reflects the importance the system places on different features during the decision-making process.

[0119] The manually corrected feature set is then deconstructed using the human decision logic. This involves analyzing the decision logic of the manual correction instructions using a logic parsing algorithm, identifying the key features that humans focus on during the correction process and their weighting, and generating a set of human decision feature weights. This weight set reflects the degree of importance humans place on different features based on their experience.

[0120] A three-dimensional difference quantification model was used to compare the feature weights of the model's decisions with those of human decision-makers. This model compared the differences across three dimensions: feature weight difference, decision logic difference, and risk response difference. The model calculated the feature weight difference degree, reflecting the differences in feature importance assessment between the system and human decision-makers; the decision logic difference degree, reflecting the differences in logical reasoning paths during the decision-making process; and the risk response difference degree, measuring the differences in risk response strategies between the system and human decision-makers. These three difference metrics comprehensively quantify the differences between system and human decision-making.

[0121] Based on a pre-configured parameter generation rule base, optimization parameters are synthesized for feature weight variance, decision logic variance, and risk response variance. Using a parameter synthesis algorithm and pre-defined rules, these three variance metrics are converted into specific optimization parameters, generating a model optimization parameter set. This parameter set includes specific guidance for adjusting feature weights, revising decision logic, and optimizing risk response strategies, enabling subsequent updates and optimizations to the dual-path inference engine.

[0122] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0123] In one embodiment, if Figure 3 As shown, the present application also provides a water supply pump station intelligent dispatching control device 300, which includes:

[0124] The data perception module 301 is used to collect and process the multi-dimensional operation data of the water supply network in real time to generate a dynamic perception data set;

[0125] The dual-path decision module 302 is used to perform scheduling decision processing based on the dynamic perception data set through the dual-path reasoning engine to generate a decision output, which includes the water pump control instructions and their interpretable basis;

[0126] Credibility assessment module 303, used to perform dynamic credibility assessment on the decision output and generate a credibility score for the instruction;

[0127] An execution decision module 304 is configured to make an instruction execution decision based on a comparison result between the instruction credibility score and a preset threshold value, and generate an execution policy instruction;

[0128] The closed-loop optimization module 305 is used to iteratively optimize the dual-path reasoning engine through a human-machine collaborative verification closed loop based on the control effect feedback data of the execution strategy instructions.

[0129] Specifically, the data perception module 301 collects and processes multidimensional operational data from the water supply network in real time. It collects data such as pressure, flow, and water quality from various sensors deployed in the water supply network. After preprocessing to remove noise and fill in missing values, the data is integrated to generate a dynamic perception dataset. The dual-path decision module 302 uses a dual-path inference engine to process the dynamic perception dataset. The main decision path performs deep feature extraction to generate a pump operating state feature vector, which is then used to derive pump control instructions. The interpretation path deconstructs decision features to generate a key decision feature set, which is then mapped to generate interpretable evidence. Both are then packaged into a decision output.

[0130] The credibility assessment module 303 performs a credibility assessment on the decision output. This involves using relevant credibility assessment models to quantitatively analyze various aspects of the decision output. Based on factors such as the consistency, stability, and accuracy of the decision, a credibility score (i.e., the instruction credibility score) is calculated for the decision output. This score reflects the reliability and trustworthiness of the decision output, providing a basis for subsequent instruction execution decisions.

[0131] The execution decision module 304 determines whether to execute the command based on a comparison of the command credibility score with a preset threshold. The preset threshold is a standard set based on system security and stability requirements. If the command credibility score is higher than or equal to the preset threshold, the decision output is deemed reliable and the corresponding execution strategy instructions are generated to adjust and control the water supply system. If the score is lower than the preset threshold, further analysis and verification of the decision output is required to ensure the safe operation of the water supply system.

[0132] The closed-loop optimization module 305 iteratively optimizes the dual-path inference engine through a human-machine collaborative verification closed-loop. This involves collecting feedback data on control effectiveness after executing strategy instructions. This data reflects the actual results of the instructions. This feedback data is compared and analyzed with a pre-stored optimization model to evaluate the model's accuracy and adaptability. Based on the evaluation results, an optimization algorithm is used to adjust the model's parameters and optimize its structure, updating the dual-path inference engine. In this way, the inference engine can continuously learn and adapt to new operating conditions, thereby improving the accuracy and reliability of its decisions.

[0133] The dual-path decision module 302 is further configured to:

[0134] Perform deep feature extraction on the dynamic perception dataset through the main decision path to generate the pump operation status feature vector;

[0135] Perform control strategy deduction based on the pump operation state feature vector to generate pump control instructions;

[0136] Deconstruct the decision features of the dynamic perception dataset through the interpretation path to generate the key decision feature set;

[0137] Conduct rule mapping based on key decision feature sets to generate interpretable evidence;

[0138] The pump control instructions and interpretable basis are collaboratively packaged to generate decision output.

[0139] The dual-path decision module 302 is further configured to:

[0140] Perform multi-objective collaborative analysis on the pump operating state feature vector to generate a set of candidate control strategies;

[0141] Perform dynamic constraint verification on the candidate control strategy set based on the pre-stored device constraint condition set to generate a feasible control strategy set;

[0142] Through the real-time performance evaluation mechanism, the feasible control strategy set is ranked by comprehensive performance to generate the target control strategy;

[0143] The target control strategy is subjected to instruction conversion and encoding processing to generate water pump control instructions.

[0144] The dual-path decision module 302 is further configured to:

[0145] Perform causal correlation analysis on dynamic perception data sets to generate feature causal correlation graphs;

[0146] Extract decision-making key paths based on the characteristic causal relationship graph and generate a set of decision-making influencing factors;

[0147] Through feature orthogonal decoupling, the decision-making factor set is reduced in dimension to generate the core decision feature set;

[0148] The core decision feature set is semantically structured and encapsulated to generate the key decision feature set.

[0149] The dual-path decision module 302 is further configured to:

[0150] Perform semantic unit parsing on the key decision feature set to generate a semantic feature tuple set;

[0151] Based on the pre-stored engineering rule knowledge base, the semantic feature tuple set is matched and a matching rule sequence is generated;

[0152] Through the dynamic situation adaptation mechanism, the matching rule sequence is logically reconstructed to generate a causal logic chain;

[0153] Perform readable semantic compilation on the causal logic chain to generate interpretable evidence.

[0154] The closed-loop optimization module 305 is further configured to:

[0155] Conduct multi-dimensional performance analysis on the control effect feedback data of the execution strategy instructions and generate a control effectiveness evaluation report;

[0156] In response to the manual correction instruction, performing feature inversion processing on the manual correction instruction to generate a manual correction feature set;

[0157] Compare the control effectiveness evaluation report with the manually corrected feature set to generate a model optimization parameter set;

[0158] Based on the model optimization parameter set, the mapping relationship of the interpretation path of the dual-path reasoning engine is updated to generate an optimized dual-path reasoning engine.

[0159] The closed-loop optimization module 305 is further configured to:

[0160] Extract and process the feature weight distribution of the control effectiveness evaluation report to generate the model decision feature weight set;

[0161] Perform manual decision logic deconstruction on the manually corrected feature set to generate a manual decision feature weight set;

[0162] The three-dimensional difference quantification model is used to compare the model decision feature weight set with the manual decision feature weight set to generate feature weight difference, decision logic difference and risk response difference.

[0163] Based on the preset parameter generation rule library, the feature weight difference, decision logic difference and risk response difference are optimized and synthesized to generate a model optimization parameter set.

[0164] In one embodiment, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0165] In one embodiment, the present application further provides a computer-readable storage medium having a computer program stored thereon, which implements the steps in the above-mentioned method embodiments when the computer program is executed by a processor.

[0166] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments. The device embodiments described above are merely illustrative, wherein the components described as separate parts may or may not be physically separated, and the parts displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the disclosed solution. A person of ordinary skill in the art can understand and implement it without expending creative work.

[0167] The above-described embodiments merely represent several implementation methods of the embodiments of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that a person skilled in the art may make various modifications and improvements without departing from the concept of the embodiments of the present application, and these modifications and improvements fall within the scope of protection of the embodiments of the present application.

Claims

1. The intelligent dispatching and control method of a water supply pump station is characterized in that: The method comprises: Collect and process multi-dimensional operation data of the water supply network in real time to generate dynamic perception data sets; Based on the dynamic perception data set, a scheduling decision is processed by a dual-path reasoning engine to generate a decision output, wherein the decision output includes a water pump control instruction and an interpretable basis thereof; Performing dynamic credibility evaluation on the decision output to generate an instruction credibility score; Performing instruction execution decision processing based on a comparison result of the instruction credibility score and a preset threshold to generate an execution strategy instruction; Based on the control effect feedback data of the execution strategy instructions, the dual-path reasoning engine is iteratively optimized through a human-machine collaborative verification closed loop.

2. The intelligent dispatching and control method for a water supply pump station according to claim 1, characterized in that: The step of performing scheduling decision processing based on the dynamic perception data set through a dual-path reasoning engine to generate a decision output includes: Performing deep feature extraction processing on the dynamic perception data set through a main decision path to generate a water pump operation state feature vector; Performing control strategy deduction processing based on the water pump operation state characteristic vector to generate a water pump control instruction; Performing decision feature deconstruction processing on the dynamic perception data set through an interpretation path to generate a key decision feature set; Perform rule mapping processing based on the key decision feature set to generate an explainable basis; The water pump control instruction and the interpretable basis are collaboratively packaged to generate the decision output.

3. The intelligent dispatching and control method for a water supply pump station according to claim 2, characterized in that: The control strategy deduction process is performed based on the water pump operation state characteristic vector to generate a water pump control instruction, including: Performing multi-objective collaborative analysis on the pump operation state feature vector to generate a candidate control strategy set; Performing dynamic constraint verification processing on the candidate control strategy set based on a pre-stored device constraint condition set to generate a feasible control strategy set; Performing comprehensive performance ranking processing on the feasible control strategy set through a real-time performance evaluation mechanism to generate a target control strategy; The target control strategy is subjected to instruction conversion and encoding processing to generate the water pump control instruction.

4. The intelligent dispatching and control method for a water supply pump station according to claim 2, characterized in that: The process of performing decision feature deconstruction processing on the dynamic perception data set through the interpretation path to generate a key decision feature set includes: Performing causal correlation analysis on the dynamic perception data set to generate a feature causal correlation graph; Extracting decision-making key paths based on the characteristic causal relationship graph to generate a set of decision-making influencing factors; Performing dimensionality reduction processing on the decision-influencing factor set through feature orthogonal decoupling processing to generate a core decision feature set; The core decision feature set is subjected to semantic structured encapsulation processing to generate the key decision feature set.

5. The intelligent dispatching and control method for a water supply pump station according to claim 2, characterized in that: The rule mapping process is performed based on the key decision feature set to generate an interpretable basis, including: Performing semantic unit parsing processing on the key decision feature set to generate a semantic feature tuple set; Performing rule matching processing on the semantic feature tuple set based on a pre-stored engineering rule knowledge base to generate a matching rule sequence; Performing logical relationship reconstruction on the matching rule sequence through a dynamic context adaptation mechanism to generate a causal logic chain; The causal logic chain is subjected to readable semantic compilation processing to generate the interpretable basis.

6. The intelligent dispatching and control method for a water supply pump station according to claim 1, characterized in that: The control effect feedback data based on the execution strategy instruction is used to iteratively optimize the dual-path reasoning engine through a human-machine collaborative verification closed loop, including: Performing multi-dimensional performance analysis on the control effect feedback data of the execution strategy instructions to generate a control effectiveness evaluation report; In response to the manual correction instruction, performing feature inversion processing on the manual correction instruction to generate a manual correction feature set; Performing a difference comparison process on the control effectiveness evaluation report and the manually corrected feature set to generate a model optimization parameter set; Based on the model optimization parameter set, a mapping relationship update process is performed on the interpretation path of the dual-path reasoning engine to generate an optimized dual-path reasoning engine.

7. The intelligent dispatching and control method for a water supply pump station according to claim 6, characterized in that: The performing difference comparison processing on the control effectiveness evaluation report and the manually corrected feature set to generate a model optimization parameter set includes: Performing feature weight distribution extraction processing on the control effectiveness evaluation report to generate a model decision feature weight set; Performing manual decision logic deconstruction processing on the manually corrected feature set to generate a manual decision feature weight set; Performing a difference comparison process on the model decision feature weight set and the manual decision feature weight set using a three-dimensional difference quantification model to generate a feature weight difference degree, a decision logic difference degree, and a risk response difference degree; Based on a preset parameter generation rule library, the feature weight difference, the decision logic difference and the risk response difference are optimized parameter synthesis processed to generate the model optimization parameter set.

8. Intelligent dispatching control device for water supply pump station, characterized in that: The device comprises: The data perception module is used to collect and process the multi-dimensional operation data of the water supply network in real time and generate a dynamic perception data set; A dual-path decision module, configured to perform scheduling decision processing based on the dynamic perception data set through a dual-path reasoning engine to generate a decision output, wherein the decision output includes a water pump control instruction and an interpretable basis thereof; A trustworthy evaluation module, configured to perform dynamic trustworthiness evaluation on the decision output and generate a trustworthiness score for the instruction; An execution decision module, configured to make an instruction execution decision based on a comparison result between the instruction credibility score and a preset threshold value, and generate an execution policy instruction; A closed-loop optimization module is used to iteratively optimize the dual-path reasoning engine through a human-machine collaborative verification closed loop based on the control effect feedback data of the execution strategy instructions.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the water supply pump station intelligent scheduling control method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the intelligent dispatching and control method for a water supply pump station according to any one of claims 1 to 7 are implemented.

Images