A multi-objective collaborative optimization pipe network automatic delivery control method

Abstract

The application discloses a kind of multi-objective collaborative optimization's pipe network automatic delivery control method, belong to process control and system automation technical field, it includes obtaining pipe network operation data, generates robust feature by adversarial domain adaptation training, real-time monitoring environmental mutation, utilize robust feature to construct dynamic multi-objective optimization model, and adopt dynamic multi-objective evolutionary algorithm to generate Pareto solution set, finally combine collaborative balance degree index and preference function to select optimal control solution.The application adopts adversarial domain adaptation technology to eliminate data domain difference, obtain robust feature which is not sensitive to working condition change, and combine dynamic multi-objective evolutionary algorithm responding to environmental mutation, can track and determine optimal balance solution between energy consumption, pressure safety and service quality in real time when pipe network load and operating environment dynamically change, improve the real-time performance, robustness and multi-objective collaborative optimization level of pipe network automatic delivery control.

Description

Technical Field

[0001] This application relates to the field of process control and system automation, and in particular to a multi-objective collaborative optimization method for automatic pipeline transmission and distribution control. Background Technology

[0002] Automatic pipeline distribution control systems are a crucial component of modern urban infrastructure, widely used in water supply, heating, and gas supply. These systems precisely regulate pumps, valves, and other actuators within the pipeline network to effectively manage key operating parameters such as pressure and flow, ensuring safe, efficient, and economical system operation. Their core function lies in generating and executing optimal control strategies based on the real-time status of the pipeline network, falling under the technical category of regulating and controlling non-electrical variables.

[0003] In existing technologies, automatic distribution control of pipeline networks typically relies on optimization algorithms based on hydraulic models. The control system first calibrates the hydraulic model using historical operating data, then combines this with real-time acquired operating data. By setting optimization objectives, such as minimizing energy consumption or maximizing pressure stability, mathematical optimization algorithms are used to calculate the optimal setpoints for the control equipment. Some advanced methods employ multi-objective optimization techniques, simultaneously considering multiple objectives such as energy consumption and pressure, providing decision-makers with a set of compromise control solutions.

[0004] However, existing control methods have some inherent technical shortcomings. First, models trained on historical data often suffer performance degradation when faced with real-time changing operating conditions due to inconsistent data distribution, resulting in insufficient generalization ability and robustness. Second, traditional control strategies are mostly static, making it difficult to respond quickly and effectively to sudden events in the pipeline network, such as pipeline bursts or surges in user demand, and lacking adaptability to dynamic environments. Finally, when dealing with multi-objective problems, existing methods often struggle to select the single optimal solution that best meets the current operational needs and balances performance across objectives from a large number of compromise solutions, making the decision-making process subjective and somewhat arbitrary. Summary of the Invention

[0005] To address the aforementioned issues, this application provides a multi-objective collaborative optimization method for automatic pipeline transmission and distribution control. It employs a combination of adversarial domain adaptation training, dynamic multi-objective evolutionary algorithms, and collaborative balance evaluation techniques. This method can overcome data distribution differences, respond quickly to sudden environmental changes, and achieve dynamic collaborative and balanced control of multiple operational objectives.

[0006] To achieve the above objectives, this application adopts the following technical solution:

[0007] Firstly, a multi-objective collaborative optimization method for automatic pipeline distribution control is provided, including:

[0008] S1. Obtain real-time operating parameters, equipment status and external environment data of the pipeline network as real-time status data, define the real-time status data as target domain data, and obtain historical operating data, historical energy consumption data and historical load data of the pipeline network and define them as source domain data.

[0009] S2. Based on the source domain data and target domain data, by performing adversarial domain adaptation training, the difference between the source domain data and the target domain data is eliminated, robust features that can characterize the essential laws of pipeline network operation are generated, and the domain adaptation error of the training process is obtained.

[0010] S3. Monitor the rate of change and domain adaptation error of the target domain data in real time, and generate environmental mutation judgment results;

[0011] S4. Using the robust features, construct a dynamic multi-objective optimization model that includes decision variables and multiple objective functions;

[0012] S5. Based on the dynamic multi-objective optimization model, in response to the environmental mutation determination result, a Pareto solution set containing several non-dominated solutions is generated using a dynamic multi-objective evolutionary algorithm;

[0013] S6. Evaluate the objective balance of the non-dominated solutions in the Pareto solution set and calculate the synergistic balance index of each solution among multiple objective functions.

[0014] S7. Obtain the preference function used to characterize the current operational preference, select the optimal control solution from the Pareto solution set in combination with the cooperative balance index, and generate control instructions to be issued to the actuator based on the optimal control solution.

[0015] Based on the above technical solutions, in the multi-objective collaborative optimization pipeline automatic transmission and distribution control method provided in this application, the technical means of combining adversarial domain adaptation training, dynamic multi-objective evolutionary algorithm and collaborative balance evaluation are adopted. This can overcome data distribution differences, respond quickly to environmental changes, and achieve dynamic collaborative and balanced control of multiple operational objectives.

[0016] In conjunction with the first aspect above, in one possible implementation, generating robust features that characterize the essential laws of pipeline network operation includes: extracting features from the source domain data and target domain data to generate control features, and classifying the control features to determine the data source and generating a domain classification loss; training the feature extraction by alternately optimizing the main task loss of feature extraction and the domain classification loss of domain classification, so that the feature extraction can optimize the domain classification error; applying the converged feature extraction to real-time data, and using the output domain-invariant features as robust features.

[0017] In conjunction with the first aspect above, in one possible implementation, the real-time monitoring of the rate of change and domain adaptation error of the target domain data to generate an environmental mutation judgment result includes: calculating the amount of change of the target domain data within a continuous time step to generate a rate of change index; comparing the rate of change index and the domain adaptation error with a threshold representing the fluctuation range of normal operation to generate a comparison result; and generating an environmental mutation judgment result indicating that a sudden change in the environment has occurred when the rate of change index or the domain adaptation error exceeds the threshold, based on the comparison result.

[0018] In conjunction with the first aspect above, in one possible implementation, the construction of a dynamic multi-objective optimization model that includes decision variables and multiple objective functions includes: defining a set of decision variables that include operating parameters of control equipment; establishing multiple objective functions that quantify transmission and distribution energy consumption costs, pipeline pressure safety, and service quality tracking respectively; and integrating the robust features as dynamic input parameters into the multiple objective functions to adjust the parameters and model constraints of the multiple objective functions in real time, thereby forming a dynamic multi-objective optimization model.

[0019] In conjunction with the first aspect above, in one possible implementation, based on the dynamic multi-objective optimization model and in response to the environmental mutation determination result, generating a Pareto solution set containing several non-dominated solutions using a dynamic multi-objective evolutionary algorithm includes: generating an initial population using the Pareto solution set from the previous control period or prediction results based on robust features; if the environmental mutation determination result is yes, adjusting the search parameters of the evolutionary operation, performing an evolutionary operation on the initial population to generate a offspring population; evaluating the initial population and offspring population using the dynamic multi-objective optimization model, and generating a Pareto solution set through non-dominated sorting.

[0020] In conjunction with the first aspect above, in one possible implementation, the objective balance evaluation of the non-dominated solutions in the Pareto solution set and the calculation of the cooperative balance index of each solution among multiple objective functions include: determining the ideal point of the Pareto solution set in each objective dimension and calculating the objective deviation value of each non-dominated solution relative to the ideal point; normalizing the objective deviation value to eliminate the influence of dimensions and generating a normalized deviation value; calculating the dispersion of the normalized deviation value of each non-dominated solution and using the dispersion as a cooperative balance index.

[0021] In conjunction with the first aspect above, in one possible implementation, obtaining the preference function characterizing the current operational preference and selecting the optimal control solution from the Pareto solution set in conjunction with the collaborative balance index includes: obtaining operating condition information reflecting the current pipeline operating conditions and determining the target weight in the preference function based on the operating condition information; applying the preference function and calculating a comprehensive evaluation value for each non-dominated solution in the Pareto solution set in conjunction with the target weight and the collaborative balance index; and selecting the non-dominated solution with the optimal comprehensive evaluation value as the optimal control solution.

[0022] In conjunction with the first aspect above, in one possible implementation, the method further includes: after issuing the control command, collecting post-control pipeline operation data generated by the control command; updating the source domain data using the post-control pipeline operation data; and adjusting the adversarial domain adaptation training based on the updated source domain data to achieve continuous optimization of the model.

[0023] In conjunction with the first aspect above, in one possible implementation, the method further includes: analyzing the optimal control solution and its corresponding pipeline operation results within a historical control cycle to generate a performance stability index; comparing the performance stability index with a threshold used to define a performance stability interval to generate a cycle adjustment decision; and dynamically adjusting the running cycle length of the dynamic multi-objective evolutionary algorithm based on the cycle adjustment decision.

[0024] Secondly, a multi-objective collaborative optimization automatic pipeline distribution control system is provided, comprising: a data acquisition module, a domain adaptation training module, an environmental monitoring module, an optimization model construction module, an optimization decision-making module, a balance evaluation module, and an optimal solution selection and instruction generation module; wherein:

[0025] The data acquisition module is used to acquire real-time operating parameters, equipment status and external environment data of the pipeline network as real-time status data, define the real-time status data as target domain data, and acquire historical operating data, historical energy consumption data and historical load data of the pipeline network and define them as source domain data.

[0026] The domain adaptation training module is used to generate robust features that can characterize the essential laws of pipeline network operation by performing adversarial domain adaptation training based on the source domain data and target domain data, and to obtain the domain adaptation error of the training process.

[0027] The environmental monitoring module is used to monitor the rate of change and domain adaptation error of the target domain data in real time and generate environmental mutation judgment results.

[0028] The optimization model building module is used to utilize the robust features to build a dynamic multi-objective optimization model that includes decision variables and multiple objective functions;

[0029] The optimization decision module is used to generate a Pareto solution set containing several non-dominated solutions based on the dynamic multi-objective optimization model and in response to the environmental mutation determination result, using a dynamic multi-objective evolutionary algorithm.

[0030] The balance evaluation module is used to evaluate the objective balance of the non-dominated solutions in the Pareto solution set and calculate the cooperative balance index of each solution among multiple objective functions.

[0031] The optimal solution selection and instruction generation module is used to obtain a preference function that characterizes the current operational preference, select the optimal control solution from the Pareto solution set in combination with the cooperative balance index, and generate control instructions to be issued to the actuators based on the optimal control solution.

[0032] Compared with the prior art, this application has the following advantages:

[0033] This application effectively eliminates the distributional differences between historical and real-time data by introducing adversarial domain adaptation training, generating robust features that can characterize the essential laws governing pipeline network operation. This makes the control model unaffected by data drift caused by factors such as pipeline aging and seasonal changes, improving the model's generalization ability and prediction accuracy under complex and variable operating conditions, and ensuring the long-term effectiveness and reliability of the control strategy.

[0034] This application establishes a sensitive environmental change monitoring and rapid response mechanism. By monitoring the rate of data change and model adaptation error in real time, it can accurately identify sudden changes in the pipeline network's operating status. In the event of environmental changes, it can automatically adjust the search strategy of the dynamic multi-objective evolutionary algorithm, thereby quickly adapting to the new operating environment and finding a new optimal control range, ensuring the stability and adaptability of the control system under dynamic and uncertain environments.

[0035] This application proposes a multi-dimensional decision-making method that balances operational preferences with objective equilibrium. It not only generates a series of non-dominated solutions but also quantifies the balance of each solution across multiple objectives by calculating a synergistic balance index, and performs a comprehensive evaluation by combining a preference function that dynamically changes with operating conditions. This ensures that the final selected optimal control solution satisfies the current operational priorities while avoiding sacrificing other key performance aspects due to excessive emphasis on one objective, achieving true synergy and balance among the objectives.

[0036] This application designs a closed-loop self-learning and adaptive optimization mechanism. By feeding back the actual operating data after control to update the model, the control method achieves continuous self-evolution, enabling it to automatically adapt to the long-term, slow changes in pipeline characteristics. Simultaneously, by analyzing performance stability and dynamically adjusting the optimization cycle, intelligent allocation of computing resources is achieved, reducing system operating costs while ensuring control accuracy, thus achieving an optimal balance between performance and efficiency.

[0037] It should be understood that the descriptions of technical features, technical solutions, beneficial effects, or similar language in this application do not imply that all features and advantages can be achieved in any single embodiment. Rather, it is understood that the description of a feature or beneficial effect means that a specific technical feature, technical solution, or beneficial effect is included in at least one embodiment. Therefore, the descriptions of technical features, technical solutions, or beneficial effects in this specification do not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions, and beneficial effects described in this embodiment can be combined in any suitable manner. Those skilled in the art will understand that embodiments can be implemented without one or more specific technical features, technical solutions, or beneficial effects of a particular embodiment. In other embodiments, additional technical features and beneficial effects may be identified in specific embodiments that do not embody all embodiments. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 A system architecture diagram of a multi-objective collaborative optimization pipeline automatic transmission and distribution control system provided in this application embodiment;

[0040] Figure 2 A flowchart illustrating a multi-objective collaborative optimization automatic pipeline distribution control method provided in this application embodiment;

[0041] Figure 3 This is a diagram illustrating the effect of adversarial domain adaptation training provided in an embodiment of this application.

[0042] Figure 4 This is a schematic diagram of the dynamic Pareto solution set responding to environmental mutations provided in the embodiments of this application. Detailed Implementation

[0043] In the description of this application, unless otherwise stated, " / " means "or," for example, A / B can mean A or B. The "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. Furthermore, "at least one" means one or more, and "multiple" means two or more. The terms "first," "second," etc., do not limit the quantity or order of execution, and "first," "second," etc., do not necessarily imply differences.

[0044] It should be noted that, in this application, the terms "exemplary" or "for example" are used to indicate that something is being described as an example, illustration, or illustration. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0045] The multi-objective collaborative optimization pipeline automatic transmission and distribution control method provided in this application embodiment can be applied to, for example... Figure 1 In the multi-objective collaborative optimization pipeline automatic transmission and distribution control system 100 shown, such as Figure 1 As shown, the system includes: a data acquisition module, a domain adaptation training module, an environmental monitoring module, an optimization model construction module, an optimization decision-making module, a balance evaluation module, and an optimal solution selection and instruction generation module;

[0046] Among them, the data acquisition module is used to acquire real-time operating parameters, equipment status and external environment data of the pipeline network as real-time status data, define the real-time status data as target domain data, and acquire historical operating data, historical energy consumption data and historical load data of the pipeline network and define them as source domain data.

[0047] The domain adaptation training module is used to generate robust features that can characterize the essential laws of pipeline network operation by performing adversarial domain adaptation training based on the source domain data and target domain data, and to obtain the domain adaptation error of the training process.

[0048] The environmental monitoring module is used to monitor the rate of change and domain adaptation error of the target domain data in real time and generate environmental mutation judgment results.

[0049] The optimization model building module is used to utilize the robust features to build a dynamic multi-objective optimization model that includes decision variables and multiple objective functions;

[0050] The optimization decision module is used to generate a Pareto solution set containing several non-dominated solutions based on the dynamic multi-objective optimization model and in response to the environmental mutation determination result, using a dynamic multi-objective evolutionary algorithm.

[0051] The balance evaluation module is used to evaluate the objective balance of the non-dominated solutions in the Pareto solution set and calculate the cooperative balance index of each solution among multiple objective functions.

[0052] The optimal solution selection and instruction generation module is used to obtain a preference function that characterizes the current operational preference, select the optimal control solution from the Pareto solution set in combination with the cooperative balance index, and generate control instructions to be issued to the actuators based on the optimal control solution.

[0053] like Figure 2 As shown in the figure, this application provides a multi-objective collaborative optimization method for automatic pipeline distribution control, including:

[0054] S1. Obtain real-time operating parameters, equipment status and external environment data of the pipeline network as real-time status data, define the real-time status data as target domain data, and obtain historical operating data, historical energy consumption data and historical load data of the pipeline network and define them as source domain data.

[0055] S2. Based on the source domain data and target domain data, by performing adversarial domain adaptation training, the difference between the source domain data and the target domain data is eliminated, robust features that can characterize the essential laws of pipeline network operation are generated, and the domain adaptation error of the training process is obtained.

[0056] S3. Monitor the rate of change and domain adaptation error of the target domain data in real time, and generate environmental mutation judgment results;

[0057] S4. Using the robust features, construct a dynamic multi-objective optimization model that includes decision variables and multiple objective functions;

[0058] S5. Based on the dynamic multi-objective optimization model, in response to the environmental mutation determination result, a Pareto solution set containing several non-dominated solutions is generated using a dynamic multi-objective evolutionary algorithm;

[0059] S6. Evaluate the objective balance of the non-dominated solutions in the Pareto solution set and calculate the synergistic balance index of each solution among multiple objective functions.

[0060] S7. Obtain the preference function used to characterize the current operational preference, select the optimal control solution from the Pareto solution set in combination with the cooperative balance index, and generate control instructions to be issued to the actuator based on the optimal control solution.

[0061] It should be noted that this application employs adversarial domain adaptation training technology, allowing the feature extractor and domain classifier to compete against each other, extracting robust features that reflect the inherent hydraulic characteristics of the pipeline network from historical source domain data and real-time target domain data with varying distributions. Through dual indicators—the rate of change of real-time data and the error during domain adaptation training—abrupt environmental changes are sensitively monitored and used as trigger signals. Using the extracted robust features, a dynamic multi-objective optimization model capable of reflecting real-time changes in the pipeline network is constructed. When a sudden environmental change occurs, this model can be efficiently solved by a dynamic multi-objective evolutionary algorithm, generating a Pareto solution set containing multiple optimal compromise schemes. Finally, a cooperative balance index is introduced to evaluate the balance of each scheme among multiple objectives, and combined with a preference function representing the current operational focus, a unique optimal control solution is selected from numerous optimal schemes, and control commands are issued.

[0062] In one possible implementation of the embodiments of this application, combined with Figure 2 The above S2 can be implemented through the following S21, S22 and S23, which are explained in detail below:

[0063] S21. Extract features from the source domain data and target domain data to generate control features, and perform domain classification on the control features to determine the data source and generate domain classification loss.

[0064] In some implementations, an adversarial network consisting of a feature extractor and a domain classifier is constructed and trained. The feature extractor receives input source or target domain data and maps it to a high-dimensional feature space, generating control features. The domain classifier receives the control features output by the feature extractor and determines whether the feature originates from source or target domain data. During training, these two modules work in opposition. On one hand, the domain classifier is optimized to minimize its domain classification loss, striving to accurately distinguish the data source. On the other hand, while optimizing the main task loss, the feature extractor is trained to maximize the domain classifier's classification loss, i.e., to confuse the domain classifier by generating indistinguishable control features. The main task loss is typically defined based on the source domain data, such as the loss for predicting pipeline energy consumption or pressure; it ensures that the extracted features are meaningful for the pipeline control task. Figure 3 The figure shows the curve of the domain classifier accuracy changing with the number of iterations during adversarial domain adaptation training. The curve gradually decreases from high accuracy and stabilizes around 0.5, indicating that the feature extraction operation successfully obfuscated the domain source of the data, thus achieving the extraction of domain-invariant robust features.

[0065] For example, in a pipeline distribution system, the feature extraction network adopts a 4-layer 1D-CNN network structure, with each layer having a convolutional kernel size of [missing information]. The activation function is LeakyReLU, which transforms the input 120-dimensional real-time operating state vector into a 64-dimensional control feature vector, containing 30 key nodes such as pressure, flow rate, water tank level, and pump station frequency. The domain classification network uses a classifier with three fully connected layers, and the number of neurons is as follows: After inputting the feature vector, the classifier outputs two probability values ​​representing the probability from either the source or target domain. For example, if the input feature vector outputs a probability of... This indicates that it was determined to be source domain data with a high degree of confidence.

[0066] S22. The feature extraction is trained by alternately optimizing the main task loss of the feature extraction and the domain classification loss of the domain classification, so that the feature extraction can optimize the domain classification error.

[0067] In some implementations, the adversarial training process can be achieved through a unified optimization objective, where the parameter updates of the feature extractor aim to minimize the main task loss while maximizing the domain classification loss. This is typically achieved by introducing a gradient reversal layer, which multiplies the gradient from the domain classification loss by a negative constant during backpropagation, thus transforming the minimization objective of the domain classifier into the maximization objective of the feature extractor. The overall network optimization objective function is expressed as: ;

[0068] in, Represents the overall optimization goal; , and These represent the network parameters of the feature extractor, the main task predictor, and the domain classifier, respectively. It is the prediction error obtained by evaluating the main task predictor on the source domain data, i.e., the main task loss; It measures the ability of a domain classifier to distinguish features between the source and target domains by evaluating the domain classification loss obtained from the domain classifier. This is a hyperparameter used to balance the weights of the main task loss and the domain classification loss during the feature extractor optimization process. Through optimization algorithms such as gradient descent, the parameters of the feature extractor and the main task predictor are adjusted. and Towards minimization The direction of the update, and the parameters of the domain classifier Then towards maximization The direction of the update forms a dynamic mini-maximum game process.

[0069] For example, the main task loss of feature extraction operation The mean squared error is defined as the error in predicting the total energy consumption of the pipeline network over the next 6 time steps (30 minutes). The domain classification loss is defined as the binary cross-entropy loss. Training uses... The optimizer uses a batch size of 128. In each iteration, the parameter update target for the feature extraction operation is: The parameter update target for the domain classification operation is For example, in a certain iteration, if for ,and for Then the feature extraction operation will mainly be geared towards maximizing The direction is adjusted to generate more "fuzzy" features.

[0070] S23. Apply the feature extraction after training convergence to real-time data, and use the output domain-invariant features as robust features.

[0071] In some implementations, when the training process converges, it means that the feature extractor has learned a feature representation that effectively supports the prediction of the main task and successfully obfuscates the differences between the source and target domains, making the classification accuracy of the domain classifier close to random guessing. At this point, the trained feature extractor is solidified. When new target domain data, i.e., real-time state data, is input, the features output by this feature extractor are domain-invariant features. These features capture the inherent laws of pipeline network operation that do not change with operating conditions, and are therefore defined as robust features for subsequent dynamic multi-objective optimization modeling.

[0072] For example, the convergence criterion is set as follows: the classification accuracy of the domain classifier is continuously increasing on the independent test set. Round iterations remain Within the range, and the energy consumption error predicted by the main task is less than At this point, training convergence is determined. In online applications, the converged feature extraction network is solidified and integrated into the real-time data processing flow of the control system. For example, given a target domain vector contaminated by sensor noise as real-time input, the robust features output by this network are significantly better than those obtained directly using... Its signal-to-noise ratio is improved This 64-dimensional vector is directly used as the initial input for the next stage of the algorithm to determine the optimal control solution.

[0073] In one possible implementation, combining Figure 2 The above-mentioned S3 can be implemented through the following S31, S32 and S33, which are explained in detail below:

[0074] S31. Calculate the change in the target domain data within a continuous time step and generate a rate of change index;

[0075] In some implementations, the rate of change of the target domain data over a continuous time step is calculated. The target domain data is a real-time status data vector containing multiple dimensions, such as pressure, flow rate, and equipment operating parameters at various monitoring points in the pipeline network. To eliminate the influence of different physical dimensions and numerical ranges, this data vector is first normalized. The difference between the normalized data vector of the current time step and the previous time step is calculated. The rate of change index is the norm of this difference vector, typically the L2 norm, and its calculation method is as follows:

[0076] ;

[0077] in, For the current moment The rate of change index; and These are the current times. and the previous time step The target domain data vector after normalization; This represents the L2 norm of the calculated vector, i.e., the Euclidean distance, which quantifies the magnitude of change in the overall operating status of the pipeline network over a short period of time.

[0078] For example, control cycle setting Minutes. Target domain data. It is dimensional vector, containing Pressure at key nodes Flow rate of the main pipeline The power of each pumping station, etc., and after Interval normalization. Rate of change index. Using the current time Compared to the previous moment The L2 norm of the normalized vector difference. For example, under normal operating conditions, Approximately When a pipe bursts or critical equipment trips in the pipeline network, Possibly The inside rapidly increased to above.

[0079] S32. Compare the rate of change index and the domain adaptation error with a threshold representing the normal operating fluctuation range to generate a comparison result;

[0080] In some implementations, the domain adaptation error defined during the previous stage of adversarial domain adaptation training is acquired in real time. This error is essentially a measure of the pre-trained domain classifier's ability to misclassify robust features extracted from the current real-time state data as source domain data. A high domain adaptation error value indicates that the current target domain data distribution has significantly deviated from the historical source domain data distribution used for training, suggesting that the underlying operating mode of the pipeline network may have changed.

[0081] For example, the threshold is set based on stable operating data over the past year. Confidence interval determination: threshold for rate of change index Set as The domain adaptation error threshold is the cross-entropy loss of the domain classifier on real-time data. Set as At a certain time step If the calculated And the domain adaptation error at the same time The following comparison result is generated: the rate of change index comparison result is "exceeding", that is... The domain adaptation error comparison result is "normal", that is... .

[0082] S33. Based on the comparison results, when the rate of change index or domain adaptation error exceeds the threshold, an environmental mutation determination result is generated to identify a sudden change in the environment.

[0083] In some implementations, the calculated rate of change and the real-time acquired domain adaptation error are compared with their respective thresholds. These two thresholds represent the reasonable fluctuation range of the pipeline network under normal operating conditions. Their values ​​are set based on statistical analysis of a large amount of historical stable operating data; for example, they can be set as the normal fluctuation mean plus three standard deviations. When the rate of change exceeds its threshold, or the domain adaptation error exceeds its threshold, a sudden change in the environment is determined, and a corresponding environmental change determination result is generated. This result is usually a Boolean signal used to trigger subsequent dynamic optimization adjustments.

[0084] For example, the mutation determination logic is set as an "OR" relationship: any indicator exceeding a threshold is considered a mutation. This is because the rate of change indicator exceeds the threshold. This immediately generates an environmental mutation determination result: True. This result then triggers a dynamic multi-objective evolutionary algorithm, for example, switching the population initialization strategy from "memory-based" to "random-based," and immediately increasing the mutation rate of the evolutionary operation from... Adjusted to This is to accelerate the search for new Pareto fronts and quickly adapt to new operating conditions.

[0085] In one possible implementation, combining Figure 2 The above S4 can be implemented through the following S41, S42 and S43, which are explained in detail below:

[0086] S41. Define a set of decision variables that include the operating parameters of the control equipment;

[0087] In some implementations, a set of decision variables is defined. Decision variables are physical quantities that the control system can directly adjust, constituting the solution space of the optimization problem. The set of decision variables mainly includes the operating parameters of all controllable equipment in the pipeline network. This set consists of the operating frequency of each variable frequency pump and the valve opening degree of each electric regulating valve. These variables together form a decision vector, where each element corresponds to the control setpoint of a specific piece of equipment, and its value range is strictly constrained by the equipment's physical performance and safe operating procedures.

[0088] For example, in a medium-sized water supply network, the set of decision variables is defined as a vector containing 15 elements. This vector consists of the real-time operating frequencies of five main variable frequency pump stations and the opening degrees of ten key electrically controlled regulating valves. For instance, the range of values ​​for the pump station frequencies is limited to... Hertz Between Hertz; the range of valve opening values ​​is limited to... arrive These ranges are preset based on the safety and operational limitations of the equipment.

[0089] S42. Establish multiple objective functions to quantify transmission and distribution energy consumption costs, pipeline pressure safety, and service quality tracking respectively;

[0090] In some implementations, a quantifiable objective function for multiple operational goals is established. To achieve multi-objective synergistic optimization, the operational goals—namely, transmission and distribution energy consumption costs, pipeline pressure safety, and service quality tracking—need to be transformed into specific mathematical functions. Transmission and distribution energy consumption cost objective function. Its goal is to minimize the energy consumption of the entire power transmission and distribution system. This function is primarily based on a power consumption model of the pump sets. The total power consumption is the sum of the power consumption of all operating pump sets, while the power consumption of a single pump is related to its flow rate and head, and consequently, to the frequency converter and valve opening in the decision variables. Therefore, It is a function with the decision vector as its independent variable. Pipeline pressure safety objective function. Its goal is to ensure that the pressure at all critical nodes in the pipeline network remains within preset safe upper and lower limits. This function is typically constructed as a penalty function; when the predicted pressure at any node exceeds its safe range, the function value increases non-linearly based on the degree and duration of the deviation. The pressure value itself is calculated using a hydraulic model based on decision vectors, therefore... It is also a function of the decision vector. Service quality tracking objective function. Its objective is to minimize the deviation between the service pressure of key user nodes and the expected pressure setpoint. This function calculates the weighted sum of the differences between the predicted pressure values ​​and the target pressure values ​​for all key service points. Similarly, the pressure at each point is a function of the decision vector, which makes... It also depends on the decision vector.

[0091] For example, multiple objective functions include: a transmission and distribution energy cost objective, minimizing the total electricity cost in the next control cycle. This function is calculated based on the real-time power curves of each pumping station and the local peak-valley electricity price schedule. A pipeline pressure safety objective: ensuring that the pressure at 20 key monitoring points must be maintained within safe pressure limits. This function is constructed as a penalty term, imposing a large quadratic penalty if the predicted pressure exceeds this range. A service quality tracking objective: minimizing the percentage deviation between the predicted water supply pressure and the scheduled expected pressure in 10 major user service areas.

[0092] S43. The robust features are integrated as dynamic input parameters into a multi-objective function to adjust the parameters and model constraints of the multi-objective function in real time, forming a dynamic multi-objective optimization model.

[0093] In some implementations, robust features are integrated to form a dynamic multi-objective optimization model. This is key to achieving dynamic adaptability of the model. In traditional methods, the hydraulic models upon which each objective function depends are static, with parameters such as pipe roughness and local head loss coefficients fixed. The robust features generated through adversarial domain adaptation training are then integrated into the model. As dynamic input parameters, they are integrated into the hydraulic model and objective function. Robustness characteristics. It contains deep information that cannot be directly measured, such as the current pipeline network operation mode and unmodeled changes in water demand. By... As one of the inputs to the hydraulic model, for example This allows pressure prediction to reflect the actual hydraulic characteristics of the pipe network in real time. Furthermore, It can also be used to dynamically adjust model constraints, such as adjusting the minimum service pressure standard based on current demand patterns. Ultimately, the optimization model is expressed as finding the optimal decision vector. To simultaneously minimize a set of dependencies and dynamic parameters The objective function is:

[0094] .

[0095] For example, the 64-dimensional robust feature vector output from the adversarial domain adaptation module is used as a dynamic calibration input to the model. For instance, this vector is input to a parametric adaptive network to adjust the Hessian-Williams coefficient (HW roughness) in the network hydraulic model in real time to correct for hydraulic model distortions caused by network aging or sudden failures. Furthermore, the robust feature vector is also used to dynamically adjust the minimum safe pressure threshold in the pressure safety objective, allowing for a safe slight reduction of the minimum pressure threshold to conserve energy when extremely low loads are predicted. The above refers to energy consumption.

[0096] In one possible implementation, combining Figure 2 The above S5 can be implemented through the following S51, S52 and S53, which are explained in detail below:

[0097] S51. Generate the initial population using the Pareto solution set from the previous control period or the prediction results based on robust features.

[0098] In some implementations, the algorithm requires an initial population to initiate the evolutionary search for the current control cycle. The generation strategy of the initial population is closely related to the operating state of the pipeline network. When the pipeline network is operating smoothly, i.e., no environmental mutations are detected, the algorithm directly uses the Pareto solution set calculated in the previous control cycle as the initial population for the current cycle. This "warm start" method inherits the high-quality solutions from the previous moment, allowing the algorithm to fine-tune and optimize based on existing good foundations, thereby improving the convergence speed. Conversely, if the environmental monitoring module issues a judgment result of an environmental mutation, it indicates that the original optimal solution region may have become invalid. At this time, the algorithm will make predictions based on the latest robust features and generate a new set of solutions that are more likely to adapt to the current new environment as the initial population, or increase the population diversity by introducing random solutions to prepare for searching for new optimal regions.

[0099] For example, a dynamic variant of the NSGA-II algorithm is used. Normally, the algorithm directly uses the Pareto solution set containing 100 non-dominated solutions obtained in the previous control cycle as the initial population for the current cycle; this is called a memory-based "warm start." If the environmental monitoring module determines that a mutation has occurred, the algorithm activates a robust feature-based prediction initialization: the current 64-dimensional robust feature vector is input into a pre-trained prediction network to generate five predicted optimal control solutions, and these five solutions are mixed with 95 random solutions as a new initial population.

[0100] S52. If the environmental mutation determination result is yes, then adjust the search parameters of the evolution operation and perform the evolution operation on the initial population to generate the offspring population.

[0101] In some implementations, the core of the algorithm lies in its response mechanism to environmental mutations. Upon receiving an environmental mutation determination result indicating a change in the environment, the algorithm immediately adjusts its internal evolutionary search parameters. For example, it automatically increases the strength and probability of the mutation operator or introduces a dedicated population diversity maintenance mechanism. This adjustment aims to enhance the algorithm's exploration capability, enabling the population to break free from the constraints of the previous optimal solution region and explore broader unknown areas in the solution space, quickly locating the new Pareto front formed by the environmental mutation. If there is no environmental mutation, the algorithm maintains normal search parameters, focusing on a refined search of the current optimal region to enhance its convergence ability. The algorithm performs evolutionary operations, including selection, crossover, and mutation, on the strategically initialized population to generate a progeny population. Then, the initial population is merged with the progeny population, and a dynamic multi-objective optimization model is used to evaluate each individual in the merged population—that is, each candidate control strategy—calculating the corresponding values ​​of three objective functions: transmission and distribution energy consumption cost, pipeline pressure safety, and service quality tracking. Figure 4The diagram shows the changes in the Pareto solution set of the dynamic multi-objective evolutionary algorithm before and after a sudden environmental change. After the change, the optimal solution set shifts towards higher costs and higher penalties. By adjusting the search parameters, the algorithm can quickly transition from the old Pareto front to a new Pareto front, demonstrating its dynamic adaptability.

[0102] For example, when the environmental mutation determination result is "yes", the algorithm immediately changes the crossover probability from the normal value. Reduce to To protect superior individuals while reducing the mutation rate from Significantly increased to The mutation operator is switched from polynomial mutation to non-uniform mutation to enhance the population's global exploration ability and encourage it to quickly jump out of the old optimal region. If the result is "no", the mutation rate remains at [value missing]. The algorithm focuses on fine-grained local search. Subsequently, it performs selection, crossover, and mutation operations on the initial population to generate a progeny population of 100 individuals.

[0103] S53. The initial population and offspring population are evaluated using the dynamic multi-objective optimization model, and a Pareto solution set is generated through non-dominated sorting.

[0104] In some implementations, the algorithm employs a non-dominated sorting technique to stratify the entire evaluated population. The principle of non-dominated sorting is that if all objective function values ​​of individual A are not inferior to those of individual B, and at least one objective function value is superior to that of individual B, then A is said to dominate B. All individuals not dominated by any other individual in the population are assigned to the first stratum, which constitutes the Pareto solution set obtained in this round of optimization. These solutions are all non-dominated solutions, representing the optimal trade-offs between different objectives under the current conditions.

[0105] For example, the initial population and the offspring population are merged into one. A mixed population of individuals. The three objective function values ​​of these 200 candidate solutions were evaluated using a dynamic multi-objective optimization model. Subsequently, NSGA-II non-dominated ranking and crowding distance calculations were performed on these 200 individuals: all individuals determined to be in the first layer constituted the Pareto solution set for this period; for example, 112 non-dominated solutions were obtained in this period, which were used for subsequent cooperative balance evaluation.

[0106] In one possible implementation, combining Figure 2 The above-mentioned S6 can be implemented through the following S61, S62 and S63, which are explained in detail below:

[0107] S61. Determine the ideal point of the Pareto solution set in each target dimension, and calculate the target deviation value of each non-dominated solution relative to the ideal point;

[0108] In some implementations, an ideal point is determined and the target deviation value is calculated. After obtaining a Pareto solution set containing several non-dominated solutions, a reference benchmark, i.e., the ideal point, needs to be determined. The ideal point is a virtual solution in the current Pareto solution set, composed of the optimal values ​​achievable by each objective function. For the three minimization objectives of transmission and distribution energy consumption cost, pipeline pressure safety, and service quality tracking, the coordinates of the ideal point are composed of the minimum value of each objective dimension in the Pareto solution set. For each non-dominated solution, its target deviation value in each objective dimension is the difference between its objective function value and the corresponding dimension value of the ideal point.

[0109] For example, when analyzing a Pareto solution set containing 100 non-dominated solutions, an ideal point is determined for three objectives: the energy cost at the ideal point is the lowest among the 100 solutions, for example, 1500 yuan; the stress safety penalty is the lowest; and the service quality deviation is the lowest. For any non-dominated solution, its objective deviation value is the absolute difference between its objective function value and the ideal point value. For example, if the energy cost of a non-dominated solution is 1800 yuan, then its energy deviation value is... Yuan.

[0110] S62. Normalize the target deviation value to eliminate the influence of dimensions and generate a normalized deviation value;

[0111] In some implementations, the target deviation values ​​are normalized. Since the physical meanings and numerical ranges of the three objective functions are different, directly comparing the target deviation values ​​is unreasonable. To eliminate the influence of dimensions, it is necessary to normalize the deviation values ​​for each objective dimension. Normalization typically uses the max-min normalization method, mapping the deviation value of each non-dominated solution in a certain objective dimension to the interval between 0 and 1. The formula for calculating the normalized deviation value is:

[0112] ;

[0113] in, It is the first The non-inferior solution is at the th . Normalized bias values ​​across each target dimension; It is the first The non-inferior solution is at the th . Original target deviation values ​​relative to the ideal point in each target dimension; and These are all solutions in the Pareto solution set at the th . The minimum and maximum values ​​of the target deviation values ​​across each target dimension are determined. This step numerically evens out the importance of all target dimensions, allowing for subsequent cross-sectional comparisons.

[0114] For example, to eliminate the dimensional differences between "yuan," "point," and "percentage," a maximum-minimum normalization method is used to process the deviation values ​​of all non-dominated solutions. For instance, the energy consumption deviation value has a maximum of 600 yuan and a minimum of 100 yuan among all solutions. For a solution with a deviation value of 300 yuan, the normalized deviation value is calculated as follows: After this processing, the deviation values ​​of all solutions on the three objectives are mapped to one. Within the dimensionless interval.

[0115] S63. Calculate the degree of dispersion of the normalized deviation value of each non-dominated solution, and use the degree of dispersion as an index of cooperative balance.

[0116] In some implementations, the dispersion of the normalized deviation values ​​is calculated and used as an indicator of cooperative balance. For each non-dominated solution, a set of normalized deviation values ​​is now available, reflecting its relative proximity to the "absolute optimum" across all objectives. A solution that performs evenly across objectives will have relatively close and concentrated normalized deviation values; conversely, an extreme solution that performs exceptionally well on some objectives but poorly on others will have highly dispersed normalized deviation values. Therefore, calculating the dispersion of these normalized deviation values ​​quantifies the solution's balance. Dispersion can be measured by calculating its standard deviation or variance. Cooperative Balance Indicator The calculation is as follows:

[0117]

[0118] in, It is the first A co-equilibrium index for non-dominated solutions; This is the number of objective functions, typically set to 3. It is the first The average of all normalized bias values ​​for each solution. Cooperative balance index. The smaller the value, the more balanced the solution is across multiple objectives and the better its synergy.

[0119] For example, the synergistic balance index The standard deviation of the three normalized biases for each non-dominated solution is calculated. A well-balanced solution will have very similar normalized biases, resulting in a standard deviation close to zero. Conversely, an extreme solution will have significantly different normalized biases, leading to a larger standard deviation. The scheduling system will... The value serves as a key indicator for measuring the synergistic balance of the control strategy among multiple objectives.

[0120] In one possible implementation, combining Figure 2The above-mentioned S7 can be implemented through the following S71, S72 and S73, which are explained in detail below:

[0121] S71. Obtain operating condition information reflecting the current pipeline network operating conditions, and determine the target weight in the preference function based on the operating condition information;

[0122] In some implementations, current pipeline network operating condition information is obtained, and the target weights in the preference function are dynamically determined based on this information. Operating condition information refers to real-time information reflecting the current operating status of the pipeline network and external demand, such as whether it is currently during peak or off-peak water usage periods, whether emergency water use events such as fires have occurred, or whether there are reports of insufficient pressure in specific areas. Based on this specific operating condition information, a pre-set rule base or model is invoked to assign different weights to the three target functions: transmission and distribution energy consumption cost, pipeline pressure safety, and service quality tracking. For example, during off-peak hours at night, energy conservation and consumption reduction are the primary tasks, so the weight of the transmission and distribution energy consumption cost target will be increased accordingly; while during peak days, ensuring user service quality is more critical, and the weight of the service quality tracking target will be increased. These weights constitute the preference function representing the current operational preferences and are a quantitative manifestation of the decision-maker's intentions.

[0123] For example, it is detected in real time that the current period is a "nighttime off-peak period" and there are no sudden events. Based on preset operational rules, a preference function is invoked to dynamically allocate weights to three objectives: the weight of the transmission and distribution energy consumption cost objective is set to 0.6; the weight of the pipeline pressure safety objective is set to 0.3; and the weight of the service quality tracking objective is set to 0.1. This set of weight vectors... This is a quantitative reflection of current operational preferences.

[0124] S72. Apply the preference function, and combine the target weight and the collaborative balance index to calculate the comprehensive evaluation value of each non-dominated solution in the Pareto solution set;

[0125] In some implementations, a preference function is applied, combined with a previously calculated collaborative balance index, to calculate a comprehensive evaluation value for each non-dominated solution in the Pareto solution set. The comprehensive evaluation value aims to provide a holistic score for each solution, integrating two dimensions: first, whether the solution aligns with current operational preferences; and second, whether a good balance has been achieved among multiple objectives. The calculation formula is as follows:

[0126] ;

[0127] in, It is the first The comprehensive evaluation value of each non-dominated solution; It is a preset hyperparameter with a value between 0 and 1, used to adjust the relative importance of operational preferences and synergy balance in the final evaluation; The first one is determined based on the current working condition information. The weights of each objective function; It is the first The non-inferior solution is at the th . The normalized objective value is calculated over several objective functions. This normalization process aims to eliminate dimensional differences between different objective functions, making the weighted sum physically meaningful. The first part of the formula measures the degree to which the solution satisfies the current operational preferences, while the second part evaluates the solution's own internal equilibrium.

[0128] For example, a combination of weighted summation and balance penalty is used to calculate the comprehensive evaluation value. For each non-dominated solution in the Pareto solution set, its weighted objective value is first calculated: the three normalized objective values ​​of the solution are multiplied by the objective weights under the current operating condition. Then sum the results. Next, add a penalty term to this weighted sum that is proportional to the co-balance index of the solution. For example, the final comprehensive evaluation value of a solution is... The comprehensive evaluation value of the other solution is .

[0129] S73. Select the non-dominated solution with the best comprehensive evaluation value as the optimal control solution.

[0130] In some implementations, the non-dominated solution with the best overall evaluation value is selected as the final optimal control solution. Since each objective function and the cooperative balance index are defined as better the smaller they are, the solution with the smallest overall evaluation value is selected. The non-dominated solution is considered the most ideal control scheme under the current operating conditions. This solution not only largely satisfies the current operational priorities, but its decision-making scheme itself also avoids excessive bias between different objectives, exhibiting good overall performance. Once the optimal control solution is determined, its corresponding decision variables, namely the frequencies of each frequency converter and the valve opening degree, constitute the specific control instructions issued to the field actuators.

[0131] For example, the overall evaluation values ​​of all 100 non-dominated solutions in the Pareto solution set are sorted. For instance, among all solutions, the overall evaluation value of a particular solution is... This is the minimum value. This solution is then selected as the optimal control solution for the current control cycle. This solution corresponds to a specific set of decision variables, such as the operating frequency of pump station P3. Hertz, control valve Opening These specific values ​​are then used as control commands and sent to the field actuators through the real-time automated control system.

[0132] In one possible implementation, after S73, the multi-objective cooperative optimization pipeline automatic transmission and distribution control method provided in this application embodiment further includes:

[0133] After the control command is issued, the post-control pipeline operation data generated by the control command is collected;

[0134] In some implementations, after generating and issuing control commands based on the optimal control solution, operational data of the pipeline network is continuously collected. This data, known as post-control pipeline network operational data, represents the actual response after the optimization decision was implemented, comprehensively recording the actual performance of various parameters such as pressure, flow rate, and equipment energy consumption under the control command. This data is the most direct and reliable basis for evaluating control effectiveness and identifying model deviations.

[0135] For example, during the 5-minute control cycle following the issuance of the optimal control solution, key data is continuously collected. This data includes: real-time power and head of all 5 pump stations, actual pressure readings at 20 key monitoring points, and actual energy consumption. For instance, if the instruction requires pump station P3 to operate at... Then what is collected is P3 at The actual energy consumption during operation and the resulting downstream pressure data form a control network operation record containing 100-dimensional data.

[0136] The source domain data is updated using the control network operation data.

[0137] In some implementations, newly acquired post-control network operation data is used to update the source domain data. Source domain data forms the basis for adversarial domain adaptation training, representing the model's known "historical experience." Dynamic expansion of the knowledge base is achieved by supplementing the source domain database with operation data containing the latest control effects. The update operation can employ various strategies; for example, new data can be directly appended to the existing database, or a sliding window mechanism can be used to replace the oldest data with new data, ensuring that the source domain data always reflects the recent operational characteristics of the network.

[0138] For example, a sliding window strategy is used to update source domain data. The source domain database always maintains the latest 10,000 historical operation records. Each time a new control network operation record is obtained, it is appended to the end of the source domain database, while the oldest record in the database is deleted. For example, after adding the 10,001st record, the 1st record is removed, ensuring that the source domain data always reflects the operation characteristics of the network over the past 35 days.

[0139] Based on the updated source domain data, the adversarial domain adaptation training is adjusted to achieve continuous model optimization.

[0140] In some implementations, the original adversarial domain adaptation training process is adjusted and retrained based on updated source domain data. As the source domain data changes, the original feature extractor may no longer be able to optimally extract domain-invariant robust features. Therefore, it is necessary to re-execute the adversarial domain adaptation training process using updated, richer, and more realistic source domain data. This can take the form of online incremental learning, i.e., fine-tuning the original model parameters to adapt to the distribution changes brought about by the new data; or it can be periodic full retraining to completely reconstruct the feature extractor and domain classifier. In this way, the adversarial domain adaptation model can continuously learn from new practices and correct its understanding of the essential laws governing pipeline network operation.

[0141] For example, the training adjustments employ an incremental learning strategy. Small-scale incremental training is performed every 24 hours: using 288 newly added data points as training batches, the parameters of the original feature extractor and domain classifier are fine-tuned for 1000 iterations. This allows the model to quickly adapt to slow changes such as sensor drift or slight network aging without requiring retraining from scratch.

[0142] Analyze the optimal control solution and its corresponding pipeline operation results within the historical control cycle to generate performance stability index;

[0143] In some implementations, the optimal control solution and its corresponding pipeline operation results within historical control cycles are analyzed to generate a performance stability index. The optimal control solution selected for each control cycle over a past period, along with the pipeline operation results generated by that solution in practical applications, are recorded and analyzed. The performance stability index aims to quantify the consistency and volatility of recent control performance. A simple performance stability index can be the magnitude of change of the optimal control solution in the decision space over multiple consecutive control cycles, or the variance of the corresponding objective function value sequence. For example, the trace or determinant of the covariance matrix of the objective function value vector sequence of the optimal control solution over the most recent N cycles can be calculated; the smaller the value, the more stable the performance.

[0144] For example, record the most recent The average variation of the optimal control solution in the decision variable space over a series of consecutive control cycles. For example, calculating the standard deviation of the periodic variations in the frequency and valve opening of each pump station over 10 cycles. If the sum of these standard deviations is less than... If the sum is greater than 1, the control strategy is considered very stable, and the performance stability index is set to low volatility. This indicates that the optimal solution changes frequently and drastically, and the performance stability index is set to high volatility.

[0145] The performance stability index is compared with a threshold used to define the performance stability range to generate a periodic adjustment decision;

[0146] In some implementations, the calculated performance stability index is compared with a preset threshold to generate a periodic adjustment decision. One or more thresholds are predefined, forming a performance stability range. When the performance stability index falls within this range, it indicates that the recent pipeline network operation is stable, and the current control strategy is performing well and stably, requiring no frequent complex global optimization. Conversely, when the index exceeds the stability range, it means that the pipeline network operation may be changing, or the current control strategy is experiencing significant fluctuations, requiring more frequent optimization calculations for a faster response. Based on this comparison, a periodic adjustment decision is generated, which clarifies whether the length of the next control cycle should be extended, shortened, or remained unchanged.

[0147] For example, two thresholds are preset: a high stability threshold of 0.08 and a low stability threshold of 0.25. The current control cycle length is set to 5 minutes. The calculated performance stability index is compared with the thresholds: if the index is less than the high stability threshold of 0.08, it indicates a hyper-stable state, and a cycle adjustment decision of "extend the cycle" is generated. If the index is between the high stability threshold of 0.08 and the low stability threshold of 0.25, it indicates a normal state, and a decision of "remain unchanged" is generated. If the index is greater than the low stability threshold of 0.25, it indicates instability, and a decision of "shorten the cycle" is generated.

[0148] Based on the aforementioned periodic adjustment decision, the running period length of the dynamic multi-objective evolutionary algorithm is dynamically adjusted.

[0149] In some implementations, the runtime of the dynamic multi-objective evolutionary algorithm is dynamically adjusted based on a periodic adjustment decision. Upon receiving the periodic adjustment decision, the control system modifies its internal timer or scheduler settings accordingly. If the decision is to extend the period, the dynamic multi-objective evolutionary algorithm will wait longer before being triggered again. During this period, it can continue executing the current optimal control solution or adopt a simpler local adjustment strategy. If the decision is to shorten the period, the algorithm's execution frequency will increase to more closely track changes in the pipeline network state. If the decision is to keep it unchanged, the runtime length remains the same.

[0150] For example, upon receiving a decision to "extend the cycle," the algorithm's execution cycle length is dynamically adjusted from the current 5 minutes to 7.5 minutes. This reduces the algorithm's execution frequency. Thus saving energy in a steady state. The computational resources are limited. Conversely, if a decision to "shorten the cycle" is received, the cycle length will be adjusted to 3 minutes to ensure that the algorithm can respond more quickly to potential system instability.

[0151] It should be noted that the electrical connections between the various units described above do not necessarily represent direct or indirect connections. Any indirect connection method is applicable to the embodiments of this application as long as it achieves the purpose of this application. The above descriptions are merely exemplary embodiments of this application and should not be construed as limiting the scope of this application.

[0152] All equivalent changes and modifications made in accordance with the teachings of this application shall still fall within the scope of this application. Other embodiments of this application will be readily apparent to those skilled in the art upon consideration of the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary technical means in the art not described in this application.

Claims

1. A multi-objective collaborative optimization method for automatic pipeline transmission and distribution control, characterized in that, The method includes: The real-time operating parameters, equipment status, and external environment data of the pipeline network are acquired as real-time status data. The real-time status data is defined as target domain data. The historical operating data, historical energy consumption data, and historical load data of the pipeline network are acquired and defined as source domain data. Based on the source domain data and target domain data, adversarial domain adaptation training is performed to eliminate the differences between the source domain data and the target domain data, generate robust features that can characterize the essential laws of pipeline network operation, and obtain the domain adaptation error of the training process. The system monitors the rate of change and domain adaptation error of the target domain data in real time and generates an environmental mutation judgment result. This real-time monitoring of the rate of change and domain adaptation error of the target domain data includes: calculating the change in the target domain data within a continuous time step and generating a rate of change index; comparing the rate of change index and domain adaptation error with a threshold representing the fluctuation range of normal operation and generating a comparison result; and, based on the comparison result, generating an environmental mutation judgment result indicating a sudden change in the environment when the rate of change index or domain adaptation error exceeds the threshold. Using the robust features, a dynamic multi-objective optimization model containing decision variables and multiple objective functions is constructed; Based on the dynamic multi-objective optimization model, and in response to the environmental mutation determination result, a Pareto solution set containing several non-dominated solutions is generated using a dynamic multi-objective evolutionary algorithm. The non-dominated solutions in the Pareto solution set are evaluated for objective balance. A cooperative balance index for each solution across multiple objective functions is calculated. This evaluation includes: determining the ideal point of the Pareto solution set in each objective dimension and calculating the objective deviation value of each non-dominated solution relative to the ideal point; normalizing the objective deviation value to eliminate dimensional influence and generating a normalized deviation value; calculating the dispersion of the normalized deviation value for each non-dominated solution and using the dispersion as a cooperative balance index. Obtain a preference function to characterize the current operational preferences, select the optimal control solution from the Pareto solution set in combination with the cooperative balance index, and generate control instructions to be issued to the actuators based on the optimal control solution.

2. The multi-objective collaborative optimization method for automatic pipeline transmission and distribution control according to claim 1, characterized in that, The robust features generated that characterize the essential laws governing pipeline network operation include: Feature extraction is performed on the source domain data and target domain data to generate control features, and the data source is determined by domain classification of the control features to generate domain classification loss; The feature extraction is trained by alternately optimizing the main task loss of feature extraction and the domain classification loss of domain classification, so that feature extraction can optimize the domain classification error. The features extracted after training convergence are applied to real-time data, and the output domain-invariant features are used as robust features.

3. The multi-objective collaborative optimization method for automatic pipeline transmission and distribution control according to claim 1, characterized in that, The construction of the dynamic multi-objective optimization model, which includes decision variables and multiple objective functions, includes: Define a set of decision variables that include the operating parameters of the control equipment; Establish multiple objective functions to quantify transmission and distribution energy consumption costs, pipeline pressure safety, and service quality tracking; The robust features are integrated as dynamic input parameters into a multi-objective function to adjust the parameters and model constraints of the multi-objective function in real time, forming a dynamic multi-objective optimization model.

4. The multi-objective collaborative optimization pipeline automatic transmission and distribution control method according to claim 1, characterized in that, Based on the dynamic multi-objective optimization model, and in response to the environmental mutation determination result, a Pareto solution set containing several non-dominated solutions is generated using a dynamic multi-objective evolutionary algorithm, including: The initial population is generated using the Pareto solution set from the previous control period or the prediction results based on robust features. If the environmental mutation determination result is yes, then the search parameters of the evolutionary operation are adjusted, and the evolutionary operation is performed on the initial population to generate the offspring population; The initial population and offspring population are evaluated using the dynamic multi-objective optimization model, and a Pareto solution set is generated through non-dominated sorting.

5. The multi-objective collaborative optimization method for automatic pipeline transmission and distribution control according to claim 1, characterized in that, The step of obtaining the preference function to characterize the current operational preference, and selecting the optimal control solution from the Pareto solution set in conjunction with the collaborative balance index, includes: Obtain operating condition information reflecting the current pipeline network operating conditions, and determine the target weight in the preference function based on the operating condition information; The preference function is applied, and the target weight and the collaborative balance index are combined to calculate the comprehensive evaluation value of each non-dominated solution in the Pareto solution set; The non-dominated solution with the best comprehensive evaluation value is selected as the optimal control solution.

6. The multi-objective collaborative optimization method for automatic pipeline transmission and distribution control according to claim 1, characterized in that, The method further includes: After the control command is issued, the post-control pipeline operation data generated by the control command is collected; The source domain data is updated using the control network operation data. Based on the updated source domain data, the adversarial domain adaptation training is adjusted to achieve continuous model optimization.

7. The multi-objective collaborative optimization method for automatic pipeline transmission and distribution control according to claim 1, characterized in that, The method further includes: Analyze the optimal control solution and its corresponding pipeline operation results within the historical control cycle to generate performance stability index; The performance stability index is compared with a threshold used to define the performance stability range to generate a periodic adjustment decision; Based on the aforementioned periodic adjustment decision, the running period length of the dynamic multi-objective evolutionary algorithm is dynamically adjusted.

8. A multi-objective collaborative optimization automatic pipeline transportation and distribution control system, characterized in that, The system is used in a multi-objective collaborative optimization pipeline automatic transmission and distribution control method as described in any one of claims 1-7, the system comprising: The data acquisition module is used to acquire real-time operating parameters, equipment status and external environment data of the pipeline network as real-time status data, define the real-time status data as target domain data, and acquire historical operating data, historical energy consumption data and historical load data of the pipeline network and define them as source domain data. The domain adaptation training module is used to generate robust features that can characterize the essential laws of pipeline network operation by performing adversarial domain adaptation training based on the source domain data and target domain data, and to obtain the domain adaptation error of the training process. An environmental monitoring module is used to monitor the rate of change and domain adaptation error of the target domain data in real time and generate an environmental mutation judgment result. The real-time monitoring of the rate of change and domain adaptation error of the target domain data includes: calculating the change in the target domain data within a continuous time step and generating a rate of change index; comparing the rate of change index and the domain adaptation error with a threshold representing the normal operating fluctuation range and generating a comparison result; and, based on the comparison result, generating an environmental mutation judgment result indicating that a sudden change in the environment has occurred when the rate of change index or the domain adaptation error exceeds the threshold. The optimization model building module is used to utilize the robust features to build a dynamic multi-objective optimization model that includes decision variables and multiple objective functions; The optimization decision module is used to generate a Pareto solution set containing several non-dominated solutions based on the dynamic multi-objective optimization model and in response to the environmental mutation determination result, using a dynamic multi-objective evolutionary algorithm. The balance evaluation module is used to evaluate the target balance of the non-dominated solutions in the Pareto solution set and calculate the cooperative balance index of each solution among multiple objective functions. The target balance evaluation of the non-dominated solutions in the Pareto solution set includes: determining the ideal point of the Pareto solution set in each objective dimension and calculating the target deviation value of each non-dominated solution relative to the ideal point; normalizing the target deviation value to eliminate the influence of dimensions and generating a normalized deviation value; calculating the dispersion of the normalized deviation value of each non-dominated solution and using the dispersion as a cooperative balance index. The optimal solution selection and instruction generation module is used to obtain a preference function that characterizes the current operational preference, select the optimal control solution from the Pareto solution set in combination with the cooperative balance index, and generate control instructions to be issued to the actuators based on the optimal control solution.

Images