A collaborative method and apparatus for predicting and optimizing parameters for pump station energy efficiency

Abstract

This invention discloses a collaborative method and apparatus for predicting and optimizing parameters in pump station energy efficiency, relating to the interdisciplinary fields of computational intelligence and optimization control. The method includes: establishing a predictive-optimization collaborative framework; using a dual-thread engine, shared memory, and a cache to achieve collaborative data interaction and model solving between a long short-term memory neural network and an improved particle swarm optimization algorithm, forming a collaborative data flow; collecting historical operating data of the pump station and performing feature engineering processing, inputting it into a two-layer long short-term memory neural network model, and outputting a predicted value for future short-term water demand; constructing a pump station energy consumption minimization optimization model based on this prediction value, setting it as a dynamic flow constraint, and solving it in conjunction with relevant operating constraints of the pipeline network and pumps; using a particle swarm optimization algorithm incorporating multiple improved strategies to solve the model, and outputting the optimal control parameter set after convergence to achieve optimized pump station energy consumption control. This invention achieves optimal pump station energy consumption through prediction-driven optimization, solving the problem of traditional scheduling lag.

Description

Technical Field

[0001] This invention relates to the field of computational intelligence and optimization control, and in particular to a method and apparatus for predicting and optimizing parameters for pump station energy efficiency. Background Technology

[0002] In the interdisciplinary field of computational intelligence and optimal control, energy-saving operation and optimized scheduling of water supply pumping stations is a core element in achieving efficient energy consumption reduction in water supply systems. Pumping station optimal scheduling is essentially a multivariable, nonlinear, and strongly coupled constrained integer programming problem, and its solution directly determines the operating efficiency and energy consumption level of the pumping station. Currently, the industry often uses conventional optimization methods such as traditional particle swarm optimization algorithms to solve pumping station scheduling problems. However, these methods have inherent technical defects, are prone to premature convergence, have weak local search capabilities, and are difficult to adapt to the instantaneous changes in operating conditions during actual pumping station operation. They cannot simultaneously meet the dual requirements of optimization algorithms for solution speed and computational accuracy, making it difficult to output the optimal scheduling scheme.

[0003] Meanwhile, in existing pump station scheduling research, water load prediction and operation optimization have long been implemented independently, lacking an effective dynamic linkage mechanism. Prediction results are only introduced into the optimization process as static input parameters, failing to be deeply integrated with the entire optimization solution process. This prevents real-time adjustments to optimization constraints and solution directions based on dynamic changes in prediction information, resulting in insufficient foresight in the generated optimized scheduling schemes. This disconnected approach fails to fully realize the value of prediction information, making it impossible to leverage short-term water demand predictions to proactively avoid unnecessary start-ups and shutdowns and inefficient operation of pump station equipment. This not only restricts the actual effectiveness of energy conservation and consumption reduction in pump stations but also fails to meet the real-time and accurate scheduling requirements of water supply systems. Therefore, there is an urgent need for a method to optimize the energy-saving operation of water supply pump stations by achieving deep dynamic collaboration between water load prediction and optimization algorithms. Summary of the Invention

[0004] The main objective of this invention is to provide a collaborative method for predicting and optimizing parameters for pump station energy efficiency.

[0005] Another objective of this invention is to propose a collaborative device for predicting and optimizing parameters for pump station energy efficiency.

[0006] The third objective of this invention is to provide an electronic device.

[0007] A fourth objective of this invention is to provide a non-transitory computer-readable storage medium.

[0008] To achieve the above objectives, a first aspect of the present invention proposes a collaborative method for prediction and parameter optimization of pump station energy efficiency, comprising:

[0009] S1. Construct a prediction-optimization collaborative framework, determine the collaborative relationship between the two-layer long short-term memory neural network water volume prediction model and the improved particle swarm optimization algorithm, realize data interaction through a dual-thread engine and shared memory, and use the output of the prediction model as the dynamic setting value of the algorithm flow constraint to form a collaborative basic data flow. S2 collects historical operating data of the pumping station and performs feature engineering processing. The processed feature data is then input into a pre-trained two-layer long short-term memory neural network model according to the standard format of the collaborative basic data stream. The model then calculates and outputs the predicted value of the short-term water demand in the future. S3. Construct an optimization model with the goal of minimizing the total operating energy consumption of the pumping station. Use the predicted value of the short-term water demand in the future as the basis for setting the dynamic upper and lower limits of the flow constraint. Combine the constraints of pipeline pressure, pump speed and maximum number of operating pumps to form a complete model that can be solved using the improved particle swarm optimization algorithm. S4 employs an improved particle swarm optimization algorithm that integrates nonlinear dynamic inertia weights, adaptive learning factors, and chaotic perturbation mechanisms to solve the complete optimization model. After the algorithm converges, it outputs the set of control parameters corresponding to the global optimal solution, thereby achieving the energy consumption optimization control objective of the pumping station through the set of control parameters.

[0010] Optionally, the data interaction via a dual-thread engine and shared memory further includes: The dual-thread engine is a collaborative optimization engine, which has two parallel sub-threads built in: a long short-term memory neural network prediction thread and an improved particle swarm optimization thread. The two use shared memory as the data interaction carrier to build a data transmission channel. The long short-term memory neural network prediction thread runs independently according to a set period, and writes the water demand prediction results obtained from each run into the prediction water quantity cache in the shared memory in real time, so as to realize the fixed-point storage of prediction data. After the improved particle swarm optimization thread is triggered, it first retrieves the latest water demand prediction data from the predicted water volume buffer, substitutes the water demand prediction data as dynamic core parameters into the constraint system of the optimization model, and then starts the solution process of the complete optimization model, forming a collaborative basic data flow in which the prediction thread and the optimization thread work together.

[0011] Optionally, the step of collecting historical operating data of the pumping station and performing feature engineering processing, inputting the processed feature data into a pre-trained two-layer long short-term memory neural network model according to the standard format of the collaborative basic data stream, and outputting a predicted value of future short-term water demand after model calculation, further includes: Historical water volume data and basic equipment operation parameters of pumping stations were collected, and derived features such as time period coding, weekday and weekend markings, and temperature grading were introduced to construct a multi-dimensional raw dataset covering basic operation data and environmental influencing factors. The original dataset is preprocessed to remove outliers and invalid data, then normalized, and a sliding window sequence is constructed to transform the time series data into feature samples that meet the input requirements of a two-layer long short-term memory neural network. The feature samples are divided into training set, validation set and test set. The training set is used to complete the iterative training of the two-layer long short-term memory neural network model. The validation set is used to adjust the model parameters to suppress overfitting. The test set is used to verify the prediction accuracy of the model. The processed feature data is input into the validated two-layer long short-term memory neural network model. The model then calculates and outputs a predicted value of the future short-term water demand, which is then written into the predicted water demand buffer.

[0012] Optionally, the construction of the optimization model aimed at minimizing the total operating energy consumption of the pumping station, using the predicted short-term water demand as the basis for setting the dynamic upper and lower limits of flow constraints, and combining the constraints of pipeline pressure, pump speed, and maximum number of operating pumps, to form a complete model that can be solved using the improved particle swarm optimization algorithm, further includes: With minimizing the total energy consumption of the pumping station throughout the entire operating period as the core optimization objective, an objective function is established. in Let be the shaft power of the i-th water pump. Let be the rotational speed of the i-th water pump. Let i be the flow rate of the i-th pump. For runtime, This is the pump start / stop penalty coefficient. These are the parameters for the start and stop status of the water pump; Based on the performance curves of water pump manufacturers, the formula for calculating the shaft power of a single pump is obtained through a fitting method. ,in, H represents the pump operating frequency, and H represents the pump head. For pump operating efficiency, the total operating power of the pump station is the sum of the shaft power of each operating pump; Based on water demand forecast Based on this, a dynamic flow constraint of 0.95 is set. ≤ ≤1.05 ,in The total water supply of the pumping station is taken into account, and the safety range of pipeline pressure, the safety range of pump speed, and the maximum number of pumps in operation are also included as hard constraints. This forms a complete constraint system that takes into account both dynamic adaptation and safe operation, and the model is solved.

[0013] Optionally, the process of determining the nonlinear dynamic inertia weight of the improved particle swarm optimization algorithm includes: Define the nonlinear dynamic inertia weight parameter and determine the calculation method. The calculation formula is as follows: ,in The inertia weight for the current iteration. For maximum inertia weight, The minimum inertia weight is given by t, where t is the current iteration number. The maximum number of iterations is preset. Among them, the nonlinear dynamic inertia weight changes nonlinearly with the algorithm iteration process. In the early stage of iteration, a larger ω value is configured to enhance the global search capability of the algorithm and expand the optimization range. In the later stage of iteration, the ω value is gradually reduced to focus on the current optimal region and improve the local search accuracy. The nonlinear change achieves a dynamic balance between global exploration and local optimization of the algorithm, adapting to the optimization needs of the improved particle swarm optimization algorithm at different iteration stages.

[0014] Optionally, the process of determining the adaptive learning factor of the improved particle swarm optimization algorithm includes: Setting individual cognitive factors Social learning factors This represents a reverse dynamic change trend. Initially, a large value is chosen, and then gradually decreased as the algorithm iterates. Initially, a small value is taken, and it is gradually increased as the algorithm iterates. Among them, the dynamic adjustment of the adaptive learning factor and the parameter change of the nonlinear dynamic inertia weight are implemented simultaneously. The coordinated adjustment of the two ensures that the algorithm always has the appropriate optimization performance throughout the entire iteration cycle.

[0015] Optionally, the chaotic perturbation mechanism of the improved particle swarm optimization algorithm further includes: In the later stages of algorithm iteration, the update status of the global optimal solution is monitored in real time. When it is detected that the global optimal solution has not changed for several consecutive generations and the algorithm has fallen into a local convergence stagnation state, the chaotic perturbation mechanism is triggered. Logistic chaotic mapping is introduced to generate chaotic sequences. Based on the chaotic sequences, a small-amplitude random perturbation is applied to the pump operating frequency parameter of the globally optimal particle position vector to break the local convergence state of the algorithm. After applying chaotic perturbation, the algorithm continues to perform optimization calculations according to the preset iteration rules. Taking the globally optimal particle position after perturbation as the new starting point, the algorithm performs a new optimization operation combining global and local factors until the algorithm meets the convergence criteria.

[0016] To achieve the above objectives, a second aspect of the present invention provides a collaborative device for predicting and optimizing parameters for pump station energy efficiency, comprising: The framework construction module is used to build a prediction-optimization collaborative framework, determine the collaborative relationship between the two-layer long short-term memory neural network water volume prediction model and the improved particle swarm optimization algorithm, realize data interaction through a dual-thread engine and shared memory, and use the output results of the prediction model as the dynamic setting value of the algorithm flow constraint to form a collaborative basic data flow. The water volume prediction module is used to collect historical operating data of the pumping station and perform feature engineering processing. The processed feature data is input into the pre-trained two-layer long short-term memory neural network model according to the standard format of the collaborative basic data stream. The model calculates and outputs the predicted value of the future short-term water demand. The energy consumption modeling module is used to build an optimization model with the goal of minimizing the total operating energy consumption of the pumping station. It uses the predicted value of the short-term water demand in the future as the basis for setting the dynamic upper and lower limits of the flow constraint. Combined with the constraints of pipeline pressure, pump speed and maximum number of operating pumps, it forms a complete model that can be solved using the improved particle swarm optimization algorithm. The optimization solution module is used to solve the complete optimization model by employing an improved particle swarm optimization algorithm that integrates nonlinear dynamic inertia weights, adaptive learning factors, and chaotic perturbation mechanisms. After the algorithm converges, it outputs the set of control parameters corresponding to the global optimal solution, and achieves the energy consumption optimization control objective of the pump station through the set of control parameters.

[0017] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0018] To achieve the above objectives, a third aspect of this application provides an electronic device, including a processor and a memory; wherein the processor reads executable program code stored in the memory to run a program corresponding to the executable program code, for implementing a collaborative method for predicting and optimizing parameters for pump station energy efficiency as described in the first aspect embodiment.

[0019] To achieve the above objectives, the fourth aspect of this application proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a collaborative method for predicting and optimizing parameters for pump station energy efficiency as described in the first aspect embodiment.

[0020] The embodiments of the present invention have the following beneficial effects: 1. Construct a dynamic linkage mechanism of "prediction-driven optimization" and embed the prediction results of long short-term memory neural network into the particle swarm optimization process throughout the process. This enables the optimization target to accurately match future water demand, greatly improves the foresight and economy of the scheduling scheme, and effectively avoids inefficient operation and redundant energy consumption caused by insufficient demand prediction.

[0021] 2. The improved particle swarm optimization algorithm overcomes the shortcomings of traditional algorithms, such as premature convergence and weak local search capabilities, while balancing solution speed and computational accuracy. Experimental results show that its convergence iterations are reduced by approximately 38% compared to traditional algorithms, optimal energy consumption is reduced by 8.7%, and overshoot is reduced from 5.2% to 2.1%, making it suitable for the rapidly changing operating conditions of pumping stations.

[0022] 3. The optimized model fully incorporates the physical and engineering constraints of pump station operation, closely matching actual operating scenarios. The output control parameters can be directly connected to the underlying control system, bridging the entire link from algorithm design to engineering application, and providing reliable technical support for the energy-efficient operation of pump stations. Attached Figure Description

[0023] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 A flowchart illustrating a collaborative method for predicting and optimizing parameters for pump station energy efficiency, provided in an embodiment of the present invention. Figure 2 This is a structural diagram of a collaborative device for predicting and optimizing parameters for pump station energy efficiency, provided in an embodiment of the present invention. Detailed Implementation

[0024] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0025] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0026] The following describes, with reference to the accompanying drawings, a method and apparatus for predicting and optimizing parameters for pump station energy efficiency according to an embodiment of the present invention.

[0027] Example 1 This invention provides a collaborative method for predicting and optimizing parameters in pump station energy efficiency. Figure 1 This is a flowchart illustrating a collaborative method for predicting and optimizing parameters in pump station energy efficiency, provided in an embodiment of the present invention. Figure 1 As shown, the method includes the following steps: Step S1: Construct a prediction-optimization collaborative framework, determine the collaborative relationship between the two-layer long short-term memory neural network water volume prediction model and the improved particle swarm optimization algorithm, realize data interaction through a dual-thread engine and shared memory, and use the output of the prediction model as the dynamic setting value of the algorithm's flow constraint to form a collaborative basic data flow.

[0028] In this embodiment, the core of constructing the collaborative framework lies in achieving deep dynamic linkage between the prediction and optimization stages, rather than a simple superposition of stages. First, it is necessary to clarify the core collaborative logic between the Long Short-Term Memory (LSTM) neural network water demand prediction model and the improved particle swarm optimization algorithm, establishing their functional positioning and data interaction relationship. Specifically, the LSTM neural network focuses on the accurate prediction of short-term future water demand. Its output of short-term predicted water demand is not a fixed parameter input, but rather a key dynamic setting value for the flow constraint in the optimization problem solved by the improved particle swarm optimization algorithm. This setting transforms the optimization problem from being oriented towards historical or current operating conditions to directly addressing future water demand, fundamentally establishing the core logic of prediction-driven optimization and giving the optimization scheme a forward-looking perspective.

[0029] Meanwhile, to ensure the stable implementation of the aforementioned collaborative logic, this embodiment of the application constructs a dual-threaded collaborative optimization engine as the software architecture carrier of the framework. Through the collaborative design of hardware and software, it provides technical support for the linkage between prediction and optimization. This collaborative optimization engine includes two independent and synchronously operating sub-threads: a Long Short-Term Memory Neural Network prediction thread and an improved particle swarm optimization thread. These two threads abandon traditional data transmission modes, using shared memory as a dedicated data interaction carrier. This interaction method effectively avoids latency and data loss issues during cross-thread data transmission, significantly improving the real-time performance and stability of data interaction, laying the foundation for the efficient operation of the collaborative framework.

[0030] Building upon this foundation, a closed-loop and efficient collaborative basic data flow is further constructed, clarifying the operational mechanisms and data interaction processes of the two main threads. The Long Short-Term Memory Neural Network (LSTM) prediction thread operates independently at fixed short-term intervals. After each prediction calculation, the generated water demand prediction result is immediately written to a pre-defined prediction water demand buffer, ensuring that the buffer data remains up-to-date. When the improved particle swarm optimization (PSO) thread is triggered, it does not use historical prediction data but instead prioritizes reading the latest water demand prediction value from the prediction water demand buffer. This value is then used as a dynamic parameter in the constraints of the optimization model before initiating the model solving process. Through the collaborative operation of shared memory and the prediction water demand buffer, the data interaction and model solving processes of the LSM and the improved PSO algorithm are linked, forming a closed-loop and efficient collaborative basic data flow. This provides a stable and reliable technical implementation platform for the collaboration between prediction and optimization, ensuring the overall consistency and accuracy of the framework's operation.

[0031] Step S2: Collect historical operating data of the pumping station and perform feature engineering processing. Input the processed feature data into the pre-trained two-layer long short-term memory neural network model according to the standard format of the collaborative basic data stream. The model will then output the predicted value of the future short-term water demand.

[0032] In this embodiment of the application, the accuracy of water demand prediction directly determines the solution effect of the subsequent optimization model. Therefore, the data acquisition and processing stage must take into account comprehensiveness, effectiveness and adaptability to ensure that high-quality data is input into the model.

[0033] The data acquisition process breaks away from the limitations of single-dimensional data collection, moving beyond simply collecting historical water volume data from pumping stations to constructing a multi-dimensional data acquisition system. In addition to core historical water volume time-series data, it simultaneously collects basic operational parameters of the pumping station equipment, covering parameters directly related to pumping station operation such as pump speed, head, efficiency, and pipeline pressure. Furthermore, it fully considers the impact of environmental and temporal factors on water demand, introducing derived features such as time-period coding, weekday and weekend markings, and temperature grading. Time-period coding corresponds to differences in water usage patterns across different time periods, weekday and weekend markings distinguish water demand characteristics for different date types, and temperature grading adapts to the fluctuations in water demand caused by seasonal and temperature changes. Through the integrated collection of multi-dimensional data, a multi-dimensional raw dataset covering basic operational data, equipment condition data, and environmental influencing factors is constructed. This allows the model input data to better reflect the complex operating conditions of the actual pumping station, laying a data foundation for improving subsequent prediction accuracy.

[0034] The feature engineering stage, serving as the link between data and the model, aims to transform the raw dataset into standardized feature samples adapted to the input requirements of a two-layer long short-term memory neural network. First, the constructed multi-dimensional raw dataset undergoes preprocessing. Outlier detection algorithms identify and remove outliers, while invalid and missing values ​​are filtered out to prevent interference with model training and ensure the validity and accuracy of the input data. Next, normalization is implemented to map feature data with different dimensions and numerical ranges to a unified interval, eliminating model training bias caused by dimensional differences and ensuring a balanced weighting of each feature's influence on model prediction. Finally, considering the processing characteristics of long short-term memory neural networks for time-series data, a sliding window sequence is constructed. The time-series data is segmented and recombined according to a preset window size and step size, transforming one-dimensional time-series data into multi-dimensional feature samples. This ensures the data format fully adapts to the input requirements of the two-layer long short-term memory neural network, completing the entire feature engineering process.

[0035] The model training and prediction phase focuses on the training optimization and predictive application of a two-layer long short-term memory neural network. In this embodiment, the processed feature samples are divided into a training set, a validation set, and a test set in a 7:2:1 ratio, each with its own function and working together to ensure model performance. The training set is used for iterative training of the model, continuously adjusting network weights and bias parameters to allow the model to gradually learn the mapping relationship between multi-dimensional features and water demand. The validation set is used for parameter tuning during training, monitoring the model's training effect in real time, and suppressing overfitting by adjusting hyperparameters such as the number of network layers and hidden units to ensure the model has good generalization ability. The test set is used for the final verification of the model's prediction accuracy, evaluating the model's actual prediction performance through quantitative indicators. Testing has verified that the two-layer long short-term memory neural network model in this embodiment has a mean absolute error of 2.1 cubic meters per hour, a root mean square error of 2.8 cubic meters per hour, and a prediction accuracy of 95.3%, significantly outperforming traditional time series prediction models and demonstrating accurate short-term water demand prediction capabilities. In practical forecasting applications, the preprocessed feature data that meets the input requirements is input into a pre-trained two-layer long short-term memory neural network model. The model can then quickly output the predicted water demand for the next short period, providing reliable and real-time dynamic data support for the subsequent construction and solution of the optimization model.

[0036] Step S3: Construct an optimization model with the goal of minimizing the total operating energy consumption of the pumping station. Use the predicted value of the short-term water demand in the future as the basis for setting the dynamic upper and lower limits of the flow constraint. Combine the constraints of pipeline pressure, pump speed and maximum number of operating pumps to form a complete model that can be solved using the improved particle swarm optimization algorithm.

[0037] In this embodiment, the optimization model is the core carrier for realizing the energy consumption optimization of the pumping station. Its construction process needs to take into account both goal orientation and engineering practicality. It should not only take the minimization of energy consumption as the core anchor for the optimization direction, but also fully conform to the physical characteristics and engineering specifications of the actual operation of the pumping station. Through the quantified objective function and multi-dimensional constraint system, it provides a standardized and computable model foundation for the accurate solution of the subsequent improved particle swarm optimization algorithm, ensuring that the model solution results can directly guide the actual operation and control of the pumping station.

[0038] Firstly, focusing on the core optimization objective of minimizing the total operating energy consumption of the pumping station, a refined and quantitative construction of the energy consumption objective function is carried out, clarifying the constituent elements and calculation logic of the total energy consumption of the pumping station. In this embodiment, the total operating energy consumption of the pumping station is decomposed into two parts: the energy consumption of continuous pump operation and the additional energy consumption of equipment start-up and shutdown. Based on this, the objective function is established as follows: This achieves comprehensive coverage and precise quantification of the total energy consumption of pumping stations. Among these... This is the real-time shaft power of the i-th pump, a parameter related to the pump speed. Actual output flow It exhibits a non-linear correlation, directly reflecting the operating energy consumption level of a single water pump; This is the actual running time of the water pump, used to calculate the cumulative operating energy consumption within a unit cycle; The pump start-stop penalty coefficient is calibrated based on the mechanical characteristics and measured energy consumption data of the pumping station equipment. It is used to quantify the additional energy consumption and mechanical losses caused by frequent pump start-stop, and to avoid excessive start-stop of equipment in order to reduce short-term operating energy consumption during the optimization process. These are parameters for the start / stop status of the water pumps, used to identify the running or stopped state of a single water pump. They are used to achieve shaft power... In this embodiment of the application, the shaft power calculation formula for a single water pump is obtained by data fitting based on the original performance curves provided by various water pump manufacturers. The mapping relationship between shaft power and key parameters such as water pump operating frequency, output flow rate, head, and operating efficiency is clarified. The total operating power of the pump station is obtained by accumulating the shaft power of all water pumps in operation. Finally, the energy consumption objective function is quantitatively constructed, so that the energy consumption optimization of the pump station has a clear and calculable quantitative objective.

[0039] Secondly, a multi-dimensional constraint system is established that balances dynamic adaptability and operational safety. This system defines scientific solution boundaries for the optimization model, ensuring that the solution results both meet future water demand and comply with the strict operational specifications of the pumping station. In this embodiment, the constraint system is built around dynamic flow constraints, combined with multiple strict operational constraints, achieving dual protection of dynamic adaptability and safe operation. The dynamic flow constraints are based on the real-time water demand prediction output in step S2. Based on this, a specific constraint range is set to 0.95. ≤ ≤1.05 , For the total water supply of the pumping station, this constraint allows the flow constraint threshold to be dynamically adjusted according to the real-time update of the water demand forecast. This breaks the limitations of static flow constraints in traditional optimization models, transforming the optimization model from a static model oriented towards fixed operating conditions into a dynamic model that can adapt to changes in future water demand, effectively improving the foresight and practical adaptability of the optimization scheme. Simultaneously, to ensure the safe and stable operation of the pumping station and water supply network, several rigid engineering constraints are incorporated into this system: a safe pressure range constraint for the water supply network, strictly limiting the upper and lower limits of the water supply network pressure to prevent excessive pressure from causing network damage and leakage, and also preventing excessively low pressure from affecting the quality of end-point water supply; a safe pump speed range constraint, limiting the pump speed to the rated safe range of the equipment to prevent mechanical failure of the pumps and reduce equipment lifespan due to overspeed operation; and a maximum number of operating pumps constraint, setting an upper limit on the number of pumps that can operate simultaneously based on the equipment configuration, power supply load, and operation and maintenance specifications of the pumping station, avoiding safety hazards caused by equipment overload operation. By organically combining dynamic flow constraints with multiple hard operational constraints, a complete constraint system is constructed that takes into account dynamic adaptation, safe operation, and energy consumption optimization, thus defining a scientific and reasonable boundary for solving the optimization model.

[0040] Finally, based on the constructed energy consumption objective function and multi-dimensional constraint system, comprehensive and detailed preparation work for solving the optimization model is carried out to ensure that the model can directly adapt to the solution requirements of the subsequent improved particle swarm optimization algorithm. In this embodiment, the solution preparation work revolves around the standardization and rationality of the model, specifically including three core contents: First, the objective function and all constraints are standardized, and the measurement units, calculation formats and data dimensions of all parameters are unified to ensure that various parameters can be directly substituted into the improved particle swarm optimization algorithm for solution, avoiding solution errors caused by inconsistent formats and units; Second, the nonlinear coupling relationship between various parameters is sorted out, the influence weight of different constraints on the energy consumption objective function is clarified, the constraint conflict problems that may occur in the solution process are predicted in advance, and corresponding conflict coordination rules are formulated to ensure the smoothness of the algorithm solution process; Third, the overall feasibility and rationality of the optimization model are verified by simulating and calculating by substituting historical operating data of the pumping station to verify whether the model can accurately reflect the multivariable, nonlinear and strongly coupled characteristics of pumping station operation, ensuring that the model fits the actual engineering application scenario of the pumping station. Through the above series of preparatory work, the optimization model has a clear solution logic, a standardized input format, and reliable engineering applicability, laying a solid foundation for the subsequent improvement of the particle swarm optimization algorithm to achieve efficient and accurate solutions, and ensuring the smoothness of the entire optimization process and the reliability and applicability of the final optimization results.

[0041] Step S4: An improved particle swarm optimization algorithm that integrates nonlinear dynamic inertia weights, adaptive learning factors, and chaotic perturbation mechanisms is used to solve the complete optimization model. After the algorithm converges, the set of control parameters corresponding to the global optimal solution is output. The energy consumption optimization control objective of the pump station is achieved through the set of control parameters.

[0042] In this embodiment, the improved particle swarm optimization algorithm is the core means to achieve accurate solution of the optimization model and output effective control parameters. It makes up for the technical defects of the traditional particle swarm optimization algorithm through three core improvement strategies, takes into account the global search capability and local search accuracy of the algorithm, and ensures the dynamic matching of the solution process with the future water demand, so as to ensure that the solution results can fit the actual operation needs of the pumping station, and ultimately achieve the core goal of energy consumption optimization.

[0043] First, particle encoding and initialization are performed to set the basic parameters and initial state for the algorithm, laying the foundation for subsequent iterative optimization. In this embodiment, based on the actual needs of pump station optimization scheduling, basic algorithm parameters such as particle swarm size, maximum number of iterations, and parameter search range are reasonably set to ensure that the parameter settings are adapted to the multivariable and nonlinear optimization solution scenario of the pump station. Then, the positions and velocities of all particles in the particle swarm are randomly initialized to ensure that the particles are evenly distributed within the preset search range, guaranteeing the comprehensiveness of the initial search of the algorithm. Simultaneously, multidimensional vector encoding of the particles is performed to integrate all control parameters required for pump station optimization scheduling into a multidimensional vector. This vector includes pump station control mode encoding, pump start / stop status parameters, pump operating frequency parameters, and PID controller adjustment parameters, realizing integrated encoding of all optimization control parameters. This allows a single particle to correspond to a complete pump station operation control scheme, completing the initial parameter setting for the entire algorithm solution process.

[0044] Next, fitness values ​​are calculated to construct an evaluation standard for algorithm iterative optimization, while simultaneously achieving dynamic linkage between predicted water demand and algorithm solutions. In this embodiment, the fitness value is the core indicator for evaluating the quality of the control scheme corresponding to the particle, and the calculation process must consider both energy consumption quantification and constraint compliance. First, the initialized particles are decoded to restore the various control parameters corresponding to the particles; then, the decoded pump operating frequency is substituted into the pump manufacturer's performance curve, and the corresponding key operating parameters such as flow rate, head, and efficiency are estimated through the curve fitting mapping relationship. Based on these parameters, the total operating power of the pumping station is calculated. The energy consumption level of the control scheme is initially quantified. Then, the decoded parameter combination is rigorously checked to ensure it meets all constraints set in step S3, including dynamic flow constraints, pipeline pressure constraints, and pump speed constraints. A large penalty value is imposed on any particle that violates any constraint, thus using a penalty mechanism to avoid control schemes that do not conform to engineering realities. Finally, the energy consumption is determined according to Fitness= The formula for calculating the particle fitness value, along with the penalty value for violating constraints, indicates that the lower the energy consumption and the better the constraint compliance of the corresponding control scheme. Throughout the fitness value calculation process, the predicted water demand output from step S2 is used as a boundary condition updated in real time, participating in each fitness evaluation and constraint check. This ensures that the algorithm's solution is always guided by the future short-term water demand, guaranteeing a dynamic match between fitness calculation and future water demand.

[0045] Next, algorithm iteration and optimization are carried out according to three core improvement strategies to comprehensively improve the algorithm's search capability and solution accuracy, effectively overcoming the shortcomings of traditional particle swarm optimization algorithms such as premature convergence and weak local search capability. In the embodiments of this application, the three improvement strategies work synergistically to achieve an organic combination of global exploration and local refinement. The nonlinear dynamic inertia weight strategy is based on... The formula is used for real-time parameter calculation, where t is the current iteration number. To determine the maximum number of iterations, a larger number is configured in the early stages of the iteration. The value is increased to enhance the algorithm's global search capability, allowing particles to find the optimal solution region within a broad search range; it is gradually reduced in the later stages of iteration. This value reduces the global search step size of particles, improves the local search accuracy of the algorithm, and allows particles to perform a refined search within the optimal solution region. The adaptive learning factor strategy specifically adjusts individual cognitive factors. With social learning factors The numerical variation pattern of individual cognitive factors The initial value is set to a relatively large value and gradually decreased as the iteration progresses, allowing particles to focus more on their own search experience in the early stages of iteration, thus ensuring the diversity of the search; the social learning factor is set to... The algorithm is initially set to a small value and gradually increased during iteration. This allows particles to focus more on the swarm optimization experience in the later stages of iteration, accelerating the algorithm's convergence speed and achieving a smooth transition from global exploration to local utilization. The chaotic perturbation mechanism specifically addresses the problem of premature convergence. In the later stages of iteration, it monitors the update status of the global optimum in real time. When it detects that the global optimum has remained unchanged for several generations, indicating that the algorithm is stuck in a local convergence stagnation state, it immediately introduces a Logistic chaotic mapping to generate a chaotic sequence with randomness and ergodicity. This sequence applies small-amplitude random perturbations to the key control parameters of the global optimum particle position vector, breaking the algorithm's local convergence state and allowing particles to escape the local optimum region and continue searching, ensuring that the algorithm eventually converges to the global optimum.

[0046] Finally, the algorithm convergence determination and optimal control parameter output are completed, transforming the algorithm solution into a directly implementable pump station operation control scheme to achieve the pump station energy consumption optimization control objective. In this embodiment, a dual convergence determination condition is set: when the algorithm iteration count reaches the preset maximum iteration count, or the change in the global optimal solution over multiple consecutive generations is less than a preset error threshold, the algorithm is determined to have converged, and the iterative optimization process is terminated. Subsequently, the globally optimal particle obtained after convergence is decoded to restore the complete control parameter set corresponding to the particle. This parameter set includes the pump station operation control mode, the start / stop status of each pump, the speed setpoint of each operating pump, and the Kp, Ti, and Td adjustment parameters of the PID controller. All parameters are standardized engineering control parameters that can be directly connected to the pump station's underlying control system. After this set of control parameters is transmitted to the pump station's execution system, the execution system can perform precise control of the pump station equipment based on the parameters, allowing the pump station to operate according to the energy-optimized control scheme, ultimately achieving the core objective of pump station energy consumption optimization control, while ensuring the safety and stability of pump station operation.

[0047] Example 2 This invention provides a collaborative device for predicting and optimizing parameters for pump station energy efficiency. Figure 2 This is a schematic diagram of a collaborative device for predicting and optimizing parameters for pump station energy efficiency, provided in an embodiment of the present invention. Figure 2 As shown, the device includes: The framework construction module 100 is used to build a prediction-optimization collaborative framework, clarify the collaborative relationship between the two-layer long short-term memory neural network water volume prediction model and the improved particle swarm optimization algorithm, realize data interaction through a dual-thread engine and shared memory, and use the output results of the prediction model as the dynamic setting value of the algorithm flow constraint to form a collaborative basic data flow. The water volume prediction module 200 is used to collect historical operating data of the pumping station and perform feature engineering processing. The processed feature data is input into the pre-trained two-layer long short-term memory neural network model according to the standard format of the collaborative basic data stream. The model calculates and outputs the predicted value of the future short-term water demand. The energy consumption modeling module 300 is used to build an optimization model with the goal of minimizing the total operating energy consumption of the pumping station. It uses the predicted value of the short-term water demand in the future as the basis for setting the dynamic upper and lower limits of the flow constraint. Combined with the constraints of pipeline pressure, pump speed and maximum number of operating pumps, it forms a complete model that can be solved using the improved particle swarm optimization algorithm. The optimization solution module 400 is used to solve the complete optimization model by adopting an improved particle swarm optimization algorithm that integrates nonlinear dynamic inertia weights, adaptive learning factors and chaotic perturbation mechanisms. After the algorithm converges, it outputs the set of control parameters corresponding to the global optimal solution, and achieves the energy consumption optimization control target of the pump station through the set of control parameters.

[0048] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0049] Example 3 To implement the methods of the above embodiments, the present invention also provides an electronic device, which includes a memory and a processor; wherein the processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement the various steps of the methods described above.

[0050] Example 4 To implement the above embodiments, this application also proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method described in the foregoing embodiments.

[0051] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0052] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0053] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

Claims

1. A method for coordinated prediction and parameter optimization of pump station energy efficiency, characterized in that, include: S1. Construct a prediction-optimization collaborative framework, determine the collaborative relationship between the two-layer long short-term memory neural network water volume prediction model and the improved particle swarm optimization algorithm, realize data interaction through a dual-thread engine and shared memory, and use the output of the prediction model as the dynamic setting value of the algorithm flow constraint to form a collaborative basic data flow. S2 collects historical operating data of the pumping station and performs feature engineering processing. The processed feature data is then input into a pre-trained two-layer long short-term memory neural network model according to the standard format of the collaborative basic data stream. The model then calculates and outputs the predicted value of the short-term water demand in the future. S3. Construct an optimization model with the goal of minimizing the total operating energy consumption of the pumping station. Use the predicted value of the short-term water demand in the future as the basis for setting the dynamic upper and lower limits of the flow constraint. Combine the constraints of pipeline pressure, pump speed and maximum number of operating pumps to form a complete model that can be solved using the improved particle swarm optimization algorithm. S4 employs an improved particle swarm optimization algorithm that integrates nonlinear dynamic inertia weights, adaptive learning factors, and chaotic perturbation mechanisms to solve the complete optimization model. After the algorithm converges, it outputs the set of control parameters corresponding to the global optimal solution, thereby achieving the energy consumption optimization control objective of the pumping station through the set of control parameters.

2. The method according to claim 1, characterized in that, The data interaction implemented via a dual-thread engine and shared memory also includes: The dual-thread engine is a collaborative optimization engine, which has two parallel sub-threads built in: a long short-term memory neural network prediction thread and an improved particle swarm optimization thread. The two use shared memory as the data interaction carrier to build a data transmission channel. The long short-term memory neural network prediction thread runs independently according to a set period, and writes the water demand prediction results obtained from each run into the prediction water quantity cache in the shared memory in real time, so as to realize the fixed-point storage of prediction data. After the improved particle swarm optimization thread is triggered, it first retrieves the latest water demand prediction data from the predicted water volume buffer, substitutes the water demand prediction data as dynamic core parameters into the constraint system of the optimization model, and then starts the solution process of the complete optimization model, forming a collaborative basic data flow in which the prediction thread and the optimization thread work together.

3. The method according to claim 2, characterized in that, The process of collecting historical operating data from pumping stations and performing feature engineering processing, then inputting the processed feature data into a pre-trained two-layer long short-term memory neural network model according to the standard format of collaborative basic data streams, and outputting a predicted value of future short-term water demand after model computation, also includes: Historical water volume data and basic equipment operation parameters of pumping stations were collected, and derived features such as time period coding, weekday and weekend markings, and temperature grading were introduced to construct a multi-dimensional raw dataset covering basic operation data and environmental influencing factors. The original dataset is preprocessed to remove outliers and invalid data, then normalized, and a sliding window sequence is constructed to transform the time series data into feature samples that meet the input requirements of a two-layer long short-term memory neural network. The feature samples are divided into training set, validation set and test set. The training set is used to complete the iterative training of the two-layer long short-term memory neural network model. The validation set is used to adjust the model parameters to suppress overfitting. The test set is used to verify the prediction accuracy of the model. The processed feature data is input into the validated two-layer long short-term memory neural network model. The model then calculates and outputs a predicted value of the future short-term water demand, which is then written into the predicted water demand buffer.

4. The method according to claim 3, characterized in that, The optimization model constructed with the goal of minimizing the total operating energy consumption of the pumping station uses the predicted short-term water demand as the basis for setting the dynamic upper and lower limits of flow constraints. Combined with constraints on pipeline pressure, pump speed, and the maximum number of operating pumps, a complete model capable of being solved using the improved particle swarm optimization algorithm is formed. This also includes: With minimizing the total energy consumption of the pumping station throughout the entire operating period as the core optimization objective, an objective function is established. in Let be the shaft power of the i-th water pump. Let be the rotational speed of the i-th water pump. Let i be the flow rate of the i-th pump. For runtime, This is the pump start / stop penalty coefficient. These are the parameters for the start and stop status of the water pump; Based on the performance curves of water pump manufacturers, the formula for calculating the shaft power of a single pump is obtained through a fitting method. ,in, H represents the pump operating frequency, and H represents the pump head. For pump operating efficiency, the total operating power of the pump station is the sum of the shaft power of each operating pump; Based on water demand forecast Based on this, a dynamic flow constraint of 0.95 is set. ≤ ≤1.05 ,in The total water supply of the pumping station is taken into account, and the safety range of pipeline pressure, the safety range of pump speed, and the maximum number of pumps in operation are also included as hard constraints. This forms a complete constraint system that takes into account both dynamic adaptation and safe operation, and the model is solved.

5. The method according to claim 4, characterized in that, The process of determining the nonlinear dynamic inertia weight in the improved particle swarm optimization algorithm includes: Define the nonlinear dynamic inertia weight parameter and determine the calculation method. The calculation formula is as follows: ,in The inertia weight for the current iteration. For maximum inertia weight, The minimum inertia weight is given by t, where t is the current iteration number. The maximum number of iterations is preset. Among them, the nonlinear dynamic inertia weight changes nonlinearly with the algorithm iteration process, and is configured with a larger weight in the early stage of iteration. The value is increased to enhance the algorithm's global search capability, expand the optimization range, and is gradually reduced in the later stages of iteration. The value focuses on improving the local search accuracy by targeting the current optimal region. It achieves a dynamic balance between global exploration and local optimization through nonlinear changes, adapting to the optimization needs of the improved particle swarm optimization algorithm at different iteration stages.

6. The method according to claim 5, characterized in that, The process of determining the adaptive learning factor in the improved particle swarm optimization algorithm includes: Setting individual cognitive factors With social learning factors It represents a reverse dynamic change trend. Initially, a large value is chosen, and then gradually decreased as the algorithm iterates. Initially, a small value is taken, and it is gradually increased as the algorithm iterates. Among them, the dynamic adjustment of the adaptive learning factor and the parameter change of the nonlinear dynamic inertia weight are implemented simultaneously. The coordinated adjustment of the two ensures that the algorithm always has the appropriate optimization performance throughout the entire iteration cycle.

7. The method according to claim 6, characterized in that, The chaotic perturbation mechanism of the improved particle swarm optimization algorithm also includes: In the later stages of algorithm iteration, the update status of the global optimal solution is monitored in real time. When it is detected that the global optimal solution has not changed for several consecutive generations and the algorithm has fallen into a local convergence stagnation state, the chaotic perturbation mechanism is triggered. Logistic chaotic mapping is introduced to generate chaotic sequences. Based on the chaotic sequences, a small-amplitude random perturbation is applied to the pump operating frequency parameter of the globally optimal particle position vector to break the local convergence state of the algorithm. After applying chaotic perturbation, the algorithm continues to perform optimization calculations according to the preset iteration rules. Taking the globally optimal particle position after perturbation as the new starting point, the algorithm performs a new optimization operation combining global and local factors until the algorithm meets the convergence criteria.

8. A collaborative device for predicting and optimizing parameters for pump station energy efficiency, characterized in that, include: The framework construction module is used to build a prediction-optimization collaborative framework, determine the collaborative relationship between the two-layer long short-term memory neural network water volume prediction model and the improved particle swarm optimization algorithm, realize data interaction through a dual-thread engine and shared memory, and use the output results of the prediction model as the dynamic setting value of the algorithm flow constraint to form a collaborative basic data flow. The water volume prediction module is used to collect historical operating data of the pumping station and perform feature engineering processing. The processed feature data is input into the pre-trained two-layer long short-term memory neural network model according to the standard format of the collaborative basic data stream. The model calculates and outputs the predicted value of the future short-term water demand. The energy consumption modeling module is used to build an optimization model with the goal of minimizing the total operating energy consumption of the pumping station. It uses the predicted value of the short-term water demand in the future as the basis for setting the dynamic upper and lower limits of the flow constraint. Combined with the constraints of pipeline pressure, pump speed and maximum number of operating pumps, it forms a complete model that can be solved using the improved particle swarm optimization algorithm. The optimization solution module is used to solve the complete optimization model by employing an improved particle swarm optimization algorithm that integrates nonlinear dynamic inertia weights, adaptive learning factors, and chaotic perturbation mechanisms. After the algorithm converges, it outputs the set of control parameters corresponding to the global optimal solution, and achieves the energy consumption optimization control objective of the pump station through the set of control parameters.

9. An electronic device, characterized in that, Including processor and memory; The processor runs a program corresponding to the executable program code stored in the memory to implement the method as described in any one of claims 1-7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-7.

Images