A Method and System for Water Supply Plant Scheduling Based on Multi-Objective Collaborative Optimization

Abstract

This invention relates to the field of urban water supply scheduling and energy-saving control technology, and discloses a water supply plant scheduling method and system based on multi-objective collaborative optimization. The method includes: acquiring real-time influencing factor data and historical water consumption sequences, performing multi-source data nonlinear fitting prediction to obtain a predicted water supply distribution range; performing random scenario simulation and deviation quantification analysis based on the distribution range and the current pump configuration to obtain a water supply deviation probability distribution; if the deviation exceeds the limit, retrieving alternative solutions based on the distribution range and evaluating energy consumption to obtain an energy consumption ranking list; performing dynamic constraint iterative optimization combined with frequency constraints to obtain an optimized pump control sequence; fine-tuning the speed of the control sequence based on the flow matching degree to obtain the final control command; and using actual water supply data for adaptive feedback updates and simulation verification. This method can achieve the dual objectives of energy consumption reduction and precise water supply matching.

Description

Technical Field

[0001] This invention relates to the field of urban water supply scheduling and energy-saving control technology, and in particular to a water supply plant scheduling method and system based on multi-objective collaborative optimization. Background Technology

[0002] With the continuous expansion of urbanization, water demand is subject to dramatic fluctuations and high uncertainty due to the interaction of multiple factors such as meteorological conditions, holiday effects, and unforeseen events. Against this backdrop, fully utilizing industrial big data technology to deeply mine and analyze massive amounts of historical water supply data and real-time environmental factors has become a key trend for achieving precise scheduling of water supply plants, reducing pump station energy consumption, and improving management levels.

[0003] In a current technology, water supply scheduling methods primarily rely on historical experience or short-term forecasts based on a single linear model to formulate operational plans. In practice, dispatchers typically estimate future fixed flow rates based on historical data from the same period and directly set the number of pumps to start / stop or variable frequency parameters, attempting to address future water demand with a static control strategy. However, this method ignores the nonlinear characteristics of future water consumption changes influenced by multiple factors, making it difficult for scheduling plans to adapt to actual dynamic demand. This discrepancy between prediction and reality forces dispatchers to frequently adjust pump status during operation, resulting not only in water and energy waste due to oversupply or user disruption due to insufficient supply, but also significantly increased energy consumption and equipment wear due to rapid switching and frequent start-stop of pump conditions. This uncertainty directly impacts the optimization process, making it exceptionally difficult to find the pump combination with the lowest energy consumption while meeting flow boundary conditions.

[0004] In summary, existing technologies have limitations in achieving both precise flow matching and minimum energy consumption for pumping stations. Summary of the Invention

[0005] This invention provides a water supply plant scheduling method and system based on multi-objective collaborative optimization to solve the problem of difficulty in achieving accurate flow matching and minimum pump station energy consumption.

[0006] Firstly, to address the aforementioned technical problems, this invention provides a water supply plant scheduling method based on multi-objective collaborative optimization, comprising:

[0007] Real-time influencing factor data of the water supply system is obtained, and combined with the pre-stored historical water consumption sequence, multi-source data nonlinear fitting prediction is performed on the real-time influencing factor data and the historical water consumption sequence to obtain the predicted distribution range of water supply.

[0008] Based on the predicted water supply distribution range and the current configuration of the pump unit operating status collected in real time, random scenario simulation and deviation quantification analysis are performed to obtain the probability distribution of water supply deviation.

[0009] According to the water supply deviation probability distribution, if the deviation probability exceeds the preset threshold, then alternative schemes are searched and energy consumption is evaluated based on the predicted water supply distribution range to obtain an energy consumption ranking list.

[0010] Based on the energy consumption ranking list, and combined with the preset frequency change rate constraint, a genetic algorithm is used to perform dynamic constraint iterative optimization to obtain the optimized pump group control sequence.

[0011] Based on the optimized pump group control sequence and the real-time operating data of the pump station, the flow matching degree is calculated. If the flow matching degree is lower than the preset flow matching degree threshold, the speed of the optimized pump group control sequence is fine-tuned to obtain the final control command.

[0012] The actual water supply data after the final control command is executed is collected, and the actual water supply data is used for adaptive feedback updates and simulation verification to obtain a continuously optimized pump unit operation status scheme.

[0013] Secondly, the present invention provides a water supply plant scheduling system based on multi-objective collaborative optimization, comprising:

[0014] The multi-source data prediction module is used to acquire real-time influencing factor data of the water supply system, and combine it with the pre-stored historical water consumption sequence to perform multi-source data nonlinear fitting prediction on the real-time influencing factor data and the historical water consumption sequence to obtain the predicted distribution range of water supply.

[0015] The random scenario simulation module is used to perform random scenario simulation and deviation quantification analysis based on the predicted water supply distribution range and the current configuration of the pump group operating status collected in real time, so as to obtain the probability distribution of water supply deviation.

[0016] The scheme energy consumption assessment module is used to perform alternative scheme retrieval and energy consumption assessment based on the water supply deviation probability distribution and the preset deviation probability threshold, and obtain an energy consumption ranking list.

[0017] The dynamic iterative optimization module is used to perform dynamic constraint iterative optimization based on the energy consumption sorting list and a preset frequency change rate constraint condition using a genetic algorithm to obtain the optimized pump group control sequence.

[0018] The real-time matching and adjustment module is used to calculate the flow matching degree based on the optimized pump group control sequence and the real-time operating data of the pump station. If the flow matching degree is lower than the preset flow matching degree threshold, the speed of the optimized pump group control sequence is finely adjusted to obtain the final control command.

[0019] The feedback closed-loop verification module is used to collect the actual water supply data after the execution of the final control command, and use the actual water supply data to perform adaptive feedback updates and simulation verification to obtain a continuously optimized pump group operation status scheme.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] (1) This invention obtains the predicted distribution range of water supply by integrating multi-source data and using the random forest algorithm, and combines Monte Carlo simulation to generate demand scenarios to determine the probability distribution of deviation. This method effectively handles the nonlinear relationship between factors such as weather and holidays and water consumption, quantifies the uncertainty of future water supply target flow, and identifies the risk of water supply surplus or deficit in advance. This effectively solves the problem that the traditional method relies on single numerical prediction, which makes it difficult for the scheduling scheme to adapt to the actual dynamic demand, improves the predictability of water supply scheduling, and reduces the risk of water waste or insufficient water supply.

[0022] (2) This invention uses a genetic algorithm to iteratively optimize the alternative schemes based on energy consumption ranking to generate a control sequence, and uses a frequency conversion adjustment module to fine-tune the pump group speed when the flow matching degree is low. This method can quickly screen out the pump group combination with the lowest energy consumption while meeting the dynamic flow boundary conditions, and eliminate the instantaneous flow deviation by real-time speed fine-tuning, thereby realizing the synergistic optimization of accurate flow matching and the lowest energy consumption of the pump station, avoiding the additional energy loss and equipment wear caused by frequent start-up and shutdown of pump groups or rapid switching of operating conditions, and significantly improving the economic efficiency of operation.

[0023] (3) This invention uses the actual water supply deviation data to update the training dataset of the random forest algorithm through a feedback loop mechanism, and re-simulates the improved prediction parameters to verify the energy loss. This mechanism constructs an adaptive learning loop, which enables the prediction model to continuously correct the parameters according to the feedback after the actual command is executed, and actively adapt to the impact of pump station equipment aging or changes in pipeline characteristics, thereby obtaining a continuously optimized pump group operation status scheme, ensuring that the water supply system always maintains a highly efficient, energy-saving and intelligent stable state in long-term operation. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the water supply plant scheduling method based on multi-objective collaborative optimization provided in the first embodiment of the present invention;

[0025] Figure 2 This is a schematic diagram of the water supply plant scheduling system based on multi-objective collaborative optimization provided in the second embodiment of the present invention. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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 are within the scope of protection of the present invention.

[0027] Reference Figure 1 The first embodiment of the present invention provides a water supply plant scheduling method based on multi-objective collaborative optimization, comprising the following steps:

[0028] S11, acquire real-time influencing factor data of the water supply system, combine it with the pre-stored historical water consumption sequence, perform multi-source data nonlinear fitting prediction on the real-time influencing factor data and the historical water consumption sequence, and obtain the predicted distribution range of water supply.

[0029] S12, based on the predicted water supply distribution range and the current configuration of the pump unit operating status collected in real time, perform random scenario simulation and deviation quantification analysis to obtain the probability distribution of water supply deviation;

[0030] S13. According to the water supply deviation probability distribution, if the deviation probability threshold is exceeded, then alternative schemes are searched and energy consumption is evaluated based on the water supply prediction distribution range to obtain an energy consumption ranking list.

[0031] S14. Based on the energy consumption sorting list and combined with the preset frequency change rate constraint, a genetic algorithm is used to perform dynamic constraint iterative optimization to obtain the optimized pump group control sequence.

[0032] S15. Based on the optimized pump group control sequence and the real-time operating data of the pump station, calculate the flow matching degree. If the flow matching degree is lower than the preset flow matching degree threshold, fine-tune the speed of the optimized pump group control sequence to obtain the final control command.

[0033] S16, Collect the actual water supply data after executing the final control command, and use the actual water supply data to perform adaptive feedback updates and simulation verification to obtain a continuously optimized pump group operation status scheme.

[0034] In step S11, real-time influencing factor data of the water supply system is acquired, and combined with pre-stored historical water consumption sequences, multi-source data nonlinear fitting prediction is performed on the real-time influencing factor data and the historical water consumption sequences to obtain the predicted water supply distribution range, including:

[0035] Real-time influencing factor data is collected from the multi-source data acquisition system, and pre-stored historical influencing factor data and pre-stored historical water consumption sequences are obtained.

[0036] Based on the historical influencing factor data, missing value imputation and normalization preprocessing are performed to obtain the historical feature vector matrix;

[0037] Based on the historical feature vector matrix and the historical water consumption sequence, a random forest algorithm is used for fitting and training to obtain a water supply regression prediction model.

[0038] Based on the real-time influencing factor data, a feature vector to be predicted is constructed, and the feature vector to be predicted is input into the water supply regression prediction model to obtain the output value distribution.

[0039] Based on the distribution of the output values, quantile limits are calculated to obtain the predicted distribution range of water supply.

[0040] First, data acquisition and historical database management operations are performed. Real-time influencing factor data is obtained by calling meteorological service interfaces (such as the National Meteorological Administration API) through multi-source data acquisition terminals, and pre-stored historical influencing factor data and historical water consumption sequences are read from the industrial database server deployed in the water supply dispatch center.

[0041] The data update mechanism in the industrial database server is as follows: The system backend sets midnight (00:00) as the update trigger time. The instantaneous flow data uploaded by the flow meters of each plant outlet within the previous natural day (00:00-23:59) is extracted through the SCADA system (data acquisition and monitoring control system). The hourly water supply flow value is obtained by integration calculation, which is the historical water consumption sequence. At the same time, the meteorological records and calendar information corresponding to this period are stored synchronously and added to the historical feature data table.

[0042] The historical water consumption sequence specifically includes water supply flow values ​​(unit: cubic meters / hour) stored with hourly timestamp indexes. The historical influencing factor data specifically includes the highest temperature, lowest temperature, rainfall, relative humidity, and holiday markers corresponding one-to-one with the timestamps (e.g., weekdays=0, weekends=1, holidays=2).

[0043] Secondly, data preprocessing and feature matrix construction are performed. The acquired historical influencing factor data undergoes quality checks; if missing data (i.e., null values ​​or NaN) is detected, linear interpolation is used to fill in the missing data. The specific calculation process of linear interpolation involves locating the time point where the missing value is located. Find the two closest known valid data points before and after this time point. and (in The fill value is calculated using the formula. :

[0044]

[0045] After the data is filled in, the feature data of each dimension are subjected to max-min normalization, and the calculation formula is as follows:

[0046]

[0047] in and The minimum and maximum values ​​of this feature dimension are respectively found in the historical dataset. The processed data are then combined column-wise to obtain the historical feature vector matrix. (N is the number of samples, M is the feature dimension); further, the construction of the historical feature vector matrix is ​​achieved by directly using four continuous variables—highest temperature, lowest temperature, rainfall, and relative humidity—as feature columns; for the categorical variable of holiday markers, one-hot encoding is used to convert it into three binary feature columns, representing "weekdays," "weekends," and "statutory holidays," respectively; after the above processing, the feature dimension M is fixed at 7, and all feature columns are concatenated to form the historical feature vector matrix.

[0048] Next, perform parameter setting and fitting training operations for the random forest model. The historical feature vector matrix... The historical water consumption sequence is used as the independent variable and as the dependent variable, input into the random forest algorithm. Before training, the maximum depth of the key hyperparameter tree (Max Depth, ...) is set. ) and the minimum number of leaf samples (Min Leaf Samples, The maximum depth of the tree. Defined as the longest path length from the root node to the farthest leaf node in the decision tree, used to limit the complexity of the model, and the minimum number of samples required for the leaf node. Defined as the minimum number of samples that a leaf node must contain after a split, used to prevent overfitting. The basis and general method for setting these two parameters is to use K-fold cross-validation (e.g., K=5) combined with a grid search method. The search range is [10, 50], and the step size is 5; set The search range is [1, 20], with a step size of 1. Iterate through all parameter combinations, calculate the mean squared error (MSE) on the validation set, and select the parameter combination that minimizes the MSE as the final set value. Use the set parameters and the Bootstrap sampling method to construct a dataset containing... A water supply regression prediction model using decision trees (e.g., 500 trees).

[0049] Finally, feature vector construction and distribution interval generation are performed. The collected real-time influencing factor data are arranged according to the same feature order as the historical feature vector matrix (e.g., [temperature, rainfall, humidity, holidays]), and statistically derived from historical data is used. and Substituting the parameters into the aforementioned normalization formula, the feature vector to be predicted is calculated. .Will The input is fed into the water supply regression prediction model, and each decision tree in the model is traversed. The regression output values ​​of all leaf nodes of the decision trees are counted to form an output set. The output value distribution is represented by this. The values ​​in set O are sorted in ascending order, based on a preset confidence level of 95% (this confidence level is based on a normal distribution). (Based on the established principles), the sequence index positions corresponding to the 2.5% lower quantile and the 97.5% upper quantile were calculated respectively. and Take the first element from the sorted set. The values ​​serve as the lower limit of the interval. , No. The values ​​serve as the upper limit of the interval. Thus, the predicted distribution range of water supply is obtained. .

[0050] In step S12, based on the predicted water supply distribution range and the current configuration of the pump unit operating status collected in real time, random scenario simulation and deviation quantification analysis are performed to obtain the probability distribution of water supply deviation, including:

[0051] Based on the predicted water supply distribution range and the current configuration of the pump group operating status collected in real time, a Monte Carlo simulation generator is used to perform random sampling to obtain a random demand scenario sequence containing instantaneous demand values.

[0052] Extract the instantaneous demand values ​​from the random demand scenario sequence and obtain the fixed water supply capacity value corresponding to the current configuration of the pump group's operating status;

[0053] Based on the difference between the instantaneous demand value and the fixed water supply capacity value, a difference sequence is obtained;

[0054] Based on the difference sequence, the water supply surplus and water supply deficit are identified, and kernel density estimation is performed on the water supply surplus and water supply deficit to obtain the water supply deviation probability distribution.

[0055] First, perform Monte Carlo random sampling based on the Mason twisting algorithm. Obtain the predicted water supply distribution range output from step S11. ,in The lower limit of the interval, The upper limit of the interval is set. To ensure the randomness and statistical independence of the simulation scenario, the Mason twisting algorithm (specifically the MT19937 variant) is selected as the Monte Carlo simulation generator. The number of simulation samples is set. The number of iterations is set based on the law of large numbers and the Monte Carlo convergence criterion, and is usually set to 10,000 to ensure that the statistical error is less than 0.1%. The random sampling process first uses the microsecond-level timestamp of the current system clock as the initial seed to initialize the state vector of the MT19937; then, 32-bit unsigned pseudo-random integers are generated based on a linear recursive relation. :

[0056]

[0057] in This is an XOR operation, where A is a rotation matrix. and These are the states after processing the high and low bits, respectively. Next, ... Divide by the largest possible integer value Perform max-min normalization to obtain standard uniformly distributed random numbers in the interval [0,1]. Finally, the inverse transform method is used to... Mapping to the prediction interval, the calculation formula is as follows:

[0058]

[0059] Repeat the above process This generates a sequence containing A sequence of random demand scenarios with instantaneous demand values. .

[0060] Secondly, the fixed water supply capacity value is acquired and calculated. The current configuration of the pump unit operating status, collected in real-time, is read through the SCADA system. This configuration includes the set of currently active pump unit numbers. and the current operating frequency of each pump unit. Calculate the current fixed water supply capacity based on the similarity law of water pumps. The calculation formula is:

[0061]

[0062] in Let k be the rated flow rate of the kth pump unit. This is the rated frequency (typically 50Hz) of the k-th pump unit. The value represents the instantaneous theoretical flow rate that the water supply system can provide without adjusting the current pump control strategy. Next, the difference sequence calculation and deviation type identification operations are performed. Each instantaneous demand value in the sequence of random demand scenarios is iterated through. Compare it with the fixed water supply capacity value The difference is calculated using the following formula:

[0063]

[0064] Thus, the length is obtained difference sequence Perform symbol recognition on the sequence; if... Defined as water supply surplus, this indicates that the current pump output exceeds demand; if , defined as the water supply shortfall, indicates that the current pump set output is insufficient.

[0065] Finally, kernel density estimation is performed. To obtain a continuous probability density function from the discrete difference sequence, kernel density estimation, a nonparametric statistical method, is used. A Gaussian kernel function is selected. The smoothing bandwidth parameter h is set and calculated using the following formula:

[0066]

[0067] in Let be the standard deviation of the difference sequence, and IQR be the interquartile range. The probability density function of the water supply deviation is calculated using the formula, which is the probability distribution of the water supply deviation:

[0068]

[0069] This distribution intuitively describes the probability density of water oversupply or undersupply that might result from maintaining the current pump set status within the current range of prediction uncertainty.

[0070] In step S13, based on the water supply deviation probability distribution, if the deviation exceeds a preset probability threshold, then alternative solutions are retrieved and energy consumption is assessed based on the predicted water supply distribution range to obtain an energy consumption ranking list, including:

[0071] If the probability distribution of water supply deviation exceeds the preset deviation probability threshold, alternative combination schemes are retrieved from the preset pump group control library.

[0072] Based on the predicted water supply distribution range, the data is discretized to obtain multiple discrete flow points.

[0073] Based on the multiple discrete flow points and the alternative combination schemes, the operating point is calculated by combining the preset pump set characteristic equation and the preset pipeline resistance coefficient to obtain the shaft power value.

[0074] Based on the shaft power value, a weighted integral is calculated to obtain the expected energy consumption value of each of the alternative combination schemes, and an energy consumption ranking list is generated based on the expected energy consumption value.

[0075] First, perform deviation threshold determination and alternative solution retrieval operations. Obtain the water supply deviation probability distribution output in step S12, and calculate the deviations exceeding the system's allowable deviation range (e.g., rated flow rate). The cumulative probability value of () is defined as the current risk probability. .Will Deviation probability threshold from the preset value A comparison is made. The deviation probability threshold is... The setup is based on mining the historical dispatch logs of the water supply system over the past three years, extracting the probability values ​​of water supply deviations corresponding to all events that led to pipeline pressure exceeding limits alarms or user complaints, thus constructing a fault risk distribution sample set; the 10th percentile of this sample set is selected as... (e.g., 0.15). The reason for choosing the 10th percentile instead of the median or other values ​​is that water supply scheduling must follow the principle of "zero tolerance for failure". This threshold setting means that as long as the current risk probability reaches 10% of the lowest risk level that caused the failure in history, the system will be judged as a high-risk state, thereby ensuring that more than 90% of potential hidden dangers can be covered, reflecting an extremely high degree of safety redundancy.

[0076] If the judgment If the system determines that the current default solution is no longer suitable for future uncertainties, a rescheduling mechanism is triggered, meaning the system must find a new optimal solution. At this point, the system accesses a pre-set pump control database, which stores all physically feasible start-up combinations for the pump station (e.g., "Pump 1 at mains frequency + Pump 2 at variable frequency"). Data items include combination ID, equipment list, and power distribution limit. The database's data source is the pump station design drawings and equipment ledger information. The update mechanism involves the system monitoring the SCADA system's equipment status signals in real time. When a pump's status changes to "fault" or "under maintenance," the database update logic is automatically triggered, marking all combinations involving that pump as "unavailable." When the status returns to "normal," they are remarked as "available." The system then retrieves all currently marked "available" alternative combinations from the database, forming a set of alternative combinations.

[0077] Next, the discretization and probability weight calculation operations of the prediction interval are performed. The predicted water supply distribution interval output in step S11 is obtained. To account for the weighting of flow uncertainty in energy consumption assessment, the output of the random forest model in step S11, which includes... A weighted mapping is performed on the set of predicted values. Specifically, the intervals are weighted... Divide the data into K (e.g., 20) discrete flow points at equal intervals. , with each Set a small bandwidth at the center The statistical set O falls within the interval Number of predicted values ​​within Furthermore, the setting of the micro-bandwidth is based on the distribution characteristics of the predicted value set. By calculating the standard deviation of the set, the micro-bandwidth is set to 0.05 times the standard deviation. This empirical value of 0.05 ensures that the bandwidth is small enough to distinguish the probability density of different flow points, while being large enough to avoid statistical fluctuations due to limited samples. Those skilled in the art can also set the bandwidth proportionally by dividing the difference in interval length by twice the number of discrete points, based on the prediction interval length and the number of discrete points.

[0078] Calculate the occurrence probability weight corresponding to each discrete flow point using the formula. :

[0079]

[0080] And normalization is performed to make This weight The probability density of future actual water demand falling near this discrete point was precisely quantified. Next, characteristic parameter fitting and operating point calculation operations were performed. Preset network resistance coefficients were invoked. With pump set characteristic equations. Pipeline resistance coefficient. The setting basis and fitting process are based on extracting the factory pressure data of the SCADA system within the last 7 days. , the most unfavorable point pressure And the high-frequency historical data of the total flow rate Q, substituted into the hydraulic formula:

[0081]

[0082] in, Given known terrain elevation differences, construct multiple sets. For the data pairs, a linear regression is performed using the least squares method, and the resulting slope is... The process of obtaining the pump set characteristic equation involves using performance curve test data (including multiple sets of flow-head and flow-efficiency coordinate points) provided by the pump manufacturer, and then employing a second-order polynomial fitting algorithm to fit the QH equation. and equation During the calculation, each alternative combination is iterated over. With each discrete flow point Calculate the required head based on the pipeline resistance model. .Will Substitution scheme The combined QH equations calculate the actual head that can be provided. ,like If the condition is not met at that flow rate point (i.e., the water cannot be delivered to the user), the solution is marked as invalid; if it is met, it is substituted into the following... Equation calculation running efficiency And use physical formulas to calculate the shaft power value corresponding to that point:

[0083]

[0084] in Let g be the density of water and g be the acceleration due to gravity. This refers to motor efficiency. It should be noted that the shaft power calculation described in the formula involves the unification of physical dimensions. In this embodiment, if the input discrete flow points... If we continue using the engineering unit "cubic meters per hour" from the previous steps, we need to introduce a unit conversion factor (i.e., divide by 3600) when substituting it into the formula for calculation, or convert it to the international standard unit "cubic meters per second" beforehand. In addition, the constant 1000 in the formula is used to convert the power unit from watt (W) to kilowatt (kW).

[0085] Finally, the weighted energy consumption integral and sorting list generation operations are performed. Based on the calculated shaft power values... and the corresponding probability weights Calculate each alternative combination scheme using the discrete expectation formula. Expected energy consumption:

[0086]

[0087] in This represents the scheduling cycle duration. If the scheme is marked as invalid at any high-weight traffic point, its total energy consumption is... Set it to infinity. After completing the calculation, assign all valid solutions as follows: Sort the data from smallest to largest to generate an energy consumption sorted list containing the scheme ID and expected energy consumption value.

[0088] It should be noted that the pump set characteristic equation (second-order polynomial) and pipeline resistance equation described in this embodiment provide a theoretical reference model under ideal fluid conditions. In practical engineering applications, as equipment operating time increases, impeller wear and pipe scaling can cause the actual characteristic curves to drift. Therefore, this invention supports the introduction of online correction factors based on measured data to dynamically correct the above polynomial coefficients, or directly using a black-box model trained based on neural networks (such as BP neural networks or RBF networks) to replace the above analytical equations for operating point calculation, thereby further improving the model's fitting accuracy. This replacement of the model form is a conventional technical extension of this invention.

[0089] In step S14, based on the energy consumption ranking list and combined with the preset frequency change rate constraint, a genetic algorithm is used for dynamic constraint iterative optimization to obtain the optimized pump group control sequence, including:

[0090] Based on the energy consumption ranking list, an encoding mapping is performed to obtain an initial population encoding that includes the frequency change rate;

[0091] Determine whether the frequency change rate corresponding to the initial population code satisfies the preset frequency change rate constraint condition. If it does, calculate the individual fitness value of the initial population code.

[0092] Based on the individual fitness values, a genetic algorithm is used to perform crossover and mutation operations to obtain the offspring population;

[0093] Determine whether the offspring population meets the preset iteration termination condition. If it does, perform global optimal individual gene encoding analysis to obtain an optimized pump group control sequence containing the original set frequency value.

[0094] First, perform population initialization and frequency change rate constraint verification operations. Based on the energy consumption ranking list generated in step S13, select the top [population] with the lowest energy consumption value. Five (e.g., 50) alternative combinations are used as the initial parent population. A real-number encoding strategy is employed to convert each combination into a gene vector. ,in This represents the set frequency of the i-th pump in the k-th scheme. A preset frequency change rate constraint is introduced, and a maximum frequency change rate threshold is set. (For example, 2.5 Hz / s). This threshold The settings are based on the acceleration and deceleration time parameters in the VFD hardware manual and the water hammer protection standards for water supply networks (such as CECS 193:2005), aiming to prevent pressure oscillations in the pipeline network caused by sudden frequency changes. The adjustment response time is set. (For example, 5 seconds), this parameter is determined based on the physical response lag time of the PLC controller's instruction scan cycle and the water pump speed. Calculate the rate of change for each gene locus in the initial population:

[0095]

[0096] in This represents the current real-time operating frequency of the i-th pump. If detected... If this occurs, a forced correction strategy will be triggered. The formula for calculating the forced correction strategy is:

[0097]

[0098] in As a sign function, this strategy ensures that all individuals satisfy the dynamic constraints by forcibly clamping the out-of-limit frequency target value to a physically reachable linear ramp boundary.

[0099] Secondly, fitness calculation and roulette wheel selection operations are performed. A multi-objective collaborative evaluation function is constructed. ,in This is the normalized energy consumption value for the scheme. The estimated pressure fluctuation variance for this scheme (estimated through a simplified hydraulic model) is calculated by performing max-min normalization on the maximum and minimum values ​​of the pressure fluctuation variance in the current population, mapping it to the [0,1] interval. Specifically, for an alternative combination scheme, a pre-set pipeline hydraulic model (which can employ iterative algorithms such as the Hardy-Cross method or a simplified model based on electro-hydraulic similarity) is used to simulate the pressure values ​​of key nodes in the pipeline network under the given discrete flow points and corresponding probability weights. The expected pressure value of this node under all discrete scenarios is calculated. The pressure fluctuation variance can be obtained by calculating the square of the difference between the pressure value of the key node and the expected pressure value, multiplying it by the corresponding probability weight, and finally summing the results. The key node is usually selected as the end of the pipeline network or the point with the highest geographical elevation. and To optimize the weighting coefficients, the following settings are provided: The weights are set using the Analytic Hierarchy Process (AHP), calculated based on a scheduling strategy matrix that prioritizes energy conservation while also considering stability. Individual fitness is then calculated.

[0100]

[0101] Based on the fitness values, the roulette wheel selection algorithm is executed. First, the selection probability of each individual is calculated:

[0102]

[0103] Then calculate the cumulative probability:

[0104]

[0105] Finally, a random number r is generated in the interval [0,1], and a number that satisfies this condition is selected. The k-th individual enters the next generation, and this process is repeated until the population is full, thus selecting high-probability high-quality individuals. Next, crossover and adaptive parameter adjustment operations are performed. Arithmetic crossover is performed on the selected individuals to generate offspring. Then, non-uniform Gaussian mutation is performed, with the mutation formula as follows:

[0106]

[0107] The disturbance amount Follows a normal distribution To balance the algorithm's global search capability in the early stages with its local convergence accuracy in the later stages, an adaptive strategy is adopted to dynamically reduce the standard deviation of variation. The calculation formula is:

[0108]

[0109] in Let be the initial standard deviation (e.g., 2.0 Hz), and t be the current iteration number. This is the preset maximum number of iterations (e.g., 100 generations). This formula ensures that the mutation range decays cubically with each iteration, guaranteeing that the algorithm can perform fine-tuning searches in later stages.

[0110] Finally, the iteration termination and gene analysis operations are performed. At the end of each iteration, the termination condition is checked; if the current iteration number is... Or the variation in the optimal fitness of the population over 20 consecutive generations is less than If the iteration fails, the iteration terminates. The globally optimal individual with the highest fitness is locked. Then, its gene coding is analyzed. Specifically, according to the indexing rules during coding mapping, the values ​​in the gene vector are... The corresponding physical device IDs (such as "1# Variable Frequency Pump" and "2# Industrial Frequency Pump") are mapped back to the original values, and the values ​​are restored to the control command format. The positions in the shutdown state (i.e., the frequency is 0Hz) are marked as "OFF", and the positions in the running state retain their precise floating-point frequency values ​​(such as 42.35Hz), thus obtaining the optimized pump group control sequence containing the original set frequency values.

[0111] It should be noted that the optimized weight coefficients described in this embodiment ( ) and frequency change rate threshold ( This is not a fixed constant; its value needs to be dynamically adjusted based on the specific water supply plant's operational strategy preferences and equipment specifications. For example, in areas with extremely high requirements for water supply stability (such as hospitals or precision manufacturing areas), the weighting coefficient for pressure fluctuation variance can be appropriately increased. The frequency change rate threshold should be set according to the specifications provided by the motor manufacturer (following IEC 60034 standard) regarding the motor insulation class and maximum permissible torque change rate, to ensure that the motor life is not compromised while pursuing speed adjustment. Those skilled in the art can adjust the above parameters according to specific operating conditions, which does not depart from the technical scope of this invention.

[0112] In step S15, based on the optimized pump group control sequence and the real-time operating data of the pump station, the flow matching degree is calculated. If the flow matching degree is lower than a preset flow matching degree threshold, the speed of the optimized pump group control sequence is fine-tuned to obtain the final control command, including:

[0113] Based on the optimized pump group control sequence and the real-time operation data of the pump station, the difference is calculated to obtain a multidimensional deviation feature vector;

[0114] The multidimensional deviation feature vector is input into a preset flow matching degree evaluation model for calculation to obtain the flow matching degree;

[0115] If the flow matching degree is lower than the preset flow matching degree threshold, the multidimensional deviation feature vector is analyzed and calculated according to the pump similarity law to obtain the speed fine adjustment amount;

[0116] The speed fine-tuning amount is superimposed on the original set frequency value corresponding to the optimized pump group control sequence to obtain pump group speed control data. Based on the pump group speed control data, equipment control instructions are generated to obtain the final control instructions.

[0117] First, analyze the optimized pump control sequence output in step S14 and extract the original set frequency values ​​of each pump that is started. The preset pump characteristic equation is invoked. This equation is based on the type test report provided by the pump manufacturer (following the GB / T 3216 standard), and the QH characteristic curve is obtained by least squares quadratic polynomial fitting at no less than 10 test operating points. Its mathematical expression is: (in (These are the fitting coefficients). Combining the similarity law of water pumps, the dynamic characteristic equations under the current set frequency are constructed as follows:

[0118]

[0119] in The power frequency is 50Hz, and a preset pipeline resistance equation is invoked:

[0120]

[0121] in S is the static head, and S is the resistance coefficient. The static head is an inherent physical parameter of the water supply system. Its value is the difference between the design water level of the clear water tank of the water supply plant and the required free head at the most unfavorable point of the pipeline network, and is converted into pressure units. This parameter comes from the design drawings of the water supply pipeline network and is used as a fixed constant input during system initialization. In actual operation, it can be verified and corrected by the measured value of the pump outlet pressure when all users are shut down and only the static pressure of the pipeline network is maintained.

[0122] Solve the system of equations simultaneously:

[0123] Obtain the solution to the system of equations This refers to the theoretical expected flow rate and theoretical expected head at that frequency. Simultaneously, the actual instantaneous flow rate at the pump station outlet is collected via the SCADA system. With actual main pipe pressure Obtain the rated flow rate of the pump set. With rated head These two parameters are set based on the optimal efficiency point (BEP) data indicated on the pump's nameplate. Calculate the normalized deviation:

[0124] ;

[0125] ;

[0126] Constructing multidimensional bias feature vectors Secondly, the evaluation model is used to calculate the traffic matching degree and threshold determination operation. The multidimensional deviation feature vector is then used... Input to a pre-defined flow matching evaluation model, which uses Gaussian radial basis functions to calculate the matching degree:

[0127]

[0128] in This is a bandwidth parameter, typically set to 0.1. It sets a preset traffic matching threshold. This threshold was determined through offline simulation testing using historical 30-day operating data. The threshold was set within the range [0.5, 0.95] with a step size of 0.05. The oscillation frequency of the control command (i.e., the number of times the direction of frequency increase / decrease changes per unit time) under different thresholds was statistically analyzed. The value corresponding to the abrupt change point of the first derivative of the oscillation frequency curve (e.g., 0.85) was selected as the threshold. To achieve a balance between response speed and system stability. If the calculated... If the current control model is mismatched, it is determined that there is a significant model mismatch and correction is required.

[0129] Finally, fine-tuning of the pump speed and limiting of the amplitude are performed based on the similarity law. According to the principle of proportionality between the head and the square of the pump speed in the pump similarity law, the speed correction coefficient is calculated:

[0130]

[0131] in Set the target pressure. Calculate the speed fine-tuning:

[0132]

[0133] The fine-tuning amount is superimposed on the original set frequency to obtain the initial corrected frequency. .right Perform amplitude limiting processing and set a lower limit for the allowed frequency. (Typically set to 30Hz, adjusted according to motor cooling and water pump NPSHr requirements) and upper limit. (Typically set to 50Hz, depending on the power grid frequency and distribution capacity). The final execution frequency is calculated using the limiting formula:

[0134]

[0135] Will The signal is mapped to the analog output signal (4-20mA) of the PLC controller or the value of the communication register, and the final control instruction containing the specific device address is generated and issued for execution.

[0136] In step S16, actual water supply data after executing the final control command is collected. This data is then used for adaptive feedback updates and simulation verification to obtain a continuously optimized pump unit operating status scheme, including:

[0137] Collect the actual water supply data after executing the final control command, and calculate the difference between the actual water supply data and the preset water supply head data to obtain the deviation feedback vector.

[0138] Based on the aforementioned deviation feedback vector, data fusion is performed with preset historical water supply records to obtain an extended training sample set;

[0139] Based on the expanded training sample set, the random forest algorithm is updated and its parameters are iteratively trained to obtain improved prediction model parameters;

[0140] Based on the improved prediction model parameters, the random scenario simulation and energy consumption calculation are repeatedly performed to obtain simulated energy loss data;

[0141] If the simulated energy loss data is greater than zero, the equipment control logic is locked to obtain a continuously optimized pump group operation status scheme.

[0142] First, data acquisition and deviation feedback vector construction are performed. The pump station outlet flow rate count is read in real time after the final control command is executed via the communication interface between the SCADA system and the PLC controller (e.g., Modbus TCP protocol). The value of the main pipe pressure transmitter This constitutes the actual water supply data. Simultaneously, it accesses the PLC's internal holding register (e.g., address 40001) to read the preset water supply head data corresponding to the current scheduling time. Based on the physical deviation calculation formula, calculate the absolute pressure deviation respectively. relative deviation from flow rate (in The expected flow predicted in step S11, (For the rated flow rate), the two deviation values ​​are concatenated sequentially to construct a two-dimensional deviation feedback vector. .

[0143] Secondly, perform data fusion and training sample set update operations. Obtain the timestamp of the current sampling moment. The system retrieves environmental characteristic data (temperature, humidity, and holiday markers) corresponding to the current moment from the real-time influencing factor database and uses this environmental characteristic data as input features. The actual water supply flow rate value As a tag Combine to generate a new standard training record A sliding window mechanism is used to update preset historical water supply records: a system with a fixed capacity is established. A first-in, first-out (FIFO) queue (e.g., 50,000 records) is used to push the new standard training records to the tail of the queue while removing the oldest record with the earliest timestamp from the head of the queue, thus obtaining an updated expanded training sample set. This process ensures that the training data always covers the most recent records. The system dynamics within each sampling period eliminate the interference of outdated data on prediction accuracy.

[0144] Next, a full retraining and parameter optimization operation is performed on the random forest model. For the same random forest algorithm model established in step S11, a full retraining process is executed using the expanded training sample set. To improve the model's adaptability to recent operating conditions, a grid search method is used to re-optimize the model's hyperparameters. Specifically, the maximum depth of the decision tree is set. The search space is [10, 50] with a step size of 5; the search space for setting the internal node splitting threshold is [2, 10] with a step size of 2. Each parameter combination in the search space is traversed. The out-of-bag error under this parameter is calculated using the Bootstrap method, and the formula is as follows:

[0145]

[0146] Where I is an indicator function. This is the OOB predicted value. Select the value that makes... Minimal parameter combination As the optimal parameters, the structural parameters of the random forest model are updated to obtain the improved prediction model parameters.

[0147] Then, simulated energy consumption verification and logic locking operations are performed. Using the improved prediction model parameters, the real-time influencing factor data of the current moment is input back into the model, and the corrected predicted flow value is output. For each pump unit k included in the final control command, its operating frequency is obtained. The flow-efficiency characteristic equation corrected using the pump similarity law is as follows:

[0148]

[0149] in To allocate the flow component to the k-th pump according to the characteristic curve of the parallel pump set, Calculate the operating efficiency of each pump unit at the corrected flow rate, using the rated frequency. At the same time, calculate the required head of the pipeline network.

[0150]

[0151] Substituting the above parameters into the pump set power equation, the corrected expected energy consumption power is calculated. :

[0152]

[0153] Where M represents the number of pump units in operation. Where is the fluid density, and g is the acceleration due to gravity. For motor efficiency. Meanwhile, assuming the pump unit continues operating under the old conditions from the previous scheduling time, calculate the theoretical baseline energy consumption power under unoptimized operating conditions. Calculate simulated energy loss data (i.e., energy saving benefits):

[0154]

[0155] in To control the cycle. If This indicates that the current control command still has positive energy-saving benefits under the updated model perspective, thus triggering the device control logic lockout. The current PID parameters and frequency command are written to the controller's non-volatile memory area, and the logic lockout time is set. The aforementioned The setting is based on the hydraulic response time constant of the water supply network. (Measured through a step response test, typically lasting 3-5 minutes), set (For example, 450 seconds), during which non-fault scheduling instructions are blocked to prevent actuator oscillations caused by frequent model fine-tuning, thereby obtaining a continuously optimized pump group operation status scheme.

[0156] It should be noted that, in the initial stage of implementation (cold start phase), if there is a lack of sufficient historical water consumption sequences or historical scheduling logs to construct a random forest model and set a deviation probability threshold, the method described in this invention may be ineffective. The system supports an initialization mode of "manual preset + rule-guided". Specifically, typical historical data from similar water supply plants can be imported as the initial training set, or the theoretical maximum / minimum flow and design pressure curves in the water supply network design drawings can be used as the initial constraint boundaries. As the system operates for a longer period of time and real-time water supply data accumulates, the system will automatically use the adaptive feedback update mechanism described in step S16 to gradually replace the initial preset parameters, thereby achieving self-evolution of model accuracy.

[0157] Reference Figure 2 The second embodiment of the present invention provides a water supply plant scheduling system based on multi-objective collaborative optimization, comprising:

[0158] The multi-source data prediction module is used to acquire real-time influencing factor data of the water supply system, and combine it with the pre-stored historical water consumption sequence to perform multi-source data nonlinear fitting prediction on the real-time influencing factor data and the historical water consumption sequence to obtain the predicted distribution range of water supply.

[0159] The random scenario simulation module is used to perform random scenario simulation and deviation quantification analysis based on the predicted water supply distribution range and the current configuration of the pump group operating status collected in real time, so as to obtain the probability distribution of water supply deviation.

[0160] The scheme energy consumption assessment module is used to perform alternative scheme retrieval and energy consumption assessment based on the water supply deviation probability distribution and the preset deviation probability threshold, and obtain an energy consumption ranking list.

[0161] The dynamic iterative optimization module is used to perform dynamic constraint iterative optimization based on the energy consumption sorting list and a preset frequency change rate constraint condition using a genetic algorithm to obtain the optimized pump group control sequence.

[0162] The real-time matching and adjustment module is used to calculate the flow matching degree based on the optimized pump group control sequence and the real-time operating data of the pump station. If the flow matching degree is lower than the preset flow matching degree threshold, the speed of the optimized pump group control sequence is finely adjusted to obtain the final control command.

[0163] The feedback closed-loop verification module is used to collect the actual water supply data after the execution of the final control command, and use the actual water supply data to perform adaptive feedback updates and simulation verification to obtain a continuously optimized pump group operation status scheme.

[0164] It should be noted that the water supply plant scheduling system based on multi-objective collaborative optimization provided in this embodiment of the invention is used to execute all the process steps of the water supply plant scheduling method based on multi-objective collaborative optimization in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0165] This invention also provides an electronic device. The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a cold-plate intelligent processing control program. When the processor executes the computer program, it implements the steps described in the various embodiments of the cold-plate intelligent processing control method, for example... Figure 1 The step S11 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above system embodiments, such as the multi-source data prediction module.

[0166] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0167] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0168] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting all parts of the electronic device via various interfaces and lines.

[0169] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0170] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the system embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0171] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that 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 for those skilled in the art.

Claims

1. A water supply plant scheduling method based on multi-objective collaborative optimization, characterized in that, include: Real-time influencing factor data of the water supply system is obtained, and combined with the pre-stored historical water consumption sequence, multi-source data nonlinear fitting prediction is performed on the real-time influencing factor data and the historical water consumption sequence to obtain the predicted distribution range of water supply. Based on the predicted water supply distribution range and the current configuration of the pump unit operating status collected in real time, random scenario simulation and deviation quantification analysis are performed to obtain the probability distribution of water supply deviation. According to the water supply deviation probability distribution, if the deviation probability exceeds the preset threshold, then alternative schemes are searched and energy consumption is evaluated based on the predicted water supply distribution range to obtain an energy consumption ranking list. Based on the energy consumption ranking list, and combined with the preset frequency change rate constraint, a genetic algorithm is used to perform dynamic constraint iterative optimization to obtain the optimized pump group control sequence. Based on the optimized pump group control sequence and the real-time operating data of the pump station, the flow matching degree is calculated. If the flow matching degree is lower than the preset flow matching degree threshold, the speed of the optimized pump group control sequence is fine-tuned to obtain the final control command. The actual water supply data after the final control command is executed is collected, and the actual water supply data is used for adaptive feedback update and simulation verification to obtain a continuously optimized pump group operation status scheme. The step of using a genetic algorithm to perform dynamic constraint iterative optimization based on the energy consumption ranking list and a preset frequency change rate constraint to obtain the optimized pump group control sequence includes: Based on the energy consumption ranking list, an encoding mapping is performed to obtain an initial population encoding that includes the frequency change rate; Determine whether the frequency change rate corresponding to the initial population code satisfies the preset frequency change rate constraint condition. If it does, calculate the individual fitness value of the initial population code. Based on the individual fitness values, a genetic algorithm is used to perform crossover and mutation operations to obtain the offspring population; Determine whether the offspring population meets the preset iteration termination condition. If it does, perform global optimal individual gene encoding analysis to obtain an optimized pump group control sequence containing the original set frequency value. Specifically, the step of performing encoding mapping based on the energy consumption ranking list to obtain an initial population encoding containing the frequency change rate is achieved through: Perform population initialization and frequency change rate constraint verification operations, and select the population with the lowest energy consumption value from the energy consumption sorting list. One alternative combination scheme is used as the initial parent population; A real-number encoding strategy is used to convert each scheme into a gene vector. ,in This represents the set frequency of the i-th water pump in the k-th scheme; Introduce a preset frequency change rate constraint and set a maximum frequency change rate threshold. Set and adjust response time Calculate the rate of change for each gene locus in the initial population: in Let i be the current real-time operating frequency of the i-th pump. If it is detected... If this occurs, a forced correction strategy will be triggered. The formula for calculating the forced correction strategy is as follows: in It is a symbolic function.

2. The water supply plant scheduling method based on multi-objective collaborative optimization according to claim 1, characterized in that, The process involves acquiring real-time influencing factor data of the water supply system, combining it with pre-stored historical water consumption sequences, and performing multi-source data nonlinear fitting prediction on the real-time influencing factor data and the historical water consumption sequences to obtain the predicted water supply distribution range, including: Real-time influencing factor data is collected from the multi-source data acquisition system, and pre-stored historical influencing factor data and pre-stored historical water consumption sequences are obtained. Based on the historical influencing factor data, missing value imputation and normalization preprocessing are performed to obtain the historical feature vector matrix; Based on the historical feature vector matrix and the historical water consumption sequence, a random forest algorithm is used for fitting and training to obtain a water supply regression prediction model. Based on the real-time influencing factor data, a feature vector to be predicted is constructed, and the feature vector to be predicted is input into the water supply regression prediction model to obtain the output value distribution. Based on the distribution of the output values, quantile limits are calculated to obtain the predicted distribution range of water supply.

3. The water supply plant scheduling method based on multi-objective collaborative optimization according to claim 1, characterized in that, The step involves performing random scenario simulation and deviation quantification analysis based on the predicted water supply distribution range and the current configuration of the pump unit operating status collected in real time, to obtain the probability distribution of water supply deviation, including: Based on the predicted water supply distribution range and the current configuration of the pump group operating status collected in real time, a Monte Carlo simulation generator is used to perform random sampling to obtain a random demand scenario sequence containing instantaneous demand values. Extract the instantaneous demand values ​​from the random demand scenario sequence and obtain the fixed water supply capacity value corresponding to the current configuration of the pump group's operating status; Based on the difference between the instantaneous demand value and the fixed water supply capacity value, a difference sequence is obtained; Based on the difference sequence, the water supply surplus and water supply deficit are identified, and kernel density estimation is performed on the water supply surplus and water supply deficit to obtain the water supply deviation probability distribution.

4. The water supply plant scheduling method based on multi-objective collaborative optimization according to claim 1, characterized in that, If the water supply deviation probability distribution exceeds a preset deviation probability threshold, then an alternative solution search and energy consumption assessment are performed based on the predicted water supply distribution range to obtain an energy consumption ranking list, including: If the probability distribution of water supply deviation exceeds the preset deviation probability threshold, alternative combination schemes are retrieved from the preset pump group control library. Based on the predicted water supply distribution range, the data is discretized to obtain multiple discrete flow points. Based on the multiple discrete flow points and the alternative combination schemes, the operating point is calculated by combining the preset pump set characteristic equation and the preset pipeline resistance coefficient to obtain the shaft power value. Based on the shaft power value, a weighted integral is calculated to obtain the expected energy consumption value of each of the alternative combination schemes, and an energy consumption ranking list is generated based on the expected energy consumption value.

5. The water supply plant scheduling method based on multi-objective collaborative optimization according to claim 1, characterized in that, The process involves calculating the flow matching degree based on the optimized pump group control sequence and the real-time operating data of the pump station. If the flow matching degree is lower than a preset flow matching degree threshold, the optimized pump group control sequence is fine-tuned to obtain the final control command, including: Based on the optimized pump group control sequence and the real-time operation data of the pump station, the difference is calculated to obtain a multidimensional deviation feature vector; The multidimensional deviation feature vector is input into a preset flow matching degree evaluation model for calculation to obtain the flow matching degree; If the flow matching degree is lower than the preset flow matching degree threshold, the multidimensional deviation feature vector is analyzed and calculated according to the pump similarity law to obtain the speed fine adjustment amount; The speed fine-tuning amount is superimposed on the original set frequency value corresponding to the optimized pump group control sequence to obtain pump group speed control data. Based on the pump group speed control data, equipment control instructions are generated to obtain the final control instructions.

6. The water supply plant scheduling method based on multi-objective collaborative optimization according to claim 2, characterized in that, The process involves collecting actual water supply data after executing the final control command, using this data for adaptive feedback updates and simulation verification, and obtaining a continuously optimized pump unit operation status scheme, including: Collect the actual water supply data after executing the final control command, and calculate the difference between the actual water supply data and the preset water supply head data to obtain the deviation feedback vector. Based on the aforementioned deviation feedback vector, data fusion is performed with preset historical water supply records to obtain an extended training sample set; Based on the expanded training sample set, the random forest algorithm is updated and its parameters are iteratively trained to obtain improved prediction model parameters; Based on the improved prediction model parameters, the random scenario simulation and energy consumption calculation are repeatedly performed to obtain simulated energy loss data; If the simulated energy loss data is greater than zero, the equipment control logic is locked to obtain a continuously optimized pump group operation status scheme.

7. A water supply plant scheduling system based on multi-objective collaborative optimization, characterized in that, The method for scheduling water supply plants based on multi-objective collaborative optimization as described in any one of claims 1 to 6 includes: The multi-source data prediction module is used to acquire real-time influencing factor data of the water supply system, and combine it with the pre-stored historical water consumption sequence to perform multi-source data nonlinear fitting prediction on the real-time influencing factor data and the historical water consumption sequence to obtain the predicted distribution range of water supply. The random scenario simulation module is used to perform random scenario simulation and deviation quantification analysis based on the predicted water supply distribution range and the current configuration of the pump group operating status collected in real time, so as to obtain the probability distribution of water supply deviation. The scheme energy consumption assessment module is used to perform alternative scheme retrieval and energy consumption assessment based on the water supply deviation probability distribution and the preset deviation probability threshold, and obtain an energy consumption ranking list. The dynamic iterative optimization module is used to perform dynamic constraint iterative optimization based on the energy consumption sorting list and a preset frequency change rate constraint condition using a genetic algorithm to obtain the optimized pump group control sequence. The real-time matching and adjustment module is used to calculate the flow matching degree based on the optimized pump group control sequence and the real-time operating data of the pump station. If the flow matching degree is lower than the preset flow matching degree threshold, the speed of the optimized pump group control sequence is finely adjusted to obtain the final control command. The feedback closed-loop verification module is used to collect the actual water supply data after the execution of the final control command, and use the actual water supply data to perform adaptive feedback updates and simulation verification to obtain a continuously optimized pump group operation status scheme.

Images