A multi-water source coordinated dispatching system for a water supply system

Abstract

The present invention relates to the field of urban water supply technology, and in particular to a multi-water source coordinated scheduling system for a water supply system, comprising: a data acquisition module for acquiring the initial historical demand sequence of each water plant; a feature reconstruction module for performing feature reconstruction based on the initial historical demand sequence to obtain a reconstructed historical demand sequence; the reconstructed historical demand sequence includes: feature information characterizing a sudden change in water intake in the initial historical demand sequence; a demand prediction module for predicting the predicted water demand of each water plant based on the reconstructed historical demand sequence; a water supply capacity module for calculating the water supply capacity of each water source; and a multi-objective optimization module for performing optimization and solving according to preset objective functions and constraints to obtain the optimal scheduling water volume for each transmission pipeline. The present invention can support early warning and dynamic response to emergencies, and realize more accurate, flexible and systematic multi-water source scheduling decisions.

Description

Technical Field

[0001] The present invention relates to the technical field of urban water supply, and in particular to a multi-water source coordinated dispatching system for a water supply system. Background Art

[0002] With the rapid development of social economy and the continuous advancement of urbanization, the demand for water resources in large cities continues to grow. However, due to factors such as climate change, environmental pollution and uneven distribution of water resources, urban water supply faces severe challenges. The traditional water supply model based on a single water source can no longer meet the complex and changing water needs of cities, and there are great risks in responding to emergencies.

[0003] Existing multi-source water scheduling technologies are generally based on static data and optimization calculations, which cannot adapt to the rapidly changing water resources and user needs. At the same time, due to the lack of forward-looking and predictive information on the future state of the water source-pipeline-water plant system, the water supply system's response to emergencies is delayed. In addition, due to the lack of a reasonable description of the topological relationship of the water supply system, the scheduling goals are often limited to balancing the supply and demand of raw water and reducing operating costs. Furthermore, existing technologies generally lack consideration of the rational allocation of various types of water sources, and ignore the contradiction between the distribution of water resources and the regional economic development pattern, and do not make systematic plans for the use of multiple water sources.

[0004] Specifically, the main problems faced by water source scheduling in urban water supply systems are:

[0005] (1) There are various types of water intake. Raw water may come from reservoirs, lakes, rivers, groundwater, or be diverted from other cities. Water plants have different priorities when selecting raw water from different sources.

[0006] (2) The number of water sources is large and scattered, and the topology of the water supply network is extremely complex;

[0007] (3) There are many constraints. When designing the scheduling plan, it is necessary to comprehensively consider factors such as reservoir capacity, raw water quality, water transfer quota, and water transmission distance.

[0008] Traditional urban raw water dispatch is still mainly based on rule dispatch, which mainly relies on the professional knowledge and experience of water managers, has poor reliability and is not conducive to promotion, and is not competent for the dispatch of complex large-scale urban water supply systems. In contrast, multi-water source coordinated dispatch technology based on reservoir networking and joint supply has become the key to solving the water supply problem of large cities.

[0009] Coordinated scheduling of multiple water sources can provide a strong water source guarantee to alleviate the contradiction between urban or regional water supply and demand, but it also makes the water supply system more complicated and puts forward new requirements for the allocation and management of local water resources. Although the existing multi-water source scheduling methods have achieved certain results in strategy optimization in complex scenarios, they often rely on empirical judgment and static planning. When dealing with the above problems, it is difficult to adapt to the rapidly changing water resource conditions and user needs. At the same time, due to the lack of forward-looking and predictive information on the future state of the water supply system consisting of water sources, pipelines, and water plants, the response to emergencies is delayed. In addition, the existing technologies generally lack consideration of the reasonable allocation of various types of water sources, and ignore the contradiction between the distribution of water resources and the regional economic development pattern, and do not make systematic planning for the use of multiple water sources. Summary of the invention

[0010] 1. Technical issues to be resolved

[0011] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a multi-water source coordinated scheduling system for a water supply system, which solves the technical problem that the prior art lacks forward-looking and predictive information on the future state of the water supply system, resulting in delayed response to emergencies.

[0012] (II) Technical solution

[0013] In order to achieve the above object, the main technical solutions adopted by the present invention include:

[0014] The present invention provides a multi-water source coordinated scheduling system for a water supply system, wherein the water supply system comprises at least two water sources and at least one water plant, and a transmission pipeline for connecting the two; the multi-water source coordinated scheduling system comprises:

[0015] A data acquisition module, used to acquire an initial historical demand sequence of each water plant, the historical demand sequence including the daily historical water intake of the water plant within a preset historical time interval;

[0016] A feature reconstruction module is used to perform feature reconstruction based on the initial historical demand sequence to obtain a reconstructed historical demand sequence; the reconstructed historical demand sequence includes: a mutation factor feature sequence; the mutation factor feature sequence is used to characterize feature information of a mutation in historical water intake in the initial historical demand sequence;

[0017] A demand forecasting module, used to forecast the predicted water demand of each water plant within a preset target scheduling time interval based on the reconstructed historical demand sequence;

[0018] Water supply capacity module, used to calculate the water supply capacity of each water source;

[0019] The multi-objective optimization module is used to optimize and solve the water demand forecast of each water plant, the water supply capacity of each water source and the topological relationship of the water supply system according to the preset objective function and constraints, so as to obtain the optimal dispatching water volume of each transmission pipeline within the preset target dispatching time interval.

[0020] Optionally, the feature reconstruction module includes:

[0021] The first mutation feature construction unit is used to roll-count the cumulative sum of target feature values ​​based on the initial historical demand sequence; and determine the mutation factor feature sequence according to whether the cumulative sum of the target feature values ​​is greater than a preset threshold.

[0022] Optionally, the target feature value is a rolling mean, and the first mutation feature construction unit includes a rolling statistics subunit, a cumulative sum subunit and a judgment subunit;

[0023] The rolling statistical subunit is used to determine the rolling mean value corresponding to the daily historical water withdrawal of the initial historical demand sequence within the rolling window according to formula (1) based on the initial historical demand sequence;

[0024] (1);

[0025] represents the rolling mean in the rolling window corresponding to the tth day, w represents the width of the current rolling window, and t represents the right end of the current rolling window. Indicates the left end of the current scroll window. represents the i-th historical water withdrawal in the current rolling window;

[0026] The cumulative sum subunit is used to determine the daily cumulative sum based on the rolling mean according to formula (2);

[0027] (2);

[0028] represents the cumulative sum corresponding to the tth day, max(·) represents the maximum value operation, represents the historical water withdrawal on day t;

[0029] The judgment subunit is used to judge whether the corresponding daily cumulative sum is greater than a preset threshold based on the daily cumulative sum; if so, the deviation between the historical water intake corresponding to the day and the rolling mean is used as the mutation factor corresponding to the day, and filled in the position corresponding to the day in the mutation factor feature sequence; if not, 0 is filled in the position corresponding to the day in the mutation factor feature sequence to obtain the mutation factor feature sequence.

[0030] Optionally, in the first mutation feature construction unit, the determining whether the daily corresponding cumulative sum is greater than a preset threshold comprises:

[0031] The preset threshold is based on a known non-mutated initial historical demand sequence and is predetermined according to formula (3);

[0032] (3);

[0033] represents the preset threshold, k represents the empirical constant, Represents the rolling standard deviation of the known non-mutated initial historical demand series.

[0034] Optionally, the feature reconstruction module further includes:

[0035] The second mutation feature construction unit is used to mark the day as a mutation point to form a mutation point indication sequence when the first mutation feature construction unit determines that the corresponding cumulative sum is greater than a preset threshold.

[0036] Optionally, the feature reconstruction module further includes a multi-dimensional feature reconstruction unit and a splicing unit;

[0037] The multidimensional feature reconstruction unit is used to decompose the initial historical demand sequence by using a seasonal and trend decomposition method based on local moving average to obtain a trend feature sequence, a seasonal feature sequence and a residual feature sequence;

[0038] The splicing unit is used to splice the mutation factor feature sequence, the mutation point indication sequence, the trend feature sequence, the season feature sequence or the residual feature sequence to obtain a reconstructed historical demand sequence.

[0039] Optionally, the demand forecasting module includes at least one of a next-month forecasting unit and a next-day forecasting unit;

[0040] The next month prediction unit is used to predict the predicted water demand of each water plant in the next month based on the first prediction model and the reconstructed historical demand sequence; the first prediction model is an LSTM (Long Short Term Memory Network) model with adapted model parameters;

[0041] The next-day prediction unit is used to predict the predicted water demand of each water plant on the next day based on the second prediction model and the reconstructed historical demand sequence; the second prediction model is a Transformer model with adapted model parameters.

[0042] Optionally, the water source includes a reservoir water source and an external water source, and the water supply capacity module includes at least one of a reservoir water source water supply capacity unit and an external water source water supply capacity unit;

[0043] The reservoir water source water supply capacity unit is used to calculate the water supply capacity of each reservoir water source;

[0044] The external water source supply capacity unit is used to calculate the water supply capacity of each external water source.

[0045] Optionally, in the reservoir water source water supply capacity unit, the calculating of the water supply capacity of each reservoir water source includes:

[0046] Based on the basic storage capacity of the reservoir, the current daily water supply, the number of days in the preset target scheduling time interval, and the total inflow and total outflow in the preset target scheduling time interval, the water supply capacity of the reservoir water source is determined according to formula (4);

[0047] (4);

[0048] represents the water supply capacity of the i-th reservoir, T represents the duration of the target scheduling time interval, represents the basic storage capacity of the i-th reservoir, represents the current daily water supply of the i-th reservoir, Indicates the number of days from the current time to the start date of the target scheduling time interval. represents the total inflow of the i-th reservoir within the target scheduling time interval, represents the total outflow of the i-th reservoir within the target scheduling time interval;

[0049] Optionally, in the external water source supply capacity unit, the calculating of the water supply capacity of each external water source includes:

[0050] Based on the total annual water supply of the external water source, the water supply volume and the duration of the target scheduling time interval, the water supply capacity of the external water source is determined according to formula (5);

[0051] (5);

[0052] represents the water supply capacity of the i-th external water source, T represents the number of days in the target scheduling time interval, represents the total annual water supply of the i-th external water source, represents the amount of water supplied by the i-th external water source, Represents the allocation coefficient for the mth month.

[0053] Optionally, a directed graph construction unit is used to construct a directed graph based on the topological relationship of the water supply system, the directed graph comprising nodes and edges, the nodes representing water sources or water plants, the edges representing transmission pipelines connecting the two, and the direction of the edges representing the transmission direction of the transmission pipelines;

[0054] The solving unit is used to optimize and solve the water volume that each transmission pipeline should transport based on the predicted water demand, the current water supply capacity of the water source and the topological relationship of the water supply system according to the preset objective function and constraints.

[0055] Optionally, in the solving unit, performing optimization solving according to a preset objective function and constraint conditions includes:

[0056] The objective function is formula (6):

[0057] (6);

[0058] min(·) means to take the minimum value operation. to represents the preset weight parameter, represents the total transmission distance of all water sources, represents the pumping energy consumption, Indicates the supply of external water sources, It represents the supply balance coefficient of reservoir water among all water sources. to Based on the directed graph, according to formulas (7) to (10), it is determined that:

[0059] (7);

[0060] (8);

[0061] (9);

[0062] (10);

[0063] It represents the dispatched water volume of the pipeline connecting node i and node j in the directed graph within the preset target dispatch time interval. represents the set of all edges in a directed graph, represents the water transport distance between node i and node j, , is an indicator function, which means that when x is less than 0, The value of is 1, when x is greater than or equal to 0, The value of is 0. It represents the elevation difference of the pipeline connecting node i and node j. represents the set of external water sources, R represents the set of reservoir water sources, represents the water supply capacity of the ith water source, It represents the average value of the ratio of the actual water supply of all reservoir sources to the water supply capacity. Indicates the quantity of water source in the reservoir.

[0064] Optionally, in the solving unit, performing optimization solving according to a preset objective function and constraint conditions includes:

[0065] The constraints are formulas (11) to (13):

[0066] (11);

[0067] (12);

[0068] (13);

[0069] P represents the set of all water plants, It represents the dispatched water volume of the pipeline connecting node i and node j in the directed graph within the preset target dispatch time interval. represents the water demand of the jth water plant, represents the water supply capacity of the ith water source, S represents the set of all water sources, It represents the water delivery capacity of the pipeline connecting node i and node j, and E represents the set of all edges in the directed graph.

[0070] (III) Beneficial effects

[0071] In the multi-water source collaborative scheduling system proposed in the present invention, the demand forecasting module predicts the predicted water demand of each water plant within the preset target scheduling time interval based on the reconstructed historical demand sequence provided by the feature reconstruction module, and provides a data basis for the multi-objective optimization module to solve the optimal scheduling water volume of each transmission pipeline. Since the reconstructed historical demand sequence includes a mutation factor feature sequence, the mutation factor feature sequence is used to characterize the characteristic information of the sudden change in water intake in the initial historical demand sequence, and thus can more accurately predict the sudden change in water demand that may occur within the target scheduling time interval, the multi-objective optimization module can generate a more reasonable scheduling water volume based on the predicted water demand, thereby supporting early warning and dynamic response to emergencies, and realizing more accurate, flexible and systematic multi-water source scheduling decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 A schematic diagram of the architecture of a multi-water source coordinated scheduling system for a water supply system provided in an embodiment;

[0073] Figure 2 A schematic diagram of the architecture of a feature reconstruction module provided in an embodiment;

[0074] Figure 3 A schematic diagram of a directed graph constructed based on a topological relationship of a specific water supply system provided in an embodiment;

[0075] Figure 4 It is a schematic diagram of the architecture of the demand forecasting module, water supply capacity module and multi-objective optimization solution module provided in the embodiment. DETAILED DESCRIPTION

[0076] In order to better explain the present invention and facilitate understanding, the present invention is described in detail below through specific implementation modes in conjunction with the accompanying drawings.

[0077] Embodiment 1

[0078] This embodiment provides a multi-water source coordinated scheduling system for a water supply system, wherein the water supply system includes at least two water sources, at least one pump station and at least one water plant, and a transmission pipeline for connecting the three. Figure 1 As shown, the multi-water source coordinated scheduling system includes a data acquisition module, a feature reconstruction module, a demand prediction module, a water supply capacity module and a multi-objective optimization module, which are as follows:

[0079] The data acquisition module is used to acquire the initial historical demand sequence of each water plant, and the historical demand sequence includes the daily historical water intake of the water plant within a preset historical time interval.

[0080] Preferably, the initial historical demand sequence can be obtained through data preprocessing based on the original daily water intake data of the water plant within a preset historical time interval. The preset historical time interval can be a time period from one year ago to the present, or other time periods selected by the user. In order to reduce the possible erroneous values, missing values, and noise in the original daily water intake data, the data preprocessing means specifically include operations such as outlier cleaning, missing value filling, and feature normalization.

[0081] Specifically, outlier cleaning processes the historical demand sequence according to the Laida criterion, and samples distributed outside the range of 3 times the standard deviation will be eliminated. The specific principles are as follows:

[0082] (14);

[0083] represents the i-th sample data, represents the mean of the sample data, represents the standard deviation of the sample data, Distributed in The probability of being in the range is 0.9973, and those distributed outside the interval are removed as outliers.

[0084] Missing value filling is implemented by linear interpolation, that is, weights are assigned according to the distance from the missing value to the two adjacent data on the left and right, and then filled using a linear function. The processing formula is:

[0085] (15);

[0086] , express Two adjacent sample data on the left and right, express Serial number, express Serial number, express Serial number.

[0087] The feature reconstruction module is used to perform feature reconstruction based on the initial historical demand sequence to obtain a reconstructed historical demand sequence.

[0088] Specifically, the reconstructed historical demand sequence includes at least a mutation factor feature sequence, and may also include other feature sequences extracted based on the initial historical demand sequence. The mutation factor feature sequence is used to characterize feature information of mutations in historical water withdrawal in the initial historical demand sequence.

[0089] The demand forecasting module is used to forecast the water demand of each water plant within a preset target scheduling time interval based on the reconstructed historical demand sequence. The preset target scheduling time interval can be specifically the next day or next month of the current time.

[0090] The water supply capacity module is used to calculate the water supply capacity of each water source. The types of water sources specifically include reservoir water sources and external water sources.

[0091] The multi-objective optimization module is used to optimize and solve based on the predicted water demand of each water plant, the water supply capacity of each water source and the topological relationship of the water supply system according to preset objective functions and constraints, so as to obtain the optimal scheduled water volume of each transmission pipeline within the preset target scheduling time interval.

[0092] The multi-water source collaborative scheduling system proposed in this embodiment can predict the predicted water demand of each water plant within the preset target scheduling time interval based on the reconstructed historical demand sequence provided by the feature reconstruction module, and provide a data basis for the multi-objective optimization module to solve the optimal scheduling water volume. Since the reconstructed historical demand sequence includes a mutation factor feature sequence, the mutation factor feature sequence is used to characterize the characteristic information of the sudden change in water intake in the initial historical demand sequence, and thus can more accurately predict the sudden change in water demand that may occur within the target scheduling time interval, the multi-objective optimization module can generate a more reasonable scheduling water volume based on the predicted water demand, thereby supporting early warning and dynamic response to emergencies, and realizing more accurate, flexible and systematic multi-water source scheduling decisions.

[0093] In a preferred implementation of this embodiment, considering that the water demand of a water plant is closely related to factors such as seasons, holidays, and population changes in the water supply area, and may also be affected by emergencies such as pipe network maintenance and water plant equipment maintenance, in order to improve the prediction effect of the demand prediction module and enable it to better identify the changing rules of trends, seasons, holidays, etc. in the water demand data, the feature reconstruction module provided in this embodiment includes a first mutation feature construction unit, a second mutation feature construction unit, a multidimensional feature reconstruction unit, and a splicing unit, as shown in FIG. Figure 2 shown.

[0094] The first mutation feature construction unit is used to calculate the cumulative sum (cumsum) of target feature values ​​based on the initial historical demand sequence in a rolling manner; and determine the mutation factor feature sequence according to whether the cumulative sum of the target feature values ​​is greater than a preset threshold.

[0095] Specifically, the target feature value is a rolling mean, and the first mutation feature construction unit includes a rolling statistics subunit, a cumulative sum subunit and a judgment subunit, which are specifically as follows:

[0096] The rolling statistical subunit is used to determine the rolling mean value corresponding to the daily historical water intake of the initial historical demand sequence within the rolling window according to formula (1) based on the initial historical demand sequence.

[0097] (1);

[0098] represents the rolling mean in the rolling window corresponding to the tth day, w represents the width of the current rolling window, and t represents the right end of the current rolling window. Indicates the left end of the current scroll window. Represents the i-th historical water withdrawal in the current rolling window.

[0099] The cumulative sum subunit is used to determine the daily cumulative sum based on the rolling mean according to formula (2).

[0100] (2);

[0101] represents the cumulative sum corresponding to the tth day, max(·) represents the maximum value operation, represents the historical water withdrawal on day t.

[0102] The judgment subunit is used to judge whether the corresponding daily cumulative sum is greater than a preset threshold based on the daily cumulative sum; if so, the deviation between the historical water intake corresponding to the day and the rolling mean is used as the mutation factor corresponding to the day, and filled in the position corresponding to the day in the mutation factor feature sequence; if not, 0 is filled in the position corresponding to the day in the mutation factor feature sequence to obtain the mutation factor feature sequence.

[0103] That is, if the judgment subunit determines that the cumulative sum corresponding to the tth day is greater than the preset threshold, the value filled in the position corresponding to the day in the mutation factor feature sequence is determined according to formula (16):

[0104] (16);

[0105] represents the mutation factor corresponding to day t, represents the historical water withdrawal on day t, Represents the rolling mean within the rolling window corresponding to day t.

[0106] The threshold value preset in the above judgment subunit is based on a known non-mutated initial historical demand sequence and is predetermined according to formula (3).

[0107] (3);

[0108] represents a preset threshold value, k represents an empirical constant, k is a positive integer, and is used to adjust the sensitivity of the judgment subunit. The smaller the k value is, the smaller the corresponding threshold value is, and the higher the sensitivity of the judgment subunit is.

[0109] The rolling standard deviation of the known non-mutated initial historical demand sequence can be calculated according to formula (17):

[0110] (17);

[0111] represents the rolling mean in the rolling window corresponding to the tth day, w represents the width of the current rolling window, and t represents the right end of the current rolling window. Indicates the left end of the current scroll window. represents the i-th historical water withdrawal in the current rolling window, Indicates the number of windows obtained by dividing the non-mutated initial historical demand sequence.

[0112] The second mutation feature construction unit is used to mark the day as a mutation point to form a mutation point indication sequence when the first mutation feature construction unit determines that the corresponding cumulative sum is greater than a preset threshold. Specifically, if the cumulative sum corresponding to the day is greater than the preset threshold, the position corresponding to the day in the mutation point indication sequence is recorded as 1, otherwise it is recorded as 0.

[0113] The above mutation factor feature sequence and mutation point indicator sequence can reflect the mutation characteristics in the historical demand sequence and improve the prediction effect of the demand forecasting module under mutation situations.

[0114] In particular, due to the large cross-regional flow of people during holidays such as the Spring Festival and National Day, the water usage pattern is quite different from the normal situation. In addition to indicating the impact of emergencies such as pipeline maintenance and water plant equipment maintenance on the historical demand sequence, the above-mentioned mutation factor characteristic sequence and mutation point indicator sequence can also more clearly indicate the change characteristics caused by holidays in the historical demand sequence, thereby improving the prediction effect of the demand forecasting module.

[0115] The multidimensional feature reconstruction unit is used to decompose the initial historical demand sequence by using the seasonal and trend decomposition (STL) method based on local moving average to obtain a trend feature sequence, a seasonal feature sequence and a residual feature sequence. For the time point t of the historical demand sequence X, the STL method first uses local regression to fit the seasonal component S t , which can be expressed by formula (18):

[0116] (18);

[0117] is the estimated value at time point t, are weights, determined by the local regression model, l is the size of the local regression window.

[0118] Next, after deseasonalization, Here, local regression is used to fit the trend component , and finally calculate the residual component The STL method is an iterative process that requires repeating the aforementioned steps many times until the changes in the trend and seasonal components are small or a predetermined number of iterations is reached.

[0119] The water demand of water plants shows a strong seasonal pattern, usually with high water demand in summer and low water demand in winter. At the same time, with the development of cities, the scale of domestic and industrial water use has gradually increased, showing certain trend characteristics. The above-mentioned multidimensional feature reconstruction unit can decompose the initial historical demand sequence from the dimensions of seasonality and trend to better reflect the corresponding characteristics. Improve the accuracy and robustness of the demand forecasting module when making predictions based on the reconstructed historical demand sequence.

[0120] The splicing unit is used to splice the mutation factor feature sequence, the mutation point indication sequence, the trend feature sequence, the season feature sequence or the residual feature sequence to obtain a reconstructed historical demand sequence.

[0121] Based on the above feature reconstruction module, the initial historical demand data can be reconstructed into a five-dimensional time series, which serves as the data basis for training, testing and using subsequent demand forecasting modules.

[0122] Based on the original daily water intake data, the data acquisition module provided in this embodiment pre-processes it, and then reconstructs the five dimensions of mutation factor, mutation time, seasonality, trend and residual through the feature reconstruction module. The reconstructed historical demand sequence can more comprehensively reflect the potential characteristics of the historical water intake data, which is helpful for the demand forecasting module to identify the law of water intake changes, thereby improving the forecasting effect of the demand forecasting module.

[0123] Embodiment 2

[0124] Based on the multi-water source coordinated scheduling system provided in Example 1, this example illustrates the specific architecture of the demand forecasting module.

[0125] like Figure 4 As shown, this embodiment provides a multi-water source coordinated scheduling system, including a demand prediction module.

[0126] The demand forecasting module is used to forecast the predicted water demand of each water plant within a preset target scheduling time interval based on the reconstructed historical demand sequence. Specifically, the demand forecasting module includes a next-month forecasting unit and a next-day forecasting unit.

[0127] The next month prediction unit is used to predict the predicted water demand of each water plant in the next month based on the first prediction model and the reconstructed historical demand sequence; the first prediction model is an LSTM model with adapted model parameters.

[0128] Specifically, the LSTM model provided in this embodiment is built based on the Pytorch framework, and the model includes 1 input layer, 2 LSTM layers, and 2 fully connected layers. The hyperparameters of the LSTM model are determined by Bayesian optimization, specifically, the optimization goal is to minimize the mean square error (MSE) of the prediction results of the LSTM model, and the hyperparameters such as the number of memory units in the LSTM layer and the number of neurons in the fully connected layer are used as optimization parameters. After multiple iterations, the hyperparameter settings of the model are determined and used for training and testing of the model.

[0129] When training the LSTM model provided in this embodiment, the above-mentioned five-dimensional reconstructed historical demand sequence can be cut into multiple groups of 5×30 data slices with 30 days as the basic unit as input data, and the actual daily historical water withdrawal of the next month can be used as output data for training and testing of the LSTM model.

[0130] When the LSTM model provided in this embodiment is used, the 5×30 reconstructed historical demand sequence of a water plant one month before the current date can be used as input data to obtain the daily demand of the water plant predicted for the next month. The total demand for the next month can be obtained by adding up the daily demand.

[0131] In actual production, extreme weather such as typhoons and rainstorms, as well as emergencies such as pipeline failures and water plant equipment maintenance, will significantly affect the production of water plants and thus affect their water demand. Therefore, it is necessary to dynamically adjust the daily water intake plan in a month according to changes in the water supply system and external environmental conditions.

[0132] The next-day prediction unit provided in this embodiment is used to predict the predicted water demand of each water plant on the next day based on the second prediction model and the reconstructed historical demand sequence; the second prediction model is a Transformer model with adapted model parameters.

[0133] Specifically, the Transformer model used in this embodiment is established based on the pytorch framework, and specifically includes a position encoding module, a mask module, a scaled point product attention module, and a multi-head attention module for its traversal construction. Considering that the prediction of water consumption belongs to a time series regression problem, this embodiment sets the decoding part of the Transformer model to a linear transformation module. Similar to the LSTM model input mentioned above, the input of the Transformer model is also a data slice obtained by time window division, and its shape is 5×15. The parameter adjustment process of the Transformer model is also implemented using Bayesian optimization, and the specific parameter settings can be adjusted according to actual conditions.

[0134] Similar to the LSTM model, when training the Transformer model provided in this embodiment, the above-mentioned five-dimensional reconstructed historical demand sequence can be cut into multiple groups of 5×15 data slices with 15 days as the basic unit as input data, and the actual historical water withdrawal 15 days after the current date can be used as output data for training and testing of the Transformer model.

[0135] The Transformer model provided in this embodiment can use a 5×15 reconstructed historical demand sequence of a water plant 15 days before the current date as input data to obtain a daily demand sequence of the water plant predicted 15 days after the current date. The first value of the demand sequence is the predicted water demand of the water plant for the next day.

[0136] This embodiment uses the deep learning model Transformer to predict the daily water demand of the water plant. This model can not only capture the complex time series relationships and nonlinear trends in historical water intake data, but its unique attention mechanism can also adjust its own weights according to the current situation, significantly improving the prediction accuracy.

[0137] The demand forecasting module provided in this embodiment uses the LSTM model and the Transformer model to predict the water demand for the next day and the next month respectively, which can effectively explore the time series change characteristics of the historical water intake and improve the prediction accuracy. The predicted water demand of each water plant in the next month can provide a basis for the optimization of the scheduling plan for the next month, and the predicted water demand of each water plant on the next day supports the water supply system to update the scheduling plan at a finer time granularity.

[0138] In addition, the above LSTM model and Transformer model can also be replaced by prediction models such as ARIMA model, GRU model or random forest model as needed.

[0139] Embodiment 3

[0140] Based on the multi-water source coordinated scheduling system provided in Example 1, this embodiment takes the water source including reservoir water source and external water source as an example to specifically illustrate the specific architecture of the water supply capacity module.

[0141] In the scenario of coordinated dispatch of multiple water sources, the water supply capacity of a water source refers to the total amount of raw water that a water source can provide to its downstream water plant every day. It is the maximum value of the water supply capacity of a water source under the condition that only the available water volume of the water source is considered. When formulating the water transfer plan for the next day, the water supply capacity of a water source is the total amount of water available for the next day; when formulating the water transfer plan for the next month, the water supply capacity of a water source is the total amount of water available for the next month divided by the total number of days in the next month.

[0142] Considering that there are many types of water sources in a city, including reservoirs, lakes, rivers, groundwater, or water transferred from other cities, and the calculation methods of available water volume of different types of water sources vary greatly. This embodiment is described for two types of water sources, namely reservoir water source and external water source, and does not exclude the inclusion of other types of water sources in this embodiment.

[0143] like Figure 4 As shown, this embodiment provides a multi-water source coordinated scheduling system, including a water supply capacity module.

[0144] The water supply capacity module includes a reservoir water source water supply capacity unit and an external water source water supply capacity unit, which are as follows:

[0145] The reservoir water source water supply capacity unit is used to calculate the water supply capacity of each reservoir water source.

[0146] The water supply capacity of the reservoir water source is mainly calculated based on the storage capacity and water level. When the storage capacity curve is known, the storage capacity can be calculated based on the water level. The guaranteed storage capacity (corresponding to the dead water level) and the basic storage capacity (corresponding to the normal water storage level) are two important indicators for calculating the available water volume. The guaranteed storage capacity refers to the amount of water that the reservoir can store at the lowest water level; the basic storage capacity (BC) refers to the amount of water between the normal operating water level and the lowest water level of the reservoir. This part of the storage capacity is the amount of water that can be adjusted and utilized by the reservoir in normal operation to meet the needs of irrigation, power generation, water supply, etc. In addition, when calculating the water supply capacity, future rainfall, inflow, flood discharge, power generation and unplanned scheduling factors must also be taken into consideration. In the present invention, they are simplified into the future total inflow IC and outflow OC.

[0147] Specifically, the method for the reservoir water source water supply capacity unit provided in this embodiment to calculate the water supply capacity of each reservoir water source includes:

[0148] Based on the basic storage capacity of the reservoir, the current daily water supply, the number of days in the preset target scheduling time interval, and the total inflow and total outflow in the preset target scheduling time interval, the water supply capacity of the reservoir water source is determined according to formula (4);

[0149] (4);

[0150] represents the water supply capacity of the i-th reservoir, T represents the duration of the target scheduling time interval, represents the basic storage capacity of the i-th reservoir, represents the current daily water supply of the i-th reservoir, Indicates the number of days from the current time to the start date of the target scheduling time interval. represents the total inflow of the i-th reservoir within the target scheduling time interval, Represents the total outflow of the i-th reservoir within the target scheduling time interval.

[0151] The external water source supply capacity unit is used to calculate the water supply capacity of each external water source.

[0152] External water sources are usually subject to annual water withdrawal control. If the annual total water withdrawal of the water supply system from external water sources exceeds the water withdrawal quota, additional fees will be required. Therefore, the principle of external water withdrawal is to ensure that the water withdrawal demand of the downstream water plant is met every month, and to ensure that the water withdrawal quota is used up before the end of the year. The dispatch of external water sources is usually carried out by water management personnel who allocate water to each month according to actual conditions to obtain an annual water supply plan.

[0153] This embodiment determines the water supply capacity of the external water source based on the pre-determined monthly allocation coefficient and the annual total water intake, wherein the monthly allocation coefficient is pre-determined based on historical water supply rules.

[0154] Specifically, the method for the external water source supply capacity unit to calculate the water supply capacity of each external water source includes:

[0155] Based on the total annual water supply of the external water source, the water supply volume and the duration of the target scheduling time interval, the water supply capacity of the external water source is determined according to formula (5).

[0156] (5);

[0157] represents the water supply capacity of the i-th external water source, T represents the number of days in the target scheduling time interval, represents the total annual water supply of the i-th external water source, represents the amount of water supplied by the i-th external water source, Represents the allocation coefficient for the mth month.

[0158] Embodiment 4

[0159] Based on the multi-water source coordinated scheduling system provided in embodiments 1 to 3, this embodiment is based on Figure 3 A directed graph constructed based on the topological relationship of a specific water supply system is shown, which specifically illustrates the specific architecture of the multi-objective optimization module.

[0160] like Figure 4 As shown, this embodiment provides a multi-water source coordinated scheduling system, including a multi-objective optimization module.

[0161] The multi-objective optimization module includes a directed graph construction unit and a solution unit, as follows:

[0162] The directed graph construction unit is used to construct a directed graph based on the topological relationship of the water supply system. The directed graph includes nodes and edges. The nodes represent water sources, pumping stations or water plants, and the edges represent transmission pipelines connecting the three.

[0163] For example, in Figure 3 In the directed graph example shown, S1 to S3 represent water sources, P1 to P3 represent water plants, and a pumping station N1 for relay is provided between the water source and the water plant. The pumping station is connected to the water source and the water plant respectively through transmission pipelines.

[0164] In addition, the directed graph also includes the first attribute dictionary, the second attribute dictionary, the third attribute dictionary and the adjacency matrix, as follows:

[0165] The first attribute dictionary is used to store the type and water supply capacity of each water source.

[0166] Specifically, Figure 3 S1 and S3 are reservoir water sources, and S2 is an external water source. After calculating the water supply capacity, the water supply capacity of each water source the next day or the next month is obtained, which are s1, s2, and s3 respectively, in units of 10,000 cubic meters per day. The attribute dictionary of the water source is finally obtained as follows:

[0167] source_attributes = {

[0168] S1: {" type": 0, "capacity": s1},

[0169] S2: {" type": 0, "capacity": s2},

[0170] S3: {"type": 1, "capacity": s3}}.

[0171] Where type indicates the water source type, 0 indicates reservoir water source, and 1 indicates external water source; capacity indicates the water supply capacity of the water source calculated above. Unless otherwise specified, the "water source" mentioned below does not distinguish between water source types.

[0172] The third attribute dictionary is used to store the predicted water demand of each water plant within a preset target scheduling time interval.

[0173] Specifically, Figure 3There are three water plants, namely P1, P2 and P3. The above-mentioned demand forecasting module predicts the predicted water demand for the next month and the next day for each water plant, in units of 10,000 cubic meters per day. For the subsequent scheduling plan calculation, the water demand for the next month is referred to when generating the scheduling plan for the next month, and the water demand for the next day is referred to when generating the scheduling plan for the next day. If there is no special explanation, the water demand of the water plant covers two time scales of the next month / next day. Finally, the water demand of each water plant is obtained, which are d1, d2, and d3 respectively. The attribute dictionary of the water plant can be expressed as:

[0174] plant_attributes = {

[0175] P1: d1,

[0176] P2: d2,

[0177] P3: d3}.

[0178] The adjacency matrix is ​​used to store the connection relationship between any two nodes.

[0179] Corresponds to Figure 3 There are 7 nodes in the directed graph, which correspond to 7 rows and 7 columns in the adjacency matrix. The elements in the matrix represent the connection relationship between the node corresponding to the row and the node corresponding to the column. Then Figure 3 The adjacency matrix corresponding to the connection relationship of the nodes in the directed graph can be expressed as:

[0180] adjacency_matrix = [

[0181] [0, 0, 0, 1, 0, 0, 0],

[0182] [0, 0, 0, 1, 0, 1, 1],

[0183] [0, 0, 0, 0, 0, 0, 1],

[0184] [0, 0, 0, 0, 1, 0, 0],

[0185] [0, 0, 0, 0, 0, 0, 0],

[0186] [0, 0, 0, 0, 0, 0, 0],

[0187] [0, 0, 0, 0, 0, 0, 0]].

[0188] In this adjacency matrix, 1 indicates direct connection, 0 indicates indirect connection, rows represent pipelines from {S1, S2, S3, N1, P1, P2, P3}, and columns represent pipelines flowing into the node. For example, the element at the 4th column of the 1st row is 1, which means that the node S1 corresponding to the row can flow to the node N1 corresponding to the column. There are 6 non-zero elements in the matrix, corresponding to Figure 3 The 6 edges in .

[0189] The second attribute dictionary is used to store the water delivery capacity, water delivery distance and elevation difference of the transmission pipeline associated with each edge. That is, the second attribute dictionary is the attribute dictionary of each edge in the directed graph, which can be expressed as follows, where the key is a node pair, corresponding to Figure 3 The starting and ending points of all edges (transmission pipelines) in the dictionary; the value of the dictionary is the attribute dictionary of the transmission pipeline, including 3 elements: maximum water transmission capacity C, water transmission distance L, and pipeline elevation difference H.

[0190] edge_attributes = {

[0191] (S1, N1): {"C": 30, "L": 40, "H": 2},

[0192] (S2, N1): {"C": 30, "L": 20, "H": 2},

[0193] (S2, P2): {"C": 30, "L": 30, "H": 1},

[0194] (S2, P3): {"C": 20, "L": 60, "H": 1},

[0195] (S3, P3): {"C": 40, "L": 0, "H": 2},

[0196] (N1, P1): {"C": 50, "L": 20, "H": -3}}.

[0197] based on Figure 3From the directed graph, we can see that the edge S3->P3 represents the transmission pipeline that supplies water from reservoir S3 to water plant P3. The maximum water transmission capacity C of the transmission pipeline is 200,000 cubic meters / day. It is convenient for water plant P3 to be built at reservoir S3, so the water transmission distance can be regarded as 0m. The elevation difference between the starting end S3 and the end end P3 of the transmission pipeline is 4m. The raw water flows under the action of gravity without pumping. S2 is an external water source, and S3 is a reservoir water source. The two jointly supply water to the water plant P3. The elevation difference from water sources S1 and S2 to water plant P1 is -1m, so it needs to be lifted by pumping station N1. The water transmission distance of S1->P1 is L=40+20=60m, and the water transmission distance of S2->P1 is L=20+20=40m. Although the sum of the water transmission capacity of S1 and S2 to N1 is 30+30=600,000 cubic meters / day, the maximum water transmission capacity of N1->P1 is 500,000 cubic meters / day, so the actual water supply of the two does not exceed 500,000 cubic meters / day.

[0198] The solving unit is used to optimize and solve the predicted water demand of each water plant, the current water supply capacity of each water source and the topological relationship of the water supply system contained in the above-mentioned directed graph according to preset objective functions and constraints to obtain the optimal scheduled water volume that each transmission pipeline should transport.

[0199] Based on the information contained in the directed graph, a multi-objective optimization algorithm is used to reasonably allocate raw water resources to meet the water plant's water intake needs, reduce water transmission costs, and balance raw water supply. The multi-objective optimization algorithm can specifically be a weighted sum method, an ideal point method, a Pareto optimization algorithm, and the like. This embodiment provides a solution unit that uses a weighted sum method to solve the optimal scheduling water volume. The optimization objectives include reducing water transmission costs and balancing raw water supply. Reducing water transmission costs can be further decomposed into minimizing transmission distance and minimizing pumping energy consumption; balancing raw water supply can be further decomposed into maximizing external water supply and balancing reservoir water supply.

[0200] Specifically, in the solving unit, the objective function is formula (6):

[0201] (6);

[0202] min(·) means to take the minimum value operation. to represents the preset weight parameter, represents the total transmission distance of all water sources, represents the pumping energy consumption, Indicates the supply of external water sources, Represents the supply balance coefficient of reservoir water among all water sources. to Reflection index parameters to The importance can be adjusted according to actual needs.

[0203] In formula (6), to Based on the directed graph, according to formulas (7) to (10), it is determined that:

[0204] (7);

[0205] (8);

[0206] (9);

[0207] (10);

[0208] It represents the dispatched water volume of the pipeline connecting node i and node j in the directed graph within the preset target dispatch time interval. Represents the set of all edges in a directed graph, that is, the set of keys in edge_attributes, represents the water delivery distance between node i and node j; , is an indicator function, which means that when x is less than 0, The value of is 1, when x is greater than or equal to 0, The value of is 0. represents the elevation difference of the pipeline connecting node i and node j. Formula (8) mainly takes into account the gravity flow ( ) does not consume pumping electricity; represents the set of external water sources; R represents the set of reservoir water sources, Represents the water supply capacity of the i-th water source It represents the average value of the ratio of the actual water supply of all reservoir sources to the water supply capacity. Indicates the quantity of water source in the reservoir.

[0209] In the above formulas (7) to (10):

[0210] Formula (7) is combined with formula (6) to minimize the transmission distance.

[0211] Formula (8) is combined with formula (6) to minimize the pumping energy consumption.

[0212] Formula (9) is combined with formula (6) to maximize the external water supply.

[0213] Formula (10) combined with formula (6) is used to balance the water supply of each reservoir and alleviate the contradiction between water resource distribution and economic development pattern.

[0214] In addition to the two optimization goals of transmission distance and pumping energy consumption, the above objective function also sets two goals: maximizing the supply of external water and balancing the supply of reservoir water sources. This enables the algorithm model to reasonably allocate multiple types of raw water resources, alleviate regional supply and demand contradictions, and better meet the actual needs of coordinated scheduling of multiple water sources in urban water supply systems.

[0215] In the solution unit, the constraints include three parts: supply and demand balance of the water plant, water supply capacity limitation of the water source node, and pipeline capacity limitation.

[0216] Specifically, the constraints are formulas (11) to (13):

[0217] (11);

[0218] (12);

[0219] (13);

[0220] P represents the set of all water plants, It represents the dispatched water volume of the pipeline connecting node i and node j in the directed graph within the preset target dispatch time interval. represents the water demand of the jth water plant, represents the water supply capacity of the ith water source, S represents the set of all water sources, It represents the water delivery capacity of the pipeline connecting node i and node j, and E represents the set of all edges in the directed graph.

[0221] The above constraints are as follows in formulas (11) to (13):

[0222] Formula (11) represents the water plant supply and demand balance: For each water plant node j, its water demand must be met .

[0223] Formula (12) represents the water supply capacity balance of water sources: for each water source node i, its maximum water supply capacity cannot be exceeded .

[0224] Formula (13) represents the water delivery capacity limit of the transmission pipeline: the flow rate of each edge cannot exceed the maximum water delivery capacity of the transmission pipeline .

[0225] The above-mentioned solution unit also includes a solver unit, which is used to perform optimization solution based on the above-mentioned objective function and constraint conditions to obtain the optimal scheduling water volume of each transmission pipeline within the preset target scheduling time interval.

[0226] Specifically, the optimization solution method may be an optimization algorithm such as genetic algorithm, simulated annealing, etc. Preferably, this embodiment uses non-dominated sorting genetic algorithm II (NSGA-II) for optimization solution.

[0227] The non-dominated sorting genetic algorithm II first randomly generates an initial population. Each individual in the population represents a potential solution, including decision variables. The value of . Then calculate the value of the objective function of each individual in the population, based on which the population is non-dominated sorted, and they are divided into different fronts according to their non-dominated relationship on the objective function. Generally, an individual is considered non-dominated if it dominates another individual (that is, it is not worse in all objectives and better in at least one objective). Then calculate the crowding of individuals within each non-dominated front. This indicator measures the degree of crowding in the neighborhood around the individual and is used to maintain population diversity. Individuals with higher crowding are more likely to be selected to enter the next generation. A new generation of population is generated through selection, crossover (recombination) and mutation operations, where the selection operation is usually based on non-dominated sorting and crowding. Furthermore, the parent population and the new generation population are merged, and then the next generation population is determined by non-dominated sorting and crowding screening. If the termination condition is met (for example, the maximum number of iterations or convergence criteria is reached), the algorithm stops.

[0228] At the end of the algorithm, the Pareto optimal solution set found is output, which represents the approximate optimal solution to the multi-objective optimization problem. It is worth noting that the result obtained by NSGA-II is not just a single optimal solution, but a set of optimal solutions that balance multiple objectives. Water management personnel can choose the appropriate solution to formulate a scheduling plan based on their preferences for different objectives.

[0229] The embodiments provided by the present invention solve the problems of poor real-time performance and delayed response to emergencies in existing multi-water source scheduling technologies. Based on online data and deep learning models, the present invention can make forward-looking predictions on the water demand of water plants and the water supply capacity of water sources, thereby supporting early warning and dynamic response to emergencies; it comprehensively describes the topological relationship of the water supply system, defines the supply and demand attributes of water sources and water plants based on directed graphs, and defines the maximum water transmission capacity, water transmission distance, and pipeline elevation difference attributes for each transmission pipeline, which can support more systematic optimization scheduling; by reasonably setting multiple optimization targets, it takes into account the various needs of the urban water supply system to meet the supply and demand balance of water plants, take into account the comprehensive allocation of multiple types of water sources, reduce operating costs, and alleviate regional supply and demand contradictions.

[0230] It should be understood by those skilled in the art that the embodiments of the present invention may be provided as methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, EEPROM, optical storage, etc.) containing computer-usable program code.

[0231] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present invention. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions.

[0232] It should be noted that the word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The invention may be implemented by means of hardware comprising a number of different components and by means of a suitably programmed computer. Among the several means listed, several of these means may be embodied by the same hardware. The use of the words first, second, third, etc., is for convenience only and does not indicate any order. These words may be understood as part of the name of the component.

[0233] In addition, it should be noted that, in the description of this specification, the description of the terms "one embodiment", "some embodiments", "embodiment", "example", "specific example" or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are contradictory.

[0234] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments after knowing the basic creative concept. Obviously, those skilled in the art may make various modifications and variations to the present invention without departing from the spirit and scope of the present invention.

Claims

1. A multi-water source coordinated dispatching system for a water supply system, the water supply system comprising at least two water sources and at least one water plant, and a transmission pipeline for connecting the two; characterized in that: The multi-water source coordinated dispatching system comprises: A data acquisition module, used to acquire an initial historical demand sequence of each water plant, the historical demand sequence including the daily historical water intake of the water plant within a preset historical time interval; A feature reconstruction module is used to perform feature reconstruction based on the initial historical demand sequence to obtain a reconstructed historical demand sequence; the reconstructed historical demand sequence includes: a mutation factor feature sequence; the mutation factor feature sequence is used to characterize feature information of a mutation in historical water intake in the initial historical demand sequence; A demand forecasting module, used to forecast the predicted water demand of each water plant within a preset target scheduling time interval based on the reconstructed historical demand sequence; Water supply capacity module, used to calculate the water supply capacity of each water source; A multi-objective optimization module is used to optimize and solve the water demand forecast of each water plant, the water supply capacity of each water source and the topological relationship of the water supply system according to the preset objective function and constraint conditions to obtain the optimal dispatching water volume of each transmission pipeline within the preset target dispatching time interval; The water source includes a reservoir water source and an external water source, and the water supply capacity module includes at least one of a reservoir water source water supply capacity unit and an external water source water supply capacity unit; The reservoir water source water supply capacity unit is used to determine the water supply capacity of the reservoir water source according to formula (4) based on the basic storage capacity of the reservoir, the current daily water supply, the number of days in the preset target scheduling time interval, and the total inflow and total outflow in the preset target scheduling time interval; （4）； represents the water supply capacity of the i-th reservoir, T represents the duration of the target scheduling time interval, represents the basic storage capacity of the i-th reservoir, represents the current daily water supply of the i-th reservoir, Indicates the number of days from the current time to the start date of the target scheduling time interval. represents the total inflow of the i-th reservoir within the target scheduling time interval, represents the total outflow of the i-th reservoir within the target scheduling time interval; The external water source supply capacity unit is used to determine the water supply capacity of the external water source based on the total annual water supply of the external water source, the water supply volume and the duration of the target scheduling time interval according to formula (5); （5）； represents the water supply capacity of the i-th external water source, T represents the number of days in the target scheduling time interval, represents the total annual water supply of the i-th external water source, represents the amount of water supplied by the i-th external water source, Represents the allocation coefficient for the mth month.

2. The multi-water source coordinated dispatching system according to claim 1 is characterized in that: The feature reconstruction module comprises: The first mutation feature construction unit is used to roll-count the cumulative sum of target feature values ​​based on the initial historical demand sequence; and determine the mutation factor feature sequence according to whether the cumulative sum of the target feature values ​​is greater than a preset threshold.

3. The multi-water source coordinated dispatching system according to claim 2 is characterized in that: The target feature value is a rolling mean, and the first mutation feature construction unit includes a rolling statistics subunit, a cumulative sum subunit and a judgment subunit; The rolling statistical subunit is used to determine the rolling mean value corresponding to the daily historical water withdrawal of the initial historical demand sequence within the rolling window according to formula (1) based on the initial historical demand sequence; （1）； represents the rolling mean in the rolling window corresponding to the tth day, w represents the width of the current rolling window, and t represents the right end of the current rolling window. Indicates the left end of the current scroll window. represents the i-th historical water withdrawal in the current rolling window; The cumulative sum subunit is used to determine the daily cumulative sum based on the rolling mean according to formula (2); （2）； represents the cumulative sum corresponding to the tth day, max(·) represents the maximum value operation, represents the historical water withdrawal on day t; The judgment subunit is used to judge whether the corresponding daily cumulative sum is greater than a preset threshold based on the daily cumulative sum; if so, the deviation between the historical water intake corresponding to the day and the rolling mean is used as the mutation factor corresponding to the day, and filled in the position corresponding to the day in the mutation factor feature sequence; if not, 0 is filled in the position corresponding to the day in the mutation factor feature sequence to obtain the mutation factor feature sequence.

4. The multi-water source coordinated dispatching system according to claim 3 is characterized in that: In the first mutation feature construction unit, the step of determining whether the daily corresponding cumulative sum is greater than a preset threshold comprises: The preset threshold is based on a known non-mutated initial historical demand sequence and is predetermined according to formula (3); （3）； represents the preset threshold, k represents the empirical constant, Represents the rolling standard deviation of the known non-mutated initial historical demand series.

5. The multi-water source coordinated dispatching system according to claim 4 is characterized in that: The feature reconstruction module also includes: The second mutation feature construction unit is used to mark the day as a mutation point to form a mutation point indication sequence when the first mutation feature construction unit determines that the corresponding cumulative sum is greater than a preset threshold.

6. The multi-water source coordinated dispatching system according to any one of claims 2 or 5, characterized in that: The feature reconstruction module also includes a multi-dimensional feature reconstruction unit and a splicing unit; The multidimensional feature reconstruction unit is used to decompose the initial historical demand sequence by using a seasonal and trend decomposition method based on local moving average to obtain a trend feature sequence, a seasonal feature sequence and a residual feature sequence; The splicing unit is used to splice the mutation factor feature sequence, the mutation point indication sequence, the trend feature sequence, the season feature sequence or the residual feature sequence to obtain a reconstructed historical demand sequence.

7. The multi-water source coordinated dispatching system according to claim 1, characterized in that: The demand forecasting module includes at least one of a next-month forecasting unit and a next-day forecasting unit; The next month prediction unit is used to predict the predicted water demand of each water plant in the next month based on the first prediction model and the reconstructed historical demand sequence; the first prediction model is an LSTM model with adapted model parameters; The next-day prediction unit is used to predict the predicted water demand of each water plant on the next day based on the second prediction model and the reconstructed historical demand sequence; the second prediction model is a Transformer model with adapted model parameters.

8. The multi-water source coordinated dispatching system according to claim 1, characterized in that: The multi-objective optimization module includes: A directed graph construction unit, used to construct a directed graph based on the topological relationship of the water supply system, wherein the directed graph includes nodes and edges, wherein the nodes represent water sources or water plants, the edges represent transmission pipelines connecting the two, and the direction of the edges represents the transmission direction of the transmission pipelines; The solving unit is used to optimize and solve the water volume that each transmission pipeline should transport based on the predicted water demand, the current water supply capacity of the water source and the topological relationship of the water supply system according to the preset objective function and constraints.

9. The multi-water source coordinated dispatching system according to claim 8, characterized in that: In the solving unit, the optimization solution is performed according to the preset objective function and constraint conditions, including: The objective function is formula (6): （6）； min(·) means to take the minimum value operation. to represents the preset weight parameter, represents the total transmission distance of all water sources, represents the pumping energy consumption, Indicates the supply of external water sources, It represents the supply balance coefficient of reservoir water among all water sources. to Based on the directed graph, according to formulas (7) to (10), it is determined that: （7）； （8）； （9）； （10）； It represents the dispatched water volume of the pipeline connecting node i and node j in the directed graph within the preset target dispatch time interval. represents the set of all edges in a directed graph, represents the water transport distance between node i and node j, , is an indicator function, which means that when x is less than 0, The value of is 1, when x is greater than or equal to 0, The value of is 0. It represents the elevation difference of the pipeline connecting node i and node j. represents the set of external water sources, R represents the set of reservoir water sources, represents the water supply capacity of the ith water source, It represents the average value of the ratio of the actual water supply of all reservoir sources to the water supply capacity. Indicates the quantity of water source in the reservoir; The constraints are formulas (11) to (13): （11）； （12）； （13）； P represents the set of all water plants, It represents the dispatched water volume of the pipeline connecting node i and node j in the directed graph within the preset target dispatch time interval. represents the water demand of the jth water plant, represents the water supply capacity of the ith water source, S represents the set of all water sources, It represents the water delivery capacity of the pipeline connecting node i and node j, and E represents the set of all edges in the directed graph.

Images