Multi-target coordinated urban rainwater utilization system

Abstract

This invention discloses a multi-objective collaborative urban rainwater utilization system, comprising: an environmental acquisition module for constructing a rainwater state fusion matrix; a zoning modeling module for performing rainwater functional zoning identification operations, constructing a rainwater functional zoning set, and generating a storage unit state set; a water quality treatment module for classifying water quality levels and generating a water quality classification result set; a trend prediction module for constructing a multivariate time series and generating storage and discharge trend prediction results through an improved TSMixer model; and a regulation and optimization module for performing scheduling optimization operations, constructing a rainwater utilization evaluation parameter set, generating scheduling feedback factors, and updating scheduling paths and execution instruction sets. This invention improves the resource utilization efficiency of storage units and the intelligent level of system response, possessing high model robustness, strong dynamic adaptability, and accurate scheduling path optimization.

Description

Technical Field

[0001] This invention relates to the field of urban rainwater management and resource utilization technology, and in particular to a multi-objective collaborative urban rainwater utilization system. Background Technology

[0002] With the continuous advancement of urbanization and the frequent occurrence of extreme weather events, the efficient utilization and scientific regulation of urban rainwater resources have become important issues in urban water resource management. Most existing urban stormwater utilization systems employ a single storage or discharge strategy, lacking collaborative modeling of multiple factors such as dynamic changes in stormwater, water quality status, and the operational load of storage units. This results in poor system scheduling flexibility and high response delays when facing sudden heavy rainfall or continuous rainy weather, easily leading to problems such as urban flooding, drainage overflows, or stormwater waste. Existing water quality treatment strategies mostly rely on static classification rules, failing to dynamically identify and classify stormwater sources, temporal characteristics, and water storage targets. This results in some reusable stormwater being misjudged and discharged, or uneven treatment loads leading to resource waste. In terms of trend prediction, traditional models often rely on univariate or low-dimensional sequence regression methods, which are difficult to characterize the nonlinear interactions between the pressure, water level, and water quality states of storage units, leading to insufficient prediction accuracy and difficulty in supporting real-time optimization and control needs. In addition, existing control strategies are mostly based on empirical rules to set thresholds for trigger control, lacking a dynamic trade-off mechanism between multiple objectives such as flood control safety, water quality compliance, and stormwater reuse rate, affecting the overall system efficiency and response robustness.

[0003] Therefore, how to provide a multi-objective collaborative urban rainwater utilization system is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0004] One objective of this invention is to propose a multi-objective collaborative urban rainwater utilization system. This invention constructs a multi-module collaborative structure including an environmental acquisition module, a zoning modeling module, a water quality treatment module, a trend prediction module, and a regulation and optimization module. It describes in detail the entire process based on rainwater state fusion matrix construction, rainwater functional zoning identification, water quality level classification, trend prediction modeling, and scheduling optimization feedback linkage. It makes full use of the storage unit state set, water quality classification result set, and improved TSMixer model to realize the dynamic scheduling and efficient utilization of urban rainwater under multi-objective conditions. It has the advantages of fine rainwater resource identification, accurate storage and discharge trend prediction, adaptive scheduling decision-making, and multi-path matching optimization.

[0005] A multi-objective collaborative urban rainwater utilization system according to an embodiment of the present invention includes: The environmental data acquisition module is used to collect rainfall, surface runoff, land permeability, groundwater level, water quality parameters and microclimate parameters in urban areas, and to construct a rainwater status fusion matrix. The zoning modeling module is used to perform rainwater functional zoning identification based on the rainwater state fusion matrix and urban plot planning information. It generates functional labels according to land use type, storage and drainage capacity level and water quality target, and constructs a rainwater functional zoning set. It also deploys storage and regulation units, collects capacity data and drainage control data of storage and regulation units, constructs a distributed storage and drainage structure and generates a storage and regulation unit state set. The water quality treatment module is used to classify water quality levels based on the rainwater state fusion matrix, and match green space irrigation paths, treatment paths and discharge paths according to the water quality levels to generate a water quality classification result set; the water quality levels include irrigable rainwater, rainwater requiring treatment and unusable rainwater; The trend prediction module is used to construct multivariate time series based on the storage unit state set, rainwater functional zoning set, and water quality classification result set, and generate storage and discharge trend prediction results through the improved TSMixer model. The regulation and optimization module is used to perform scheduling optimization operations based on the storage and discharge trend prediction results, generate scheduling paths and execution instruction sets; construct a rainwater utilization evaluation parameter set, generate scheduling feedback factors, and if the scheduling feedback factors meet preset update conditions, update the weight parameters in the scheduling optimization operation and generate updated scheduling paths and execution instruction sets.

[0006] Preferably, the steps between modules include the following: S1. Collect rainfall, surface runoff, land permeability, groundwater level, water quality parameters and microclimate parameters in urban areas to construct a rainwater status fusion matrix; S2. Based on the rainwater status fusion matrix and urban plot planning information, perform rainwater functional zoning identification operation, generate functional labels according to land use type, storage and drainage capacity level and water quality target, and construct rainwater functional zoning set; S3. Deploy storage and regulation units according to the rainwater functional zoning set, collect capacity data and drainage control data of the storage and regulation units, construct a distributed storage and drainage structure and generate a storage and regulation unit status set; S4. Based on the rainwater state fusion matrix, classify the water quality levels, and match the green space irrigation path, treatment path and discharge path according to the water quality level to generate a water quality classification result set; the water quality level includes irrigable rainwater, rainwater that needs to be treated and unusable rainwater; S5. Construct a multivariate time series based on the storage unit state set, rainwater functional zoning set, and water quality classification result set, and generate storage and discharge trend prediction results through the improved TSMixer model; S6. Based on the predicted storage and discharge trends, perform scheduling optimization operations to generate scheduling paths and execution instruction sets; S7. Construct a rainwater utilization evaluation parameter set, generate a scheduling feedback factor, and if the scheduling feedback factor meets the preset update conditions, update the weight parameters in the scheduling optimization operation and generate the updated scheduling path and execution instruction set.

[0007] Preferably, S1 specifically comprises: Rainfall sensors are installed within urban areas to record rainfall intensity and duration, generating rainfall data; surface flow monitoring devices record runoff velocity and flow rate changes, generating surface runoff; based on the surface structure, cover type, and soil type of a plot, infiltration characteristic indicators are determined, generating plot permeability; groundwater level sensors measure groundwater level changes, generating groundwater level; water sample parameter detection devices obtain pH value, conductivity, turbidity, suspended solids concentration, and pollutant concentration, generating water quality parameters; and meteorological measurement devices record temperature, humidity, wind speed, and air pressure values, generating microclimate parameters. Rainfall, surface runoff, land permeability, groundwater level, water quality parameters, and microclimate parameters are processed with time alignment and structural uniformity to construct a rainwater state fusion matrix. The rows of the rainwater state fusion matrix correspond to time series sampling points, and the columns correspond to rainfall, surface runoff, land permeability, groundwater level, water quality parameters, and microclimate parameters. Each cell corresponds to a standardized data value under a specified time and a specified indicator.

[0008] Preferably, S2 specifically includes: Based on rainfall, surface runoff, plot permeability, groundwater level, water quality parameters and microclimate parameters in the rainwater state fusion matrix, and combined with plot use data and plot boundary data, a set of plot spatial units is established. For each land parcel spatial unit, the land use type is identified, and a land use type label is generated; the land use types include residential land, industrial land, commercial land, public service land, transportation land, and green space. Based on the land permeability and surface runoff, the storage and drainage capacity values ​​are calculated, and the land is classified into high storage and low drainage, medium storage and medium drainage, and low storage and high drainage according to the set grade standards, and storage and drainage capacity grade labels are generated. Based on the pollutant index values ​​in the water quality parameters and the regional water quality requirements, the water quality target labels are generated by dividing the target into primary, secondary and tertiary levels according to the set intervals. By associating land use type labels, water storage and drainage capacity level labels, and water quality target labels with corresponding land parcel spatial units, a functional label structure is constructed. Land parcel spatial units with the same functional label structure and spatially adjacent to each other are merged to form rainwater functional zones, generating a rainwater functional zone set.

[0009] Preferably, S3 specifically includes: A storage and regulation unit is set up in each rainwater functional zone; the storage and regulation unit includes one or more structural combinations of rainwater collection tanks, infiltration ponds, constructed wetlands, modular water storage structures and underground storage and regulation devices; Record the structural volume parameters, actual water storage upper limit and historical water storage records of each regulation and storage unit to generate capacity data; Configure the drainage control components of the storage unit, record the status of the outlet valve, the discharge frequency and the discharge flow rate, and generate discharge control data; construct a time series structure of capacity data and discharge control data for each storage unit, and generate the storage unit state vector; Spatial mapping is performed on the state vectors of all storage units to construct a distributed storage and drainage structure corresponding to the rainwater functional zoning set; the state vectors of all storage units are integrated to generate a storage unit state set.

[0010] Preferably, the water quality classification based on the rainwater state fusion matrix specifically includes: Water quality parameter data are extracted from the rainwater state fusion matrix; the water quality parameter data includes pH value, conductivity, turbidity, suspended solids concentration and pollutant concentration; Construct a water quality grade determination range and set multi-dimensional boundary parameters for irrigable rainwater grade, rainwater requiring treatment grade, and unusable rainwater grade; Based on the extracted water quality index data and water quality grade determination intervals, each time series sample in the state fusion matrix is ​​labeled with a grade according to the rule-based determination method to generate a water quality grade label set; the water quality grade label set is combined with the corresponding time index to construct the water quality grade time series result.

[0011] Preferably, the step of matching green space irrigation paths, treatment paths, and discharge paths according to water quality levels to generate a water quality classification result set specifically involves: Based on the records marked as irrigable rainwater in the water quality grade time series results, and according to the principles of functional zoning concentration, irrigation radius and shortest distance to storage unit, the corresponding green space irrigation paths are selected to generate an irrigation path subset; Based on the records marked as rainwater requiring treatment, the matching priority is calculated according to the capacity utilization rate of the storage unit, the load status of the water treatment node, and the pipeline network bearing capacity coefficient. Treatment paths are selected and a subset of treatment paths is generated. Based on records marked as unusable rainwater, emission paths are selected according to emission channel level, emission threshold and regional emission risk level, and a subset of emission paths is generated; The irrigation path subset, treatment path subset, and discharge path subset are time-indexed and structurally concatenated to construct a water quality classification result set.

[0012] Preferably, the improved TSMixer model is as follows: An input structure is constructed by concatenating the state set of the storage unit, the rainwater functional zoning set, and the water quality classification result set according to a unified time index to form an input sequence tensor. In the input sequence tensor, rows represent time sampling points, and columns represent standardized values ​​of storage unit water storage capacity, storage unit discharge frequency, storage unit discharge rate, functional zoning type, water quality level, and microclimate index. A bidirectional hybrid structure is constructed, and linear hybridization processing is performed on the input sequence tensor along both the variable dimension and the time dimension to generate a variable hybrid vector and a time hybrid vector. The variable hybrid vector and the time hybrid vector are then weighted and fused to form a hybrid residual path. The hybrid residual path is then normalized to obtain a fused vector. A hierarchical trend structure is constructed, and multi-scale convolution processing is performed based on the fusion vector to extract short-cycle trend vectors, medium-cycle trend vectors and long-cycle trend vectors respectively; the three types of trend vectors are concatenated according to their dimensions to form a trend combination vector. An embedded structure for regulation targets is constructed, and a target weight vector is generated based on the priority of functional zoning labels, the frequency of water quality jumps, and the pressure index of the storage unit. The trend combination vector is dynamically weighted according to the target weight vector to obtain a weighted trend vector. A linear prediction structure is constructed, the weighted trend vector is input into the linear mapping path, and the storage and discharge trend prediction results are output; the storage and discharge trend prediction results include the water storage trend value, discharge trend value and pressure prediction value of the regulation and storage unit in the future time period.

[0013] Preferably, S6 specifically includes: Based on the water storage trend value, discharge trend value and pressure prediction value of the regulation and storage unit in the water storage and discharge trend prediction results, a scheduling input sequence is constructed; and a scheduling target vector is set according to the water storage benefit target, reuse target, water quality target and discharge risk target. Based on the water storage trend value and water storage benefit target of the storage unit, water storage regulation judgment is performed to generate water storage regulation sub-vector; based on the emission trend value and emission risk target, emission regulation judgment is performed to generate emission regulation sub-vector; based on the pressure prediction value and pressure parameter of the storage unit, pressure regulation judgment is performed to generate pressure regulation sub-vector; the water storage regulation sub-vector, emission regulation sub-vector and pressure regulation sub-vector are concatenated to generate regulation fusion vector; Based on the regulation fusion vector and rainwater functional zoning labels, storage unit spatial location index, and corresponding green space irrigation paths, treatment paths, and discharge paths in the water quality classification results, a path candidate set is constructed; the matching degree value of each path is calculated according to the scheduling target vector; priority paths are selected based on the matching degree values ​​to generate scheduling paths; Based on the scheduling path, the status of the drainage control components of the storage unit is matched, and valve opening instructions, discharge frequency instructions, and discharge flow rate instructions are generated. Based on the scheduling path, the water storage structure parameters of the storage unit are matched, and water storage regulation instructions are generated. The valve opening instructions, discharge frequency instructions, discharge flow rate instructions, and water storage regulation instructions are combined to form an execution instruction set.

[0014] Preferably, S7 specifically includes: The system collects real-time water storage records from the storage unit, the reuse amount output from the storage unit to the green space irrigation path, the treated water output from the water treatment path, and the actual irrigation amount from the green space irrigation path to construct a rainwater utilization assessment parameter set. The rainwater utilization assessment parameter set includes four types of assessment indicators: water storage utilization rate, reuse achievement rate, water quality treatment efficiency, and irrigation satisfaction. Based on the difference between the evaluation indicators and the scheduling targets, evaluation difference vectors are constructed respectively; water storage deviation factor, reuse deviation factor, water quality deviation factor and irrigation deviation factor are calculated according to the set feedback rules, and then concatenated into scheduling feedback factor. The scheduling feedback factor is compared with the matching degree weight vector in the previous scheduling path to determine whether the preset update conditions are met. If they are met, the weight vector is proportionally updated to construct the updated scheduling target vector. Based on the updated scheduling target vector, the scheduling path selection operation is re-executed to construct the updated scheduling path; the valve opening command, discharge frequency command, discharge flow rate command and water storage control command are adjusted synchronously to generate the updated execution command set.

[0015] The beneficial effects of this invention are: This invention addresses the problems of fragmented data collection, coarse-grained functional zoning, ambiguous water quality identification, and non-adaptive scheduling paths in existing urban rainwater utilization systems by constructing a collaborative structure comprising an environmental acquisition module, a zoning modeling module, a water quality treatment module, a trend prediction module, and a regulation and optimization module. It employs unified time alignment and structural processing to construct a rainwater state fusion matrix, combines land use type, storage and drainage capacity level, and water quality targets to construct a rainwater functional zoning set, and collects storage unit capacity data and drainage control data in a distributed storage and drainage structure to generate a storage unit state set. During water quality identification, it sets water quality level judgment intervals, constructs water quality level time series results, and matches green zones according to the levels. The system identifies irrigation, treatment, and discharge paths to generate a water quality classification result set. In trend modeling, an improved TSMixer model is introduced, fusing storage unit state sets, rainwater functional zoning sets, and water quality classification result sets to construct a multivariate time series. Based on a bidirectional hybrid structure and multi-scale convolutional structure, a regulation target embedding structure is introduced to generate storage and discharge trend prediction results. In the scheduling phase, a scheduling target vector is constructed. Based on the trend prediction results, water storage, discharge, and pressure regulation sub-vectors are concatenated to perform path matching and command set generation. A rainwater utilization evaluation parameter set is also constructed, and a scheduling feedback factor is built by setting feedback rules to achieve dynamic updates of weight parameters and optimization of scheduling paths. Ultimately, this system achieves accurate water quality classification, refined functional zoning matching, efficient trend prediction, and adaptive path scheduling for urban rainwater, improving the intelligence, accuracy, and sustainable operation capability of the rainwater utilization system. Attached Figure Description

[0016] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a block diagram of a multi-objective collaborative urban rainwater utilization system proposed in this invention; Figure 2 Here is a flowchart of a multi-objective collaborative urban rainwater utilization system proposed in this invention; Figure 3 This is a schematic diagram of the structure of the improved TSMixer model proposed in this invention; Figure 4 This is a data flow diagram of a multi-objective collaborative urban rainwater utilization system proposed in this invention. Detailed Implementation

[0017] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0018] refer to Figure 1A multi-objective collaborative urban rainwater utilization system includes: The environmental data acquisition module is used to collect rainfall, surface runoff, land permeability, groundwater level, water quality parameters and microclimate parameters in urban areas, and to construct a rainwater status fusion matrix. The zoning modeling module is used to perform rainwater functional zoning identification based on the rainwater state fusion matrix and urban plot planning information. It generates functional labels according to land use type, storage and drainage capacity level and water quality target, and constructs a rainwater functional zoning set. It also deploys storage and regulation units, collects capacity data and drainage control data of storage and regulation units, constructs a distributed storage and drainage structure and generates a storage and regulation unit state set. The water quality treatment module is used to classify water quality levels based on the rainwater state fusion matrix, and match green space irrigation paths, treatment paths and discharge paths according to the water quality levels to generate a water quality classification result set; the water quality levels include irrigable rainwater, rainwater requiring treatment and unusable rainwater; The trend prediction module is used to construct multivariate time series based on the storage unit state set, rainwater functional zoning set, and water quality classification result set, and generate storage and discharge trend prediction results through the improved TSMixer model. The regulation and optimization module is used to perform scheduling optimization operations based on the storage and discharge trend prediction results, generate scheduling paths and execution instruction sets; construct a rainwater utilization evaluation parameter set, generate scheduling feedback factors, and if the scheduling feedback factors meet preset update conditions, update the weight parameters in the scheduling optimization operation and generate updated scheduling paths and execution instruction sets.

[0019] In this implementation, an environmental data acquisition module collects multi-source environmental data, including rainfall, surface runoff, land permeability, groundwater level, water quality parameters, and microclimate parameters, to construct a rainwater status fusion matrix. This enables comprehensive perception of urban rainwater conditions and enhances the system's ability to detect dynamic rainwater processes. A zoning modeling module identifies functional zones based on the fusion matrix and land planning information. Functional labels are generated according to land use type, storage and drainage capacity level, and water quality targets, constructing a rainwater functional zone set. Simultaneously, storage and regulation units are deployed and a distributed storage and drainage structure is constructed, generating a storage and regulation unit status set, thus improving the structured and zonal control capabilities of rainwater management. Furthermore, a water quality treatment module classifies water quality based on the fusion matrix and matches irrigation, treatment, and discharge paths according to different water quality levels, generating a water quality classification result set. This enables precise graded utilization of rainwater resources, improving water utilization efficiency and safety. Furthermore, by setting up a trend prediction module, a multivariate time series model is constructed based on storage and regulation status, water quality classification, and functional zoning. An improved TSMixer model is then used to generate storage and discharge trend prediction results, effectively improving the modeling capability and prediction accuracy for future rainwater evolution. Simultaneously, by setting up a regulation and optimization module, scheduling optimization is performed based on the storage and discharge trend results, generating scheduling paths and instruction sets. A rainwater utilization evaluation parameter set is also constructed to generate scheduling feedback factors, achieving dynamic feedback and optimization updates in the scheduling process. This further enhances the system's collaborative regulation efficiency and adaptive response capability under complex rainwater scenarios. This implementation constructs a multi-objective urban rainwater utilization system with a clear structure, intelligent response, and accurate predictions, significantly enhancing the full-cycle regulation capability and comprehensive utilization level of rainwater resources.

[0020] refer to Figure 2-3 In this embodiment, the steps between modules include the following: S1. Collect rainfall, surface runoff, land permeability, groundwater level, water quality parameters and microclimate parameters in urban areas to construct a rainwater status fusion matrix; S2. Based on the rainwater status fusion matrix and urban plot planning information, perform rainwater functional zoning identification operation, generate functional labels according to land use type, storage and drainage capacity level and water quality target, and construct rainwater functional zoning set; S3. Deploy storage and regulation units according to the rainwater functional zoning set, collect capacity data and drainage control data of the storage and regulation units, construct a distributed storage and drainage structure and generate a storage and regulation unit status set; S4. Based on the rainwater state fusion matrix, classify the water quality levels, and match the green space irrigation path, treatment path and discharge path according to the water quality level to generate a water quality classification result set; the water quality level includes irrigable rainwater, rainwater that needs to be treated and unusable rainwater; S5. Construct a multivariate time series based on the storage unit state set, rainwater functional zoning set, and water quality classification result set, and generate storage and discharge trend prediction results through the improved TSMixer model; S6. Based on the predicted storage and discharge trends, perform scheduling optimization operations to generate scheduling paths and execution instruction sets; S7. Construct a rainwater utilization evaluation parameter set, generate a scheduling feedback factor, and if the scheduling feedback factor meets the preset update conditions, update the weight parameters in the scheduling optimization operation and generate the updated scheduling path and execution instruction set.

[0021] In this embodiment, S1 specifically refers to: Rainfall sensors are installed within urban areas to record rainfall intensity and duration, generating rainfall data; surface flow monitoring devices record runoff velocity and flow rate changes, generating surface runoff; based on the surface structure, cover type, and soil type of a plot, infiltration characteristic indicators are determined, generating plot permeability; groundwater level sensors measure groundwater level changes, generating groundwater level; water sample parameter detection devices obtain pH value, conductivity, turbidity, suspended solids concentration, and pollutant concentration, generating water quality parameters; and meteorological measurement devices record temperature, humidity, wind speed, and air pressure values, generating microclimate parameters. Rainfall, surface runoff, land permeability, groundwater level, water quality parameters, and microclimate parameters are processed with time alignment and structural uniformity to construct a rainwater state fusion matrix. The rows of the rainwater state fusion matrix correspond to time series sampling points, and the columns correspond to rainfall, surface runoff, land permeability, groundwater level, water quality parameters, and microclimate parameters. Each cell corresponds to a standardized data value under a specified time and a specified indicator.

[0022] In this embodiment, S2 specifically refers to: Based on rainfall, surface runoff, land permeability, groundwater level, water quality parameters, and microclimate parameters in the rainwater state fusion matrix, combined with land use data and land boundary data, a spatial overlay processing operation is performed to generate a set of land spatial units. The spatial overlay processing operation includes matching the coordinates of various parameters in the rainwater state fusion matrix with the land boundary, and generating independent spatial units according to the land boundary contour. For each land parcel spatial unit, the land use type is identified. Based on the classification tags in the land parcel use data, residential land, industrial land, commercial land, public service land, transportation land and green space are matched to generate a land use type tag. The land use type tag corresponds one-to-one with the land parcel spatial unit, forming a land use type mapping table. The storage and drainage capacity values ​​are calculated based on the land permeability and surface runoff. The storage and drainage capacity values ​​are generated using a linear combination model. The parameter weights of the linear combination model are obtained by fitting historical rainfall and runoff observation data. The storage and drainage capacity values ​​are divided into high storage and low drainage, medium storage and medium drainage, and low storage and high drainage according to the set level standards, and storage and drainage capacity level labels are generated. Based on the pollutant index values ​​in the water quality parameters and the regional water quality requirements, a pollutant index vector is constructed. The pollutant index vector includes five dimensions: pH value, chemical oxygen demand, total phosphorus, total nitrogen and dissolved oxygen. According to the correspondence of the index values ​​in the set grade interval, it is divided into primary targets, secondary targets and tertiary targets, and water quality target labels are generated. The land use type label, water storage and drainage capacity level label, and water quality target label are associated with corresponding land parcel spatial units to construct a functional label structure. The functional label structure consists of a land use type field, a water storage and drainage capacity level field, and a water quality target level field, and is bound to a unique land parcel spatial unit code to form a three-dimensional label mapping table. Spatial units with the same functional label structure and spatial adjacency are merged. The spatial adjacency relationship is calculated based on the overlap rate of the plot boundaries and the adjacency matrix. Spatial units that meet the set adjacency threshold constitute an aggregate unit. All aggregate units are used as rainwater response areas with consistent functions and similar attributes to generate rainwater functional zones, forming a rainwater functional zone set.

[0023] In this embodiment, S3 specifically refers to: A storage and regulation unit is set up in each rainwater functional zone; the storage and regulation unit includes one or more structural combinations of rainwater collection tanks, infiltration ponds, constructed wetlands, modular water storage structures and underground storage devices; each storage and regulation unit is assigned a unique structural number and bound to the rainwater functional zone identifier to establish a structural mapping relationship; The structural volume parameters of each water storage unit are collected and recorded, including the total design volume, effective water storage volume, and bottom unusable volume. The theoretical upper limit water level and minimum discharge water level are calculated based on the installation elevation and outlet location to generate structural boundary data. Water storage change data after each rainfall event are collected and recorded, including the initial water level, maximum water level, final remaining water level, and time node, to construct a historical water storage time series and generate actual water storage record data. Based on the structural boundary data and actual water storage record data, the dynamic effective water storage rate of each water storage unit is calculated. The remaining usable water storage is standardized to the [0, 1] interval using a proportional calculation formula to generate capacity data. Configure the drainage control component of the storage unit, collect information on the status changes of the outlet valve, record the time, opening angle or control voltage, duration and flow velocity of the outlet measuring point for each valve opening and closing, and generate discharge behavior records; perform statistical analysis on the discharge frequency based on the discharge behavior records, calculate the average drainage volume per unit time by combining the drainage flow velocity and duration of each discharge, construct the discharge frequency sequence and discharge flow velocity sequence, and generate discharge control data; For each storage unit, a joint time series structure of capacity data and drainage control data is constructed, aligned according to a unified time index, and a storage unit state vector is generated. The state vectors of all storage units are spatially mapped to the rainwater functional zone identifier according to their spatial location to generate a distributed storage and drainage structure. An aggregation operation is performed on the state vectors of storage units in each rainwater functional zone to integrate all storage unit state vectors and generate a storage unit state set.

[0024] In this embodiment, the water quality classification based on the rainwater state fusion matrix specifically includes: Water quality parameter data are extracted from the rainwater state fusion matrix; the water quality parameter data includes pH value, conductivity, turbidity, suspended solids concentration and pollutant concentration; Construct water quality grade determination ranges. Based on urban area water environment requirements and rainwater reuse standards, set multi-dimensional boundary parameters for irrigable rainwater grade ranges, rainwater grade ranges requiring treatment, and unusable rainwater grade ranges according to pollutant concentration thresholds, conductivity gradient ranges, turbidity upper limit values, and pH ranges. Based on historical rainwater samples and water quality monitoring records, the multidimensional boundary parameters are calibrated by interval division fitting method so that each level interval meets the stability requirements of sample distribution. Based on the extracted water quality parameter data, the pH value, conductivity, turbidity, suspended solids concentration and pollutant concentration of each time series sample are compared with multidimensional boundary parameters one by one, and corresponding water quality grade labels are generated through rule-based judgment to construct a water quality grade label set. The water quality grade label set is associated with the corresponding time index to generate water quality grade time series results.

[0025] In this embodiment, the step of matching green space irrigation paths, treatment paths, and discharge paths according to water quality levels to generate a water quality classification result set specifically involves: Based on the time series results of water quality grades, time series samples marked as irrigable rainwater are extracted, and their corresponding time indices, functional zone identifiers, and storage unit location indices are obtained to construct an irrigable sample set. For each irrigable sample, based on a preset green space zoning mapping relationship, connectable green space nodes are selected, and the irrigation radius, functional zone concentration, and path distance between the storage unit and the green space node are calculated. According to the path combination with the minimum path distance, the maximum irrigation radius coverage, and the highest functional zone concentration, the corresponding green space irrigation path is matched to generate an irrigation path subset. The path distance is calculated using the Euclidean formula, and the functional zone concentration is calculated by fitting the number of irrigable green spaces per unit area. Based on the time series results of water quality grades, time series samples marked as rainwater requiring treatment are extracted to construct a set of samples requiring treatment. For each sample requiring treatment, the corresponding storage unit state vector is extracted, and the storage capacity utilization rate and effluent status are calculated. Load status data of the current water treatment node are collected, and the remaining treatment capacity of the current node is fitted based on the historical load sequence. Combined with the carrying capacity coefficient of the regional drainage network, the matching priority of each reachable path is comprehensively evaluated through a weighted scoring function. The treatment path is selected according to the path with the highest score, and a subset of treatment paths is generated. Optionally, in the scoring function, the weight of the storage capacity utilization rate is set to 0.4, the weight of the water treatment node load status is set to 0.35, and the weight of the network carrying capacity coefficient is set to 0.25. Based on the time series results of water quality grades, time series samples marked as unusable rainwater are extracted to construct an unusable sample set. For each unusable sample, its corresponding discharge channel grade, historical discharge threshold, and regional discharge risk level are extracted. A set of reachable paths for discharge channels is constructed, and the discharge safety score for each channel is calculated. The discharge safety score is obtained by fitting a weighted model that sets the discharge threshold compliance rate, historical exceedance frequency, and regional risk level. The discharge channel path with the highest score is selected to generate a subset of discharge paths. The irrigation path subset, treatment path subset, and discharge path subset are time-mapped according to their corresponding time indices; a structure splicing module is constructed to splice the three types of path subsets to form a water quality classification result set with unified structure and time alignment.

[0026] In this embodiment, the improved TSMixer model is specifically as follows: An input structure is constructed by concatenating the state set of the water storage unit, the rainwater functional zoning set, and the water quality classification result set according to a unified time index to form an input sequence tensor. The rows of the input sequence tensor represent time sampling points, and the columns represent standardized values ​​of water storage capacity, discharge frequency, discharge rate, functional zoning type, water quality level, and microclimate index of the water storage unit. The standardized values ​​are linearly transformed through a max-min normalization function to ensure that data from different sources have a unified scale and improve the model fusion accuracy. A bidirectional mixing structure is constructed. A one-dimensional linear mixing operation is performed on the input sequence tensor along the variable dimension to extract the interrelationships between various features and generate a variable mixing vector. A one-dimensional linear mixing operation is also performed on the input sequence tensor along the time dimension to extract trend change features within the time series and generate a time mixing vector. The variable mixing vector and the time mixing vector are then weighted and fused according to set mixing weights to form a mixed residual path. The mixing weights are obtained by fitting an attention function trained using gradient descent. The mixed residual path is then normalized to generate a fusion vector, thereby mitigating the numerical offset problem after mixing at different scales and improving model stability. A hierarchical trend structure is constructed, and convolutional paths with different kernel sizes are input in parallel to the fused vector to extract short-cycle trend vectors, medium-cycle trend vectors, and long-cycle trend vectors respectively. The convolutional kernel size is set to three scales: less than 1 / 2 of the set sampling period, approximately equal to the sampling period, and greater than the sampling period, to fit short-term fluctuations, medium-term evolution, and long-term trend changes respectively. The three types of trend vectors are concatenated along the channel dimension to generate a trend combination vector, which improves the model's ability to perceive periodic changes. An embedded structure for regulation targets is constructed, generating a target weight vector based on the functional zoning label priority, water quality jump frequency, and storage unit pressure index. The functional zoning label priority is divided into four levels according to a set rule, corresponding to core green space, secondary green space, general road green space, and low priority area, with priority values ​​of 4, 3, 2, and 1, respectively. The water quality jump frequency is calculated by statistically analyzing the number of water quality level jumps within the last five time windows and normalized to the interval [0, 1]. The storage unit pressure index is the average pressure change rate per unit time, which is transformed using a normal distribution function to generate regulation influencing factors. The above three factors are fitted using a weighted summation function to generate a target weight vector. A dynamic weighting operation is performed on the trend combination vector according to the target weight vector to generate a weighted trend vector, improving the model's response sensitivity in important functional areas. A linear prediction structure is constructed, and a weighted trend vector is input into the linear mapping path to output the water storage and discharge trend prediction results. The water storage and discharge trend prediction results include the water storage trend value, discharge trend value and pressure prediction value of the regulation and storage unit in the future time period. The time period length and prediction step size corresponding to the prediction value are set according to the historical data sampling interval and the functional area regulation and response cycle. A weight sharing mechanism is adopted in the linear mapping path to improve parameter utilization and reduce model redundancy.

[0027] This implementation method achieves bidirectional fusion of multi-source heterogeneous data, hierarchical modeling of multi-period trends, and weighted output with multi-objective task orientation through an improved TSMixer model. While realizing precise prediction of water storage and discharge trends, it improves the model's prediction sensitivity to sudden water quality jumps and high-priority areas, and has technical effects such as high accuracy, high responsiveness and high stability.

[0028] In this embodiment, the improved TSMixer model is improved based on the TSMixer model in the following ways: A regulatory target embedding structure is introduced, and a target weight vector is constructed based on the functional zoning label priority, water quality jump frequency, and pressure index of the storage unit. The feature attention distribution of the multi-scale trend vector is adjusted through dynamic weighting. Functional zoning type and water quality level features are introduced into the model input structure, and a unified time index is constructed to improve the environmental adaptability of the time series structure. A hybrid residual path structure is added to the fusion path within the model. After fusing the hybrid vectors of variable dimension and time dimension, residual connection and normalization operations are performed to improve the fusion effect of multimodal sequence features. In the output structure, the weighted trend vector is processed by linear mapping to output the storage and discharge trend prediction results including the storage trend value, discharge trend value, and pressure prediction value of the storage unit. The accuracy and generalization ability of storage and discharge prediction in complex functional zoning environments are demonstrated.

[0029] In this embodiment, S6 specifically refers to: Based on the water storage trend value, discharge trend value and pressure prediction value of the regulation and storage unit in the water storage and discharge trend prediction results, a scheduling input sequence is constructed; a scheduling target vector is set according to the water storage benefit target, reuse target, water quality target and discharge risk target. The scheduling target vector is numerically quantified by a multi-target normalization mapping function to form a target constraint vector with adjustable weight. Based on the water storage trend value and water storage benefit target of the storage unit, water storage regulation is determined, the difference between the deviation value of the regulation trend and the target expectation is calculated, and the interval threshold rule is called to generate a water storage regulation sub-vector; based on the emission trend value and emission risk target, emission regulation is determined, the difference between the emission risk index and the target risk upper limit is calculated, and an emission regulation sub-vector is generated; based on the pressure prediction value and the pressure parameter of the storage unit, pressure regulation is determined, the residual between the pressure over-limit rate and the set safety threshold is calculated, and a pressure regulation sub-vector is output; the water storage regulation sub-vector, emission regulation sub-vector and pressure regulation sub-vector are concatenated to generate a regulation fusion vector; the regulation fusion vector is normalized through a weight reconstruction mechanism to ensure that the contribution of each regulation dimension remains comparable; A path candidate set is constructed based on the regulation fusion vector, rainwater functional zoning labels, spatial location index of storage units, and corresponding green space irrigation paths, treatment paths, and discharge paths in the water quality classification results. A path matching function is constructed, and a multi-factor scoring mechanism is established by combining parameters such as path target satisfaction rate, path physical distance, and path time delay. The path scoring function is fitted by the scheduling target vector to generate a path matching degree value. The path with the highest score is selected as the priority path by sorting the matching degree values, and the corresponding scheduling path is output. The scoring function is obtained by training through historical scheduling samples. Based on the scheduling path, the status of the drainage control components of the storage unit is matched, and the recommended discharge flow rate is calculated by combining the outlet pipe diameter of the storage unit with real-time water level data. Valve opening command, discharge frequency command, and discharge flow rate command are generated. Based on the scheduling path, the water storage structure parameters of the storage unit are matched, and the suggested water storage cycle is calculated by combining the target water storage height and the current water storage capacity. Water storage regulation command is generated. Valve opening command, discharge frequency command, discharge flow rate command, and water storage regulation command are combined in a structured manner to generate an execution command set.

[0030] In this embodiment, S7 specifically refers to: The system collects real-time water storage records from the storage unit, the reuse amount output from the storage unit to the green space irrigation path, the treated effluent from the water treatment path, and the actual irrigation amount from the green space irrigation path to construct data input sequences for rainwater utilization assessment. Based on these data input sequences, the water storage utilization rate is calculated, with the ratio of reused water to stored water as the water storage utilization rate value. The reuse achievement rate is calculated by comparing the reused water output to the green space irrigation path with the target reused water, generating a reuse achievement rate value. The degree of water quality improvement is calculated based on the difference in water quality levels between the influent and effluent from the treatment path, and the change in pollutant indicators before and after treatment is set as the water quality treatment efficiency value. The ratio of actual irrigation amount to the target irrigation amount from the green space irrigation path is calculated to generate an irrigation satisfaction value. The water storage utilization rate value, reuse achievement rate value, water quality treatment efficiency value, and irrigation satisfaction value are combined to form a rainwater utilization assessment parameter set. Based on the water storage benefit target, reuse target, water quality target, and irrigation target set in the scheduling target vector, the values ​​of various indicators in the evaluation parameter set are compared respectively, and the difference between the target value and the evaluation value is calculated. The four types of differences are recorded as water storage evaluation difference, reuse evaluation difference, water quality evaluation difference, and irrigation evaluation difference, respectively, and an evaluation difference vector is constructed. According to the preset feedback rules, error response mapping processing is performed on each type of evaluation difference to generate water storage deviation factor, reuse deviation factor, water quality deviation factor, and irrigation deviation factor. A nonlinear fitting function is set in the feedback rules to segment and adjust the difference interval to ensure that high deviation values ​​generate stronger feedback responses. The nonlinear fitting function is obtained by fitting historical scheduling execution data and error response intensity distribution. The four types of deviation factors are concatenated to generate scheduling feedback factors. The scheduling feedback factor is compared with the corresponding matching degree weight vector in the previous scheduling path, and the matching degree adjustment value for each type of factor is calculated. According to the set update condition rules, it is determined whether the threshold condition for executing the update is met. If at least two deviation factors exceed the preset deviation threshold, the proportional update operation is triggered. The proportional update operation introduces a dynamic learning rate coefficient to dynamically adjust the update step size according to the change of the feedback factor to avoid over-adjustment. The matching degree weight vector is updated to generate the updated scheduling target vector. Based on the updated scheduling target vector, the scheduling path matching degree calculation operation is re-executed. The path information is re-compared with the original path set, the matching degree value of each path is recalculated and sorted, and the path with the highest score is selected to construct the updated scheduling path. According to the control requirements corresponding to the updated scheduling path, the parameter values ​​of valve opening command, discharge frequency command, discharge flow rate command and water storage regulation command are adjusted synchronously, and the control commands are regenerated according to the path structure. The above commands are then concatenated to construct the updated execution command set.

[0031] Example 1: To verify the feasibility of this invention in practice, it was applied to a new district drainage pilot area in a typical city at risk of urban flooding. This area is a mixed-use area with green spaces, public buildings, residences, and roads. It is low-lying and rainwater accumulates quickly, making it a key node for rainwater storage and resource utilization in regional governance.

[0032] In this scenario, the pilot area initially deploys rainfall sensors, surface flow velocity monitoring devices, water quality sampling units, and microclimate acquisition nodes to construct a rainwater status fusion matrix. Combined with urban land use planning, the system automatically divides functional zones and establishes storage units such as infiltration ponds, constructed wetlands, and underground storage tanks based on the hydrological characteristics within each zone. Each storage unit is bound to its real-time capacity status, discharge frequency, and flow velocity parameters, and various water quality indicators, such as pH, conductivity, and pollutant concentration, are collected in real time. The system continuously builds trend prediction models before, during, and after rainfall, and uses an improved TSMixer model to predict the changes in water storage pressure and discharge trends over the next 6 hours, providing a data foundation for the scheduling module.

[0033] During application, the system operates on a three-month cycle, collecting key indicators such as rainwater utilization efficiency, processing capacity, and flood control response before and after deployment. A control group was also established to compare the system with traditional methods. Throughout the cycle, the system experienced five moderate to heavy rain events and two periods of continuous rainfall, which severely tested the stability and intelligence of the system's scheduling and response.

[0034] The table below shows key data indicators for various aspects of the test area under the traditional rainwater management model before the deployment of this system.

[0035] Table 1. Statistical table of key indicators of rainwater utilization under the traditional management model for three months.

[0036] As shown in Table 1, under the traditional rainwater management model, the rainwater reuse rate is consistently below 25%, while the storage unit experiences significant overpressure during heavy rains, averaging as many as 9 times per month. This forces some rainwater to be directly discharged, resulting in a high proportion of excessive discharge (14.1%). Simultaneously, the system's average response time exceeds 17 minutes, easily leading to control lags during sudden rainstorms, and the water quality compliance rate fails to remain stable above 80%.

[0037] After the system was switched to the multi-objective collaborative urban rainwater utilization system proposed in this invention, significant improvements were achieved under the same conditions through mechanisms such as rainwater state fusion, functional zoning identification, and dynamic path scheduling. The following is a comparison of key indicators three months after the system was deployed.

[0038] Table 2. Statistical table of key indicators for rainwater utilization after deployment of the system of the present invention.

[0039] Analysis of Table 2 reveals that this system nearly doubled rainwater utilization, reaching an average of 38.77%. The number of overpressure events in the storage unit decreased from an average of 9 times per month to only 2 times, significantly alleviating the system's pressure buildup. The average response time was controlled at around 6 minutes, reflecting the system's strong real-time trend prediction and rapid regulation capabilities. Simultaneously, the rate of exceeding emission standards decreased significantly, and the water quality compliance rate remained stable at over 90%, fully demonstrating the system's highly efficient decision-making level in multi-path matching and water quality classification.

[0040] This embodiment achieves a high degree of intelligence and goal coordination throughout the entire process, from rainwater harvesting, storage control, path matching to water quality utilization. It solves the problems of fragmented multi-source data, delayed response, fixed paths, and resource waste in traditional models. Under complex meteorological and heterogeneous site conditions, the system, with the trend prediction capabilities and feedback optimization mechanism provided by the improved TSMixer model, not only ensures the safety of storage but also improves the efficiency of rainwater resource reuse and environmental friendliness.

[0041] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A multi-objective coordinated urban rainwater utilization system, characterized in that, include: The environmental data acquisition module is used to collect rainfall, surface runoff, land permeability, groundwater level, water quality parameters and microclimate parameters in urban areas, and to construct a rainwater status fusion matrix. The zoning modeling module is used to perform rainwater functional zoning identification based on the rainwater state fusion matrix and urban plot planning information. It generates functional labels according to land use type, storage and drainage capacity level and water quality target, and constructs a rainwater functional zoning set. It also deploys storage and regulation units, collects capacity data and drainage control data of storage and regulation units, constructs a distributed storage and drainage structure and generates a storage and regulation unit state set. The water quality treatment module is used to classify water quality levels based on the rainwater state fusion matrix, and match green space irrigation paths, treatment paths and discharge paths according to the water quality levels to generate a water quality classification result set; the water quality levels include irrigable rainwater, rainwater requiring treatment and unusable rainwater; The trend prediction module is used to construct multivariate time series based on the storage unit state set, rainwater functional zoning set, and water quality classification result set, and generate storage and discharge trend prediction results through the improved TSMixer model. The regulation and optimization module is used to perform scheduling optimization operations based on the storage and discharge trend prediction results, generate scheduling paths and execution instruction sets; construct a rainwater utilization evaluation parameter set, generate scheduling feedback factors, and if the scheduling feedback factors meet preset update conditions, update the weight parameters in the scheduling optimization operation and generate updated scheduling paths and execution instruction sets.

2. The multi-objective coordinated urban rainwater utilization system according to claim 1, characterized in that, The steps between modules are as follows: S1. Collect rainfall, surface runoff, land permeability, groundwater level, water quality parameters and microclimate parameters in urban areas to construct a rainwater status fusion matrix; S2. Based on the rainwater status fusion matrix and urban plot planning information, perform rainwater functional zoning identification operation, generate functional labels according to land use type, storage and drainage capacity level and water quality target, and construct rainwater functional zoning set; S3. Deploy storage and regulation units according to the rainwater functional zoning set, collect capacity data and drainage control data of the storage and regulation units, construct a distributed storage and drainage structure and generate a storage and regulation unit status set; S4. Based on the rainwater state fusion matrix, classify the water quality levels, and match the green space irrigation path, treatment path and discharge path according to the water quality level to generate a water quality classification result set; the water quality level includes irrigable rainwater, rainwater that needs to be treated and unusable rainwater; S5. Construct a multivariate time series based on the storage unit state set, rainwater functional zoning set, and water quality classification result set, and generate storage and discharge trend prediction results through the improved TSMixer model; S6. Based on the predicted storage and discharge trends, perform scheduling optimization operations to generate scheduling paths and execution instruction sets; S7. Construct a rainwater utilization evaluation parameter set, generate a scheduling feedback factor, and if the scheduling feedback factor meets the preset update conditions, update the weight parameters in the scheduling optimization operation and generate the updated scheduling path and execution instruction set.

3. The multi-objective coordinated urban rainwater utilization system according to claim 2, characterized in that, Specifically, S1 is: Rainfall sensors are installed in urban areas to record rainfall intensity and duration, generating rainfall data; surface flow monitoring devices record runoff velocity and flow rate changes, generating surface runoff; and based on the surface structure, cover type, and soil type of the land parcel, infiltration characteristic indicators are determined to generate the land parcel permeability. The groundwater level is generated by measuring changes in the groundwater level using an underground liquid level sensor. pH, conductivity, turbidity, suspended solids concentration and pollutant concentration are obtained using water sample parameter detection devices to generate water quality parameters; temperature, humidity, wind speed and air pressure are recorded using meteorological measurement devices to generate microclimate parameters. Rainfall, surface runoff, land permeability, groundwater level, water quality parameters, and microclimate parameters are processed with time alignment and structural uniformity to construct a rainwater state fusion matrix. The rows of the rainwater state fusion matrix correspond to time series sampling points, and the columns correspond to rainfall, surface runoff, land permeability, groundwater level, water quality parameters, and microclimate parameters. Each cell corresponds to a standardized data value under a specified time and a specified indicator.

4. The multi-objective coordinated urban rainwater utilization system according to claim 2, characterized in that, Specifically, S2 is: Based on rainfall, surface runoff, plot permeability, groundwater level, water quality parameters and microclimate parameters in the rainwater state fusion matrix, and combined with plot use data and plot boundary data, a set of plot spatial units is established. For each land parcel spatial unit, the land use type is identified, and a land use type label is generated; the land use types include residential land, industrial land, commercial land, public service land, transportation land, and green space. Based on the land permeability and surface runoff, the storage and drainage capacity values ​​are calculated, and the land is classified into high storage and low drainage, medium storage and medium drainage, and low storage and high drainage according to the set grade standards, and storage and drainage capacity grade labels are generated. Based on the pollutant index values ​​in the water quality parameters and the regional water quality requirements, the water quality target labels are generated by dividing the target into primary, secondary and tertiary levels according to the set intervals. By associating land use type labels, water storage and drainage capacity level labels, and water quality target labels with corresponding land parcel spatial units, a functional label structure is constructed. Land parcel spatial units with the same functional label structure and spatially adjacent to each other are merged to form rainwater functional zones, generating a rainwater functional zone set.

5. The multi-objective coordinated urban rainwater utilization system according to claim 2, characterized in that, Specifically, S3 is: A storage and regulation unit is set up in each rainwater functional zone; the storage and regulation unit includes one or more structural combinations of rainwater collection tanks, infiltration ponds, constructed wetlands, modular water storage structures and underground storage and regulation devices; Record the structural volume parameters, actual water storage upper limit and historical water storage records of each regulation and storage unit to generate capacity data; Configure the drainage control components of the storage unit, record the status of the outlet valve, the discharge frequency and the discharge flow rate, and generate discharge control data; construct a time series structure of capacity data and discharge control data for each storage unit, and generate the storage unit state vector; Spatial mapping is performed on the state vectors of all storage units to construct a distributed storage and drainage structure corresponding to the rainwater functional zoning set; the state vectors of all storage units are integrated to generate a storage unit state set.

6. The multi-objective coordinated urban rainwater utilization system according to claim 2, characterized in that, The water quality classification based on the rainwater state fusion matrix is ​​as follows: Water quality parameter data are extracted from the rainwater state fusion matrix; the water quality parameter data includes pH value, conductivity, turbidity, suspended solids concentration and pollutant concentration; Construct a water quality grade determination range and set multi-dimensional boundary parameters for irrigable rainwater grade, rainwater requiring treatment grade, and unusable rainwater grade; Based on the extracted water quality index data and water quality grade determination intervals, each time series sample in the state fusion matrix is ​​labeled with a grade according to the rule-based determination method to generate a water quality grade label set; the water quality grade label set is combined with the corresponding time index to construct the water quality grade time series result.

7. The multi-objective synergic urban rainwater utilization system according to claim 2, characterized in that, The process of matching green space irrigation paths, treatment paths, and discharge paths according to water quality levels to generate a water quality classification result set is as follows: Based on the records marked as irrigable rainwater in the water quality grade time series results, and according to the principles of functional zoning concentration, irrigation radius and shortest distance to storage unit, the corresponding green space irrigation paths are selected to generate an irrigation path subset; Based on the records marked as rainwater requiring treatment, the matching priority is calculated according to the capacity utilization rate of the storage unit, the load status of the water treatment node, and the pipeline network bearing capacity coefficient. Treatment paths are selected and a subset of treatment paths is generated. Based on records marked as unusable rainwater, emission paths are selected according to emission channel level, emission threshold and regional emission risk level, and a subset of emission paths is generated; The irrigation path subset, treatment path subset, and discharge path subset are time-indexed and structurally concatenated to construct a water quality classification result set.

8. The multi-objective synergic urban rainwater utilization system according to claim 2, characterized in that, The improved TSMixer model is specifically as follows: An input structure is constructed by concatenating the state set of the storage unit, the rainwater functional zoning set, and the water quality classification result set according to a unified time index to form an input sequence tensor. In the input sequence tensor, rows represent time sampling points, and columns represent standardized values ​​of storage unit water storage capacity, storage unit discharge frequency, storage unit discharge rate, functional zoning type, water quality level, and microclimate index. A bidirectional hybrid structure is constructed, and linear hybridization processing is performed on the input sequence tensor along both the variable dimension and the time dimension to generate a variable hybrid vector and a time hybrid vector. The variable hybrid vector and the time hybrid vector are then weighted and fused to form a hybrid residual path. The hybrid residual path is then normalized to obtain a fused vector. A hierarchical trend structure is constructed, and multi-scale convolution processing is performed based on the fusion vector to extract short-cycle trend vectors, medium-cycle trend vectors and long-cycle trend vectors respectively; the three types of trend vectors are concatenated according to their dimensions to form a trend combination vector. An embedded structure for regulation targets is constructed, and a target weight vector is generated based on the priority of functional zoning labels, the frequency of water quality jumps, and the pressure index of the storage unit. The trend combination vector is dynamically weighted according to the target weight vector to obtain a weighted trend vector. Construct a linear prediction structure, input the weighted trend vector into a linear mapping path, and output the storage and discharge trend prediction results; The predicted storage and discharge trends include the water storage trend value, discharge trend value, and pressure prediction value of the regulating and storage unit for the future time period.

9. The multi-objective synergic urban rainwater utilization system according to claim 2, characterized in that, Specifically, S6 is: Based on the water storage trend value, discharge trend value and pressure prediction value of the regulation and storage unit in the water storage and discharge trend prediction results, a scheduling input sequence is constructed; and a scheduling target vector is set according to the water storage benefit target, reuse target, water quality target and discharge risk target. Based on the water storage trend value and water storage benefit target of the storage unit, water storage regulation judgment is performed to generate water storage regulation sub-vector; based on the emission trend value and emission risk target, emission regulation judgment is performed to generate emission regulation sub-vector; based on the pressure prediction value and pressure parameter of the storage unit, pressure regulation judgment is performed to generate pressure regulation sub-vector; the water storage regulation sub-vector, emission regulation sub-vector and pressure regulation sub-vector are concatenated to generate regulation fusion vector; Based on the regulation fusion vector and rainwater functional zoning labels, storage unit spatial location index, and corresponding green space irrigation paths, treatment paths, and discharge paths in the water quality classification results, a path candidate set is constructed; the matching degree value of each path is calculated according to the scheduling target vector; priority paths are selected based on the matching degree values ​​to generate scheduling paths; Based on the scheduling path, the status of the drainage control components of the storage unit is matched, and valve opening instructions, discharge frequency instructions, and discharge flow rate instructions are generated. Based on the scheduling path, the water storage structure parameters of the storage unit are matched, and water storage regulation instructions are generated. The valve opening instructions, discharge frequency instructions, discharge flow rate instructions, and water storage regulation instructions are combined to form an execution instruction set.

10. The multi-objective synergic urban rainwater utilization system according to claim 2, characterized in that, Specifically, S7 is: The system collects real-time water storage records from the storage unit, the reuse amount output from the storage unit to the green space irrigation path, the treated water output from the water treatment path, and the actual irrigation amount from the green space irrigation path to construct a rainwater utilization assessment parameter set. The rainwater utilization assessment parameter set includes four types of assessment indicators: water storage utilization rate, reuse achievement rate, water quality treatment efficiency, and irrigation satisfaction. Based on the difference between the evaluation indicators and the scheduling targets, evaluation difference vectors are constructed respectively; water storage deviation factor, reuse deviation factor, water quality deviation factor and irrigation deviation factor are calculated according to the set feedback rules, and then concatenated into scheduling feedback factor. The scheduling feedback factor is compared with the matching degree weight vector in the previous scheduling path to determine whether the preset update conditions are met. If they are met, the weight vector is proportionally updated to construct the updated scheduling target vector. Based on the updated scheduling target vector, the scheduling path selection operation is re-executed to construct the updated scheduling path; the valve opening command, discharge frequency command, discharge flow rate command and water storage control command are adjusted synchronously to generate the updated execution command set.

Images