Multi-objective water resources optimal scheduling method and system based on dream optimization algorithm

Abstract

The application belongs to the technical field of water resources scheduling management, and specifically provides a multi-objective water resources optimal scheduling method and system based on a dream optimization algorithm, which comprises the following steps: collecting runoff, rainfall and regenerated water prediction messages, checking and analyzing the analyzed prediction values and corresponding expected values and standard deviations; calculating the minimum and expected available water quantity; taking the minimum available water quantity as a constraint, constructing a multi-objective optimization model, and generating a fixed allocation scheme; according to the real-time hydrological data in the scheduling period, the expected available water quantity is corrected in a rolling manner, and the allocable surplus is calculated; when the allocable surplus is greater than zero, a gain allocation model is constructed based on the actual water consumption state of the terminal, and the dream optimization algorithm is used for solving; the actual water output quantity of the terminal is compared with the instruction value, the deviation rate is calculated, and the rolling prediction model parameters and the confidence are updated online. The two-stage scheduling strategy adopted by the method realizes dynamic adjustment of the scheduling scheme according to the actual hydrological working condition, and effectively solves the problem of large deviation between the traditional scheduling scheme and the actual supply and demand.

Description

Technical Field

[0001] This invention belongs to the field of water resource scheduling and management technology, specifically, it relates to a multi-objective water resource optimization scheduling method and system based on the dream optimization algorithm. Background Technology

[0002] Optimal water resource allocation is a core technology for achieving rational allocation and efficient utilization of regional water resources, playing a vital role in ensuring regional water supply security and supporting sustainable economic and social development. However, the current water resource allocation technology system still has many areas for improvement, specifically as follows: First, the existing scheduling model is mainly static, which has limited adaptability. Most current scheduling methods adopt a "single prediction, single optimization allocation" working mode, that is, a fixed scheduling plan is formulated and implemented based on hydrological forecast data at the beginning of the scheduling cycle, and generally no dynamic adjustment is made according to the actual water inflow conditions during the scheduling cycle.

[0003] Second, the processing methods for multi-source forecast data are relatively crude and lack refinement. When processing multi-source forecast data such as runoff, rainfall, and reclaimed water production, existing technologies generally directly use the expected or average forecast values ​​for optimization calculations. Some scheduling methods that take into account forecast uncertainty also process the data simply by taking the minimum value or adding a safety factor, lacking refined water volume estimation methods based on probability and statistics.

[0004] Third, most dispatching systems are open-loop architectures, lacking iterative optimization capabilities. Existing mainstream water resource dispatching systems adopt an open-loop control mode. After the dispatching center issues dispatching instructions, it does not systematically collect key feedback data such as the execution effect of water users, the deviation between actual water output and instructions, and the actual accuracy of prediction models. It lacks a closed-loop dispatching mechanism of "execution-feedback-correction".

[0005] The existing patent technology, with publication number CN121707173A and titled "A Multi-Objective Water Resource Optimization Scheduling Method Based on Dream Optimization Algorithm," discloses four macro-level implementation steps: regional basic data collection, scheduling target setting, optimization algorithm calculation, and multi-objective decision-making. This patent technology focuses on the overall scheduling process framework but does not disclose detailed implementation details such as system hardware architecture, data communication protocols, message formats, prediction uncertainty quantification methods, water allocation processes, rolling prediction models, and real-time feedback correction. Based solely on this disclosure, it is difficult for those skilled in the art to directly build and implement a complete water resource optimization scheduling system. Summary of the Invention

[0006] The technical problem to be solved by this invention is to provide a multi-objective water resource optimization scheduling method and system based on the dream optimization algorithm. The two-stage scheduling strategy adopted enables the scheduling scheme to be dynamically adjusted according to the actual hydrological conditions, which effectively solves the problem of large deviation between traditional scheduling schemes and actual supply and demand.

[0007] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a multi-objective water resource optimization scheduling method based on the dream optimization algorithm, comprising the following steps: S1. Collect runoff forecast data messages, rainfall forecast data messages, and reclaimed water production forecast data messages. Verify and parse each data message to obtain the parsed forecast values ​​and corresponding distribution parameters. The distribution parameters include the expected values ​​of the forecast values ​​within the preset scheduling period. and the corresponding standard deviation ; S2. Based on the predicted values ​​obtained from step S1 and the corresponding distribution parameters, calculate the minimum available water volume prediction value and the expected available water volume prediction value. S3. Using the lowest predicted available water volume as the upper limit of available water volume constraint, construct and solve a multi-objective optimization allocation model to generate a first-stage fixed allocation scheme. The fixed allocation scheme remains constant during execution and includes the allocated water volume values ​​of each water terminal controller. S4. During the execution of the current scheduling cycle, based on the real-time monitoring data reports from the hydrological monitoring terminal, the short-term rolling prediction program is used to perform rolling correction prediction of the available water volume for the remaining period in the current scheduling cycle, and outputs the corrected expected available water volume prediction value. Based on the corrected expected available water volume prediction value and the minimum available water volume prediction value, the allocable surplus water volume is calculated. S5. When the allocable surplus water is greater than zero, the second stage of dynamic gain allocation is executed. Based on the actual water usage status data of each water terminal, the allocable surplus water is used as the allocable increment constraint to construct a gain allocation model. The dream optimization algorithm is used to solve the gain allocation model. S6. Collect terminal operation status messages reported by each water terminal controller at predetermined time intervals, extract actual water output data, compare the actual water output with the allocation instruction value for the corresponding time period, calculate the execution deviation rate, use the execution deviation rate as feedback input, update the model parameters of the short-term rolling prediction program online, and compare the actual water inflow data with the predicted value to update the distribution parameters of each water source. Steps S4 to S6 are executed cyclically at predetermined time intervals within the current scheduling cycle until the current scheduling cycle ends.

[0008] In the preferred embodiment, the calculation process for the minimum predicted available water volume and the expected predicted available water volume in step S2 is as follows: Expected values ​​of predicted values ​​for each water source As the mean, As a distribution parameter, it is assumed that the predicted available water volume of each water source follows a normal distribution. Set probability threshold The probability threshold is determined based on the inverse function of the cumulative distribution function of the standard normal distribution. quantile coefficients ; For each water source, calculate the probability threshold for that water source using the following formula. The lower confidence limit below The lower confidence limit indicates whether, under the predicted distribution of the water source, there is a certain level of confidence. The probability that the actual usable water volume is not less than this value: ; Lower confidence limit for all water sources By summing the results, we obtain the predicted minimum available water volume within the scheduling cycle. : ; The expected values ​​of each water source By summing the results, we obtain the expected available water volume forecast for the scheduling period. : .

[0009] In the preferred embodiment, in step S3, a non-dominated sorting genetic algorithm is used to solve the multi-objective optimization allocation model, generating a set of Pareto optimal solutions. A solution is selected from the Pareto optimal solution set as the first-stage fixed allocation scheme according to a preset decision rule. The decision rule is as follows: based on the hydrological warning status and regional water supply priority configuration within the current scheduling cycle, the optimal compromise solution under the corresponding objective function weight allocation is selected from the Pareto optimal solution set. When the region is in a drought warning state, priority is given to ensuring domestic water supply, and the solution with the highest domestic water supply guarantee rate is selected. When the region is in a water-rich state, priority is given to ensuring economic benefits, and the solution with the greatest comprehensive benefits is selected. When the region is in a normal state, the entropy-weighted TOPSIS multi-attribute decision method is used, with the entropy weight of each objective function as the weight, to calculate the relative closeness of each Pareto solution, and the solution with the highest closeness is selected as the final scheme.

[0010] In the preferred embodiment, in step S4, the short-term rolling forecast program employs an autoregressive moving average model, using the latest measured data from each water source as the observed values ​​input into the autoregressive moving average model, and outputting a corrected predicted value of the expected available water volume. .

[0011] In a preferred embodiment, in step S4, the allocable surplus water is the corrected predicted value of the expected available water volume. Compared with the minimum available water forecast The difference is calculated using the following formula: .

[0012] In the preferred embodiment, in step S5, the actual water usage status data of each water terminal includes the current actual water supply of each terminal, the current water level of the reservoir, the water demand completion rate, and the output value per unit of water volume.

[0013] In the preferred embodiment, in step S5, the objective function of the gain allocation model is to maximize the comprehensive benefits of surplus water allocation, the decision variable is the additional water allocation value of each water terminal controller, and the constraints include the total additional water allocation not exceeding the allocable surplus water, the upper limit constraint of the water transmission capacity of each terminal, and the upper limit constraint of the reservoir capacity of the water storage facility.

[0014] In the preferred embodiment, in step S6, the model parameters of the short-term rolling prediction program are updated online using the recursive least squares method, and the distribution parameters of each water source are updated using the weighted recursive average method.

[0015] The present invention also provides a water resource optimization scheduling system based on two-stage allocation, for implementing the above-described method, comprising: The dispatch center server, deployed in the water resources dispatch management center, is equipped with a processor, memory module, communication interface and persistent storage device. It is used to receive and cache multi-source prediction data, calculate the minimum available water volume prediction value and the expected available water volume prediction value, generate and issue the first-stage fixed allocation scheme, calculate the allocable surplus water volume, generate and issue the second-stage dynamic gain allocation scheme, and receive and roll over feedback data from the terminal. Multiple hydrological monitoring terminals are deployed at river monitoring sections and rainfall monitoring stations. They are equipped with water level sensors, flow sensors, communication modules and microcontrollers to collect raw monitoring data of runoff and rainfall and report them to the dispatch center server through the communication network. A weather forecast server, deployed at the weather forecast center, is connected to the dispatch center server via a communication link and is used to provide rainfall forecast data messages; The monitoring terminal for the reclaimed water treatment plant is deployed at the outlet of the reclaimed water treatment plant and is equipped with a flow sensor, a communication module and a microcontroller. It is used to collect data on the amount of reclaimed water produced and report it to the dispatch center server through the communication network. Multiple water terminal controllers are deployed in each water unit and are equipped with programmable logic controllers or embedded microcontrollers, communication modules, flow meter interfaces and valve / pump group drive interfaces. They are used to receive and execute fixed allocation instruction messages and incremental allocation instruction messages issued by the dispatch center server, and report terminal operation status messages. The communication network, including wired communication links and wireless communication links, is used for data packet transmission between the aforementioned hardware nodes.

[0016] In a preferred embodiment, the dispatch center server dynamically adjusts the time interval for each hydrological monitoring terminal to report real-time monitoring data through remote configuration messages: when the rate of change of the measured data continuously exceeds a preset high threshold, the reporting interval is shortened; when the rate of change of the measured data continuously falls below a preset low threshold, the reporting interval is extended.

[0017] The present invention provides a multi-objective water resource optimization scheduling method and system based on a dream optimization algorithm, which has the following beneficial effects: 1. This invention abandons the traditional static scheduling mode of single prediction and single allocation, and constructs a complete closed-loop rolling optimization scheduling process. This invention adopts a two-stage scheduling strategy: first, it completes basic water allocation based on conservative prediction results to ensure basic water supply security; then, it continuously conducts real-time hydrological monitoring and rolling water volume correction predictions within the scheduling cycle. For actual surplus inflow, it accurately allocates water to high-efficiency water-using units or reserve reservoirs through a dynamic gain allocation mechanism; simultaneously, it can initiate early warning intervention when inflow is low. The entire process is cyclically iterated, enabling the scheduling scheme to be dynamically adjusted according to actual hydrological conditions, effectively solving the problem of large deviations between traditional scheduling schemes and actual supply and demand, and balancing efficient water resource utilization with water supply risk prevention and control.

[0018] 2. To address the issues of inconsistent prediction accuracy and difficulty in managing uncertainty across multiple water sources, this invention proposes a minimum available water quantity quantification method based on the accumulation of confidence lower limits. The accumulated confidence lower limits of multiple water sources serve as the upper limit constraint for initial fixed water quantity allocation, effectively mitigating the adverse effects of prediction errors, ensuring stable basic water supply under extreme prediction deviations, and improving system robustness. Furthermore, this invention dynamically iteratively updates the distribution parameters of each water source based on the cumulative error between actual inflow and prediction data, enabling differentiated weight allocation for high- and low-precision water sources. This mechanism overcomes the shortcomings of existing technologies that cannot distinguish between confidence differences in multi-source prediction data, achieving an optimal balance between water supply security and water resource utilization efficiency.

[0019] 3. By collecting real-time operational data such as water output, flow rate, water level, and equipment operating status from each water terminal, and comparing the deviation rate between the measured water volume and the commanded water volume, a two-layer parameter collaborative online update is achieved based on the deviation data: First, the parameters of rolling prediction models such as ARMA and Kalman filtering are dynamically optimized through recursive least squares method to continuously improve the accuracy of hydrological prediction; second, the distribution parameters of each water source are updated synchronously to ensure that the confidence assessment and the actual prediction accuracy match in real time. This invention forms a positive iterative closed loop, allowing the system's prediction accuracy and scheduling control capabilities to be continuously upgraded with operational data, significantly improving the adaptive performance and long-term operational stability of the scheduling system. Attached Figure Description

[0020] The accompanying drawings, which are provided to further illustrate the invention and constitute a part of this invention, do not constitute an undue limitation thereof. In the drawings: Figure 1 This is a schematic diagram of the overall process of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention.

[0022] Example 1: This embodiment provides a water resource optimization scheduling system based on two-stage allocation, including: The dispatch center server executes a multi-objective water resource optimization scheduling method based on the dream optimization algorithm. The dispatch center server communicates with multiple hydrological monitoring terminals, meteorological forecast servers, reclaimed water treatment plant monitoring terminals, and multiple water terminal controllers through a communication network.

[0023] The dispatch center server is deployed in the water resources dispatch management center and is equipped with a processor, memory module, communication interface and persistent storage device. It is used to receive and cache multi-source prediction data, calculate the minimum available water volume prediction value and the expected available water volume prediction value, generate and issue the first-stage fixed allocation scheme, calculate the allocable surplus water volume, generate and issue the second-stage dynamic gain allocation scheme, and receive and roll over feedback data from the terminal.

[0024] The dispatch center server dynamically adjusts the time interval for each hydrological monitoring terminal to report real-time monitoring data through remote configuration messages: when the rate of change of the measured data continuously exceeds the preset high threshold, the reporting interval is shortened; when the rate of change of the measured data continuously falls below the preset low threshold, the reporting interval is extended.

[0025] Multiple hydrological monitoring terminals are deployed at river monitoring sections and rainfall monitoring stations. They are equipped with water level sensors, flow sensors, communication modules and microcontrollers to collect raw monitoring data of runoff and rainfall and report them to the dispatch center server via the communication network.

[0026] The weather forecast server, deployed at the weather forecast center, is connected to the dispatch center server via a communication link and is used to provide rainfall forecast data messages.

[0027] The monitoring terminal for the reclaimed water treatment plant is deployed at the outlet of the reclaimed water treatment plant and is equipped with a flow sensor, a communication module and a microcontroller. It is used to collect data on the amount of reclaimed water produced and report it to the dispatch center server via the communication network.

[0028] Multiple water terminal controllers are deployed in each water unit and are equipped with programmable logic controllers or embedded microcontrollers, communication modules, flow meter interfaces and valve / pump group drive interfaces. They are used to receive and execute fixed allocation instruction messages and incremental allocation instruction messages issued by the dispatch center server, and report terminal operation status messages.

[0029] The communication network, including wired communication links and wireless communication links, is used for data packet transmission between the aforementioned hardware nodes.

[0030] Example 2: A multi-objective water resource optimization scheduling method based on dream optimization algorithm, such as Figure 1 As shown, it includes the following steps: S1. Receive and cache multi-source prediction data Collect runoff forecast data, rainfall forecast data, and reclaimed water production forecast data. Verify and parse each data message to obtain the parsed forecast values ​​and corresponding distribution parameters. The distribution parameters include the expected values ​​of the forecast values ​​within a preset scheduling period. and the corresponding standard deviation ; Specifically, the dispatch center server receives runoff prediction data messages reported by the hydrological monitoring terminal, rainfall prediction data messages sent by the meteorological forecast server, and reclaimed water production prediction data messages reported by the reclaimed water treatment plant monitoring terminal. It verifies and parses each data message and stores the parsed prediction values ​​and corresponding distribution parameters in the server's memory cache.

[0031] The data packets are transmitted using the TCP / IP protocol stack. The packet format includes a header, a body, and a trailer. The header includes a data source identifier field (used to distinguish between hydrological monitoring stations, reservoirs, or meteorological satellites) and a forecast period field (corresponding to runoff / rainfall forecasts for the next few hours or days). The body includes a predicted value field (specifically referring to the predicted inflow, reservoir capacity changes, or available water volume for that period) and a distribution parameter field (reflecting the reliability of the prediction and used for risk assessment in subsequent allocation decisions). The trailer includes a checksum field. The dispatch center server uses a CRC checksum algorithm. The received data packets are verified and validated. Messages that pass the verification are parsed and stored in the memory cache. An index is created according to the data source type and forecast period. Messages that fail the verification are discarded and a retransmission request message is sent back to the corresponding data source node. Subsequently, the dispatch center uses the forecast data of each source point in the cache, combined with the current water demand plan and priority weight of each water terminal (including hydropower, farmland irrigation, urban domestic water use, etc.), to call a multi-objective optimization algorithm to dynamically allocate water resources, generate allocation instructions and send them to the execution mechanism of each terminal to achieve supply and demand balance and maximize comprehensive benefits.

[0032] S2. Calculate the minimum and expected available water volume forecasts. Based on the analyzed predicted values ​​and corresponding distribution parameters obtained in step S1, calculate the minimum available water volume prediction value and the expected available water volume prediction value.

[0033] Specifically, the dispatch center server reads the expected values ​​of the predicted values ​​of each water source within a preset dispatch period from the memory cache. and the corresponding standard deviation ; and based on the expected value and the corresponding standard deviation Calculate the minimum and expected water availability forecasts; The calculation process for the minimum and expected available water quantities is as follows: Expected values ​​of predicted values ​​for each water source As the mean, As a distribution parameter, it is assumed that the predicted available water volume of each water source follows a normal distribution. Set probability threshold The probability threshold is determined based on the inverse function of the cumulative distribution function of the standard normal distribution. quantile coefficients ; For each water source, calculate the probability threshold for that water source using the following formula. The lower confidence limit below The lower confidence limit indicates whether, under the predicted distribution of the water source, there is a certain level of confidence. The probability that the actual usable water volume is not less than this value: ; Lower confidence limit for all water sources By summing the results, we obtain the predicted minimum available water volume within the scheduling cycle. : ; The expected values ​​of each water source By summing the results, we obtain the expected available water volume forecast for the scheduling period. : .

[0034] S3, Execute the first phase of fixed allocation Using the lowest predicted available water volume as the upper limit of available water volume constraint, a multi-objective optimization allocation model is constructed and solved to generate a first-stage fixed allocation scheme. The fixed allocation scheme remains constant during execution and includes the allocated water volume values ​​of each water terminal controller.

[0035] The multi-objective optimization allocation model uses the water allocation amount of each water user terminal in each scheduling period as the decision variable, and takes maximizing the water supply guarantee rate of each water user unit and maximizing the allocation fairness among each water user terminal as the objective function. The constraints include the minimum water demand of each water user terminal, the upper limit of the water supply network transmission capacity, the total amount constraint that the water supply of each water source does not exceed the minimum available water volume prediction value, and the priority constraint of water demand of each water user terminal.

[0036] Specifically, the dispatch center server uses the predicted minimum available water volume value. As the upper limit of available water volume constraint, a multi-objective optimization allocation model is constructed and solved to generate a first-stage fixed allocation scheme. The fixed allocation scheme includes the allocated water volume values ​​of each water terminal controller. The dispatch center server encapsulates the fixed allocation scheme into a fixed allocation instruction message and sends it to each water terminal controller through the communication network. After receiving and parsing the fixed allocation instruction message, each water terminal controller drives the corresponding actuator to adjust the water supply to the instruction value. The fixed allocation scheme remains constant during execution.

[0037] In this embodiment, a non-dominated sorting genetic algorithm is used to solve the multi-objective optimization allocation model, generating a set of Pareto optimal solutions. According to the preset decision rules, a scheme is selected from the Pareto optimal solution set as the first-stage fixed allocation scheme.

[0038] The decision-making rules are as follows: Based on the hydrological early warning status and regional water supply priority configuration within the current scheduling cycle, the optimal compromise solution under the corresponding objective function weight allocation is selected from the Pareto optimal solution set. When the region is in a drought early warning state, priority is given to ensuring domestic water supply, and the solution with the highest domestic water supply guarantee rate is selected. When the region is in a water-rich state, priority is given to ensuring economic benefits, and the solution with the greatest comprehensive benefits is selected. When the region is in a normal state, the entropy weight TOPSIS multi-attribute decision-making method is adopted, using the entropy weight of each objective function as the weight, calculating the relative closeness of each Pareto solution, and selecting the solution with the highest closeness as the final solution.

[0039] S4. Receive real-time monitoring data and calculate the allocable surplus water volume. During the execution of the current scheduling cycle, based on the real-time monitoring data reports from the hydrological monitoring terminal, the short-term rolling prediction program is used to perform rolling correction predictions on the available water volume for the remaining period within the current scheduling cycle, outputting the corrected expected available water volume prediction value, and calculating the allocable surplus water volume based on the corrected expected available water volume prediction value and the minimum available water volume prediction value.

[0040] The short-term rolling forecast program employs an autoregressive moving average model, using the latest measured data from each water source as the input to the model, and outputs a corrected predicted value for the expected available water volume. .

[0041] The allocable surplus water is the revised expected forecast of available water. Compared with the minimum available water forecast The difference is calculated using the following formula: .

[0042] S5. When the allocable surplus water is greater than zero, execute the second stage of dynamic gain allocation. When the distributable surplus water volume When the value is greater than zero, the dispatch center server starts the second-stage gain allocation process. Based on the actual water usage status data of each water terminal, the available surplus water volume is used as the available incremental constraint to construct a gain allocation model, and the dream optimization algorithm is used to solve the gain allocation model.

[0043] Specifically, the dispatch center server reads the actual water usage status data of each water terminal from the memory cache. The actual water usage status data includes the current actual water supply of each terminal, the current water level of the reservoir, the water demand completion rate, and the output value per unit of water volume.

[0044] The dispatch center server allocates surplus water volume As an allocable incremental constraint, a gain allocation model is constructed. The objective function of the gain allocation model is to maximize the comprehensive benefit of surplus water allocation. The decision variable is the additional water allocation value of each water terminal controller. The constraint condition includes that the total additional water allocation does not exceed [a certain value]. The upper limit constraints on the water delivery capacity of each terminal and the upper limit constraints on the reservoir capacity of water storage facilities are used, and the dream optimization algorithm solution program stored in the server memory is called to solve the gain allocation model.

[0045] S6. Receive terminal execution feedback data and perform rolling corrections. The system collects terminal operation status messages reported by each water terminal controller at predetermined time intervals, extracts actual water output data, compares the actual water output with the allocation command values ​​for the corresponding time period, calculates the execution deviation rate, uses the execution deviation rate as feedback input to update the model parameters of the short-term rolling prediction program online, and compares the actual water inflow data with the predicted values ​​to update the distribution parameters of each water source. .

[0046] Specifically, the recursive least squares method is used to update the model parameters of the short-term rolling forecast program online, and the weighted recursive average method is used to update the distribution parameters of each water source.

[0047] Steps S4 to S6 are executed cyclically at predetermined time intervals within the current scheduling cycle until the current scheduling cycle ends.

[0048] Example 3: This embodiment uses a watershed water resources scheduling system as an example to describe in detail the specific implementation of the water resources optimization scheduling method of the present invention. The watershed water resources scheduling system covers an area of ​​approximately 8,000 square kilometers, involving domestic water supply, industrial water supply, and ecological water replenishment in three administrative regions. The system hardware architecture includes: one scheduling center server deployed in the computer room of the watershed water resources scheduling and management center; 28 hydrological monitoring terminals deployed at monitoring sections of major rivers, rainfall monitoring stations, and reservoir inflow monitoring points within the watershed; one meteorological forecast server deployed at the regional meteorological forecast center, connected to the scheduling center server via a dedicated communication link; five reclaimed water treatment plant monitoring terminals deployed at the outlets of reclaimed water treatment plants within the watershed; and a total of 235 water user terminal controllers, including 85 domestic water pump station controllers, 110 industrial water supply network regulating valve controllers, and 40 ecological water replenishment gate controllers. All the above hardware nodes are connected via a hybrid network of fiber optic wired communication and 4G wireless communication.

[0049] A multi-objective water resource optimization scheduling method based on a dream optimization algorithm includes the following steps: Step 1: The scheduling center server receives and caches multi-source prediction data.

[0050] The dispatch center server initiates the data receiving process at a preset time point (24 hours before the start of the current dispatch cycle in this embodiment). The hydrological monitoring terminal has the following built-in components: a data acquisition module, including a radar level gauge and an ultrasonic flow meter, for collecting cross-sectional water level and flow velocity; a data preprocessing module, using an ARM Cortex-M7 series microprocessor and running an embedded real-time operating system, for performing outlier removal and time-series smoothing on the collected raw water level and flow velocity data; and a communication module, including a wired Ethernet interface and a 4G / 5G wireless communication module, for encapsulating and reporting the preprocessed data according to a preset message format. The runoff prediction data is calculated by the hydrological monitoring terminal's data preprocessing module based on the collected current water level and flow velocity data, combined with the locally stored cross-sectional water level-flow relationship curve, through table lookup and linear interpolation.

[0051] The dispatch center server receives runoff prediction data messages reported by 28 hydrological monitoring terminals via the communication network, rainfall forecast data messages sent by the meteorological forecast server, and reclaimed water production prediction data messages reported by the monitoring terminals of five reclaimed water treatment plants via the communication interface.

[0052] Each data message is transmitted using the TCP / IP protocol stack. Specifically, the message header is a fixed 32 bytes long, with the data source identifier field being a 4-byte unsigned integer used to uniquely identify the data source node number (to distinguish between hydrological monitoring stations, reservoirs, or meteorological satellites); the start and end timestamps of the forecast period are each an 8-byte Unix timestamp (unit: seconds) (corresponding to runoff / rainfall forecasts for the next few hours or dates); and the message length field is a 2-byte unsigned integer (unit: bytes). The message body is a variable-length field, with the predicted value field being a 4-byte single-precision floating-point number (IEEE 754 standard) (specifically referring to the predicted inflow, reservoir capacity change, or available water volume for the specified period), the distribution parameter being a 2-byte unsigned integer (representing a percentage, ranging from 0 to 100, with actual confidence level = field value / 100) (reflecting prediction reliability and used for risk assessment in subsequent allocation decisions), and the upper and lower limits of the confidence interval being 4-byte single-precision floating-point numbers. The message tail is a 4-byte CRC-32 checksum (unsigned integer). The message format includes three fields: message header, message body, and message footer. The message header contains a data source identifier field, a forecast period start timestamp field, a forecast period end timestamp field, and a message length field. The message body contains a predicted value field, the corresponding distribution parameters, and upper and lower confidence interval fields. The message footer contains a checksum field, which is calculated using the CRC-32 cyclic redundancy check algorithm.

[0053] After receiving each data packet, the dispatch center server first extracts the checksum field at the end of the packet. It then performs a checksum calculation on the packet header and body using the CRC-32 checksum algorithm, comparing the result with the checksum field value at the end. If they match, the packet verification is successful. The server parses the forecast period identifier and data source identifier from the packet header, as well as the predicted values ​​and distribution parameters from the packet body, and stores them in the server's memory cache. A hash index table is then created according to the data source type and forecast period for subsequent fast retrieval. If they do not match, the packet verification fails. The server discards the packet, records the packet loss event log, and sends a retransmission request packet containing the packet sequence number to the corresponding data source node via the communication network. Subsequently, based on the forecast data from each source point in the cache, combined with the current water demand plans and priority weights of each water user terminal (including hydropower, farmland irrigation, and urban residential water use), the dispatch center uses a multi-objective optimization algorithm to dynamically allocate water resources, generating allocation instructions and issuing them to the execution mechanisms of each terminal to achieve supply and demand balance and maximize overall benefits.

[0054] Step 2: The dispatch center server calculates the minimum available water volume forecast and the expected available water volume forecast.

[0055] The dispatch center server reads the predicted values ​​and corresponding distribution parameters of each water source within a preset dispatch period from the memory cache. For the first... Predicted values ​​for each water source The server reads the variance estimate corresponding to the predicted value from the memory cache. or standard deviation The variance estimate is output by each prediction model when generating the predicted value, and represents the degree of dispersion of the predicted value around the expected value.

[0056] The server assumes that the predicted available water volume of each water source follows a predictive value. As the mean, with The variance follows a normal distribution, i.e. Based on the above normal distribution assumption, the server calculates the lower confidence limit of each water source under a preset probability threshold according to the quantile calculation method of the standard normal distribution.

[0057] Specifically, for the first A water source, with a probability threshold parameter set by the server. ,in The value range is greater than 0 and less than 1. In this embodiment... The value is set to 95%. The server determines the probability corresponding to the inverse function of the cumulative distribution function of the standard normal distribution. quantile coefficients .when When taking 95%, The value is set to 1.645 (corresponding to the one-sided confidence lower limit). The server follows the formula... Calculate the probability of this water source. The lower confidence limit at a certain confidence level, which indicates whether the predicted distribution of the water source is... The probability that the actual usable water volume is not lower than this value.

[0058] After the server iterates through all water sources, it sets the confidence lower limit for each water source. By summing the results, we obtain the predicted minimum available water volume within the scheduling cycle. The minimum available water volume prediction value represents the water volume within a preset probability threshold. The actual total inflow of water from each water source shall not be less than The confidence level is .

[0059] The server simultaneously reads the expected values ​​of the predicted values ​​for each water source from the memory cache. This expected value is output by each prediction model when generating the predicted values, representing the mathematical expectation of the predicted value under its probability distribution. The server will output the expected values ​​of each water source. By summing the results, we obtain the expected available water volume forecast for the scheduling period. .

[0060] Taking the data from a certain river monitoring section in this embodiment as an example: the expected value of the predicted runoff for the next seven days reported by the hydrological monitoring terminal. The predicted volume is 1.2 million cubic meters per day. The standard deviation of this predicted value is... The daily volume is 180,000 cubic meters (obtained from the prediction model based on historical prediction error statistics). The server sets the probability threshold p to 95%, corresponding to the quantile coefficient. If the value is 1.645, then the server calculates the lower confidence limit for the water source. This means that, at a 95% confidence level, the daily runoff of the water source over the next seven days will not be less than 903,900 cubic meters.

[0061] After the server iterates through all water sources, it sums up the confidence lower limits of each source to obtain the minimum predicted available water volume for the current scheduling cycle (seven days). At the same time, the server will set the expected values ​​for each water source. By summing the results, we can obtain the predicted value of the expected available water volume. The server will... and The data is stored in the server's memory in the form of structured data and synchronously written to the server's persistent storage device as a scheduling log record.

[0062] Step 3: The dispatch center server performs the first phase of fixed allocation based on the minimum available water volume prediction value, generates a fixed allocation plan, and distributes it to each water terminal controller.

[0063] The dispatch center server uses the minimum available water volume predicted in step two. As the upper limit of available water volume constraints, a first-stage multi-objective optimization allocation model is constructed. The decision variables of this model are the allocated water volume for each water user terminal during each time period within the scheduling cycle. The objective function is to maximize the water supply guarantee rate for each water user unit. Constraints include the minimum water demand constraint for each water user unit, the upper limit constraint of the water supply network's transmission capacity, and the water supply volume from each water source not exceeding [a certain limit]. Total quantity constraints.

[0064] The dispatch center server calls a multi-objective optimization solution program stored in the server's memory. This program uses real-number encoding to encode the water allocation for each water terminal during each dispatch period as a gene locus on a chromosome, with each individual corresponding to a complete allocation scheme matrix. Non-dominated sorting employs a fast non-dominated sorting method, dividing the population into multiple Pareto fronts; crowding distance is calculated using Euclidean distance in the objective space. Selection is performed using binary tournament selection, crossover using simulated binary crossover (SBX), and mutation using polynomial mutation. Constraint handling follows a feasibility-first rule, meaning any individual violating constraints has a lower priority than all feasible individuals in the non-dominated sorting. The solution program uses a non-dominated sorting genetic algorithm (NSGA-II) as its engine. The server sets the algorithm parameters as follows: population size of 200 individuals, maximum number of generations of 500, crossover probability of 80%, and mutation probability of 5%. The server runs the solution program... Iterative calculations under constraints generate a set of Pareto optimal solutions, each solution corresponding to an allocation scheme, which includes the water allocation values ​​for each water-using terminal at each time period.

[0065] The dispatch center server selects a solution from the Pareto optimal solution set as the first-stage fixed allocation scheme according to the currently preset decision rules. In this embodiment, the current state is under a blue drought warning, and the preset decision rule is to prioritize ensuring domestic water supply. The server selects the solution with the highest domestic water supply guarantee rate from the Pareto optimal solution set as the final scheme.

[0066] The first-phase fixed allocation scheme includes the allocated water volume, valve opening percentage, or pump operating frequency for each of the 235 water terminal controllers. The dispatch center server encapsulates the corresponding allocation data into fixed allocation instruction messages based on the network address of each water terminal controller. Each fixed allocation instruction message includes a header, body, and trailer. The header contains a target terminal identifier field and an instruction type field (in this embodiment, the instruction type field is set to "01", indicating a fixed allocation instruction). The body contains the allocated water volume, valve opening percentage, or pump operating frequency, and a timestamp indicating the effective time. The trailer contains a CRC checksum field.

[0067] The dispatch center server sequentially sends the aforementioned fixed allocation instruction messages to 235 water terminal controllers via the communication network. Upon receiving the fixed allocation instruction message, each water terminal controller extracts the CRC checksum field at the end of the message, performs a checksum calculation, and compares the result with the CRC value carried in the message. If the checksum passes, the terminal controller parses the instruction value in the message body and controls the corresponding actuator through the drive interface—for pump station controllers, this adjusts the inverter output frequency to regulate the water supply. The drive interface is an RS-485 serial communication interface based on the Modbus RTU protocol, and the terminal controller communicates with the actuator using a master-slave query-response method. The terminal controller internally stores the flow-opening characteristic curve or flow-frequency characteristic curve of the actuator, stored in the form of a piecewise linear interpolation table. After the terminal controller parses the water allocation command value, it calculates the corresponding valve opening percentage or inverter frequency value based on the current pipeline pressure (collected in real time by the pressure sensor built into the terminal controller) by querying the interpolation table. Then, it sends the control value to the programmable logic controller (PLC) or inverter of the actuator via Modbus write register instructions. For valve controllers, it adjusts the electric actuator to adjust the valve opening; for gate controllers, it adjusts the hoist to adjust the gate opening—adjusting the water supply of this terminal to the command value. Terminal controllers that fail verification do not execute the command and reply to the dispatch center server via the communication network with a retransmission request message containing the terminal identifier and message sequence number. After receiving the retransmission request, the dispatch center server retransmits the corresponding fixed allocation command message.

[0068] Once the aforementioned first-phase fixed allocation scheme is issued and implemented, each water terminal controller will maintain the current fixed water allocation without changing according to short-term forecast data fluctuations unless it receives a cancellation or correction instruction from the dispatch center server.

[0069] It should be noted that the multi-objective optimization solution program can employ various known multi-objective evolutionary algorithms as the solution engine, including but not limited to the Non-Dominated Sorting Genetic Algorithm (NSGA-II), the Dream Optimization Algorithm, and the Decomposition-Based Multi-Objective Evolutionary Algorithm (MOEA / D). Those skilled in the art can select an appropriate algorithm based on actual computing resources and accuracy requirements. The core of this invention lies in the two-stage rolling optimization scheduling architecture, which consists of a first-stage fixed allocation based on the minimum available water volume constraint accumulated with confidence lower limits and a second-stage dynamic gain allocation based on real-time surplus, rather than the specific optimization algorithm itself.

[0070] Step 4: The dispatch center server receives real-time monitoring data and calculates the allocable surplus water.

[0071] During the current scheduling cycle, each hydrological monitoring terminal continuously collects real-time monitoring data such as actual runoff and actual rainfall at predetermined time intervals. In this embodiment, each hydrological monitoring terminal collects data once per hour and encapsulates the collected real-time monitoring data into real-time monitoring data messages, which are then reported to the scheduling center server via the communication network.

[0072] After receiving real-time monitoring data packets from various hydrological monitoring terminals, the dispatch center server verifies and validates the packets. Messages that pass verification are parsed and stored in the server's real-time database, while messages that fail verification are discarded and logged.

[0073] The dispatch center server reads the latest acquired measured data from the real-time database and calls the short-term rolling prediction program stored in the server's memory. In this embodiment, the short-term rolling prediction program uses an autoregressive moving average (ARMA) model, and the model order is set to the autoregressive order. Equal to 2. Moving average order The value equals one. The server inputs the latest measured data sequence into the ARMA model, performs rolling correction predictions on the available water volume for the remaining period within the current scheduling cycle, and outputs the corrected expected available water volume prediction value. .

[0074] The server follows the formula Calculate the allocable surplus water. Server makes a judgment. The value: If If the value is greater than zero, proceed to step five to perform the second stage of gain allocation; if... If the surplus is less than or equal to zero, the server generates a surplus insufficiency warning message, sends it to the administrator terminal via the communication network, suspends the second-stage allocation process, and proceeds to step six for feedback correction. The warning message is structured data in JSON format, containing a warning type code (0xE001 indicates surplus insufficiency), the current scheduling cycle identifier, the predicted water shortage value, and suggested handling measures. The administrator terminal is an industrial control computer running water resource scheduling and monitoring client software, which provides warning pop-up prompts and audible and visual alarms. After receiving the warning, the administrator can send manual intervention instructions to the scheduling center server through the client software, including but not limited to: adjusting the distribution parameters of each water source, activating backup water sources, or initiating emergency water conservation plans. Upon receiving the manual intervention instructions, the server executes the corrected allocation according to the administrator's instructions, and the priority of the manual intervention instructions is higher than that of the automatic rolling optimization process. After receiving the manual intervention instructions, the server suspends the execution of automatic rolling optimization steps four to six within the current scheduling cycle (i.e., it no longer triggers recalculation and distribution periodically) until the administrator sends a command to resume automatic optimization through the client software, at which point the server restarts the rolling optimization process.

[0075] Taking the data from the third day of the scheduling cycle in this embodiment as an example: Based on the actual runoff data from the previous three days, combined with the latest rainfall correction data issued by the meteorological forecast server, the server obtains the corrected expected available water volume prediction value after rolling prediction using the ARMA model. The volume is 7.8 million cubic meters (the sum of the predicted volume from the fourth to the seventh day and the actual water volume received in the first three days). (Server calculation) 1 million cubic meters, If the result is greater than zero, proceed to step five.

[0076] Step 5: The dispatch center server performs the second-stage dynamic gain allocation based on the allocable surplus water volume, generates a gain allocation scheme, and sends it to the corresponding water terminal controller.

[0077] When the available surplus water When the value is greater than zero, the dispatch center server initiates the second-stage gain allocation process. The dispatch center server reads the actual water usage status data of each water terminal from the memory cache. The actual water usage status data includes the current actual water supply of each terminal, the current water level of the reservoir, the water demand completion rate, and the output value per unit of water volume.

[0078] The dispatch center server allocates surplus water volume As an allocable incremental constraint, a gain allocation model is constructed. The objective function of the gain allocation model is to maximize the comprehensive benefit of surplus water allocation. The decision variable is the additional water allocation value of each water terminal controller. The constraint condition includes that the total additional water allocation does not exceed the specified value. Constraints on the upper limit of water conveyance capacity at each terminal and the upper limit of reservoir capacity of water storage facilities.

[0079] The dispatch center server invokes the dream optimization algorithm solver stored in the server's memory to solve the gain allocation model. The dream optimization algorithm solver executes the following sub-steps: Sub-step S5.1: Population initialization.

[0080] The dispatch center server generates an initial population of N individuals within the search space. The lower and upper bounds of the search space correspond to the minimum and maximum additional water allocation values ​​for each water terminal, respectively, and N is a preset population size parameter. The position vector of each individual corresponds to a set of gain allocation schemes, and the values ​​of each dimension represent the additional water allocation value for the corresponding water terminal.

[0081] Sub-step S5.2: Iterative calculation during the exploration phase.

[0082] The scheduling center server follows a preset threshold for the number of iterations during the exploration phase. Perform the following dream simulation operation: The scheduling center server divides the current population into five subgroups based on the memory capacity parameters of each individual, with each subgroup corresponding to a different forgetting dimension parameter. The forgetting dimension number parameter represents the number of decision variable dimensions that an individual randomly forgets and re-updates during a dream iteration.

[0083] For each subgroup, the scheduling center server performs the following operations: resets the position information of all individuals in the subgroup to the position information of the best individual in the subgroup in the previous iteration; and adjusts the forgetting dimension parameter according to the subgroup. Randomly selected from all decision variable dimensions One dimension; only for the selected dimensions. The location information of each dimension is calculated and updated according to the dream update formula, while the unselected dimensions retain their current values.

[0084] The dream update formula is: ,in Indicates the first The individual in the first The position of the next iteration. Indicates the first The subgroup in the th The optimal individual position in the next iteration. A random number between 0 and 1. For those between 0 and Random angle values ​​between.

[0085] The above operation simulates the "partial forgetting" mechanism in human dreams: individuals forget and reconstruct only some dimensions of information in dreams, while retaining the original memory of the remaining dimensions.

[0086] Sub-step S5.3: Fine-grained search during the development phase.

[0087] When the number of iterations exceeds the threshold for the exploration phase iterations Afterwards, the scheduling center server enters the development phase. During the development phase, the scheduling center server uses the current globally optimal individual's position as the center and performs a local fine-grained search within a preset radius to generate new candidate solutions.

[0088] In each iteration, the scheduling center server follows the formula ,in This represents the current globally optimal individual position. This is the preset local search radius decay coefficient. A random number between -1 and 1. Local search radius. As the number of iterations increases, the decay rate gradually decreases to simulate the cognitive process from "dream divergent exploration" to "memory solidification and convergence".

[0089] Sub-step S5.4: Fitness assessment and population renewal.

[0090] In each iteration, the scheduling center server calculates the fitness value of each individual according to the objective function of the gain allocation model. The scheduling center server sorts the current population from high to low fitness value, retains the top N individuals with the highest fitness value as the next generation population, and eliminates the remaining individuals.

[0091] Sub-step S5.5: Iteration termination and output of optimal solution.

[0092] The scheduling center server repeatedly executes sub-steps S5.2 to S5.4 until the preset maximum number of iterations threshold is reached. The scheduling center server outputs the position vector of the current globally optimal individual as the second-stage dynamic gain allocation scheme.

[0093] The second-stage dynamic gain allocation scheme includes the identifier of the water terminal controller that needs to allocate additional water and its corresponding additional water allocation value.

[0094] The dispatch center server encapsulates the gain allocation scheme into incremental allocation command messages and sends them to the corresponding water terminal controllers via the communication network. Upon receiving the incremental allocation command message, each water terminal controller performs a checksum and verification. If the verification passes, it parses the additional water volume value in the message body and adds the additional water volume to the current operating state: for valve controllers, it calculates the corresponding additional opening value based on the additional water volume value and adds this additional opening value to the current opening; for pump station controllers, it calculates the corresponding additional frequency value based on the additional water volume value and adds this additional frequency value to the current frequency.

[0095] Step 6: The scheduling center server receives the terminal execution feedback data and continuously corrects the model parameters.

[0096] During the execution of allocation instructions, each water terminal controller continuously collects local actual operating data, including actual water output, instantaneous pipeline flow rate, water tank level change, actual valve opening degree, or actual pump operating frequency. Each terminal controller encapsulates the above actual operating data into a terminal operating status message at predetermined time intervals (once per hour in this embodiment) and reports it to the dispatch center server via the communication network.

[0097] After receiving the operational status reports from each terminal, the dispatch center server extracts the actual water output data from each terminal and compares this actual water output with the allocation instruction values ​​issued during the corresponding time period (including the sum of the first-stage fixed allocation instruction values ​​and the second-stage incremental allocation instruction values), according to the formula... Calculate the execution deviation rate for each terminal, where This represents the actual water output. Allocate water volume according to the command.

[0098] The server aggregates the execution deviation rate data from each terminal and uses it as feedback input to call the parameter correction program stored in the server's memory. This parameter correction program uses the recursive least squares method to update the regression coefficients of the ARMA rolling prediction model in step four online, thereby gradually improving the accuracy of subsequent rolling predictions.

[0099] Simultaneously, the dispatch center server compares the actual water inflow data with the theoretical values ​​calculated based on the predicted data in step two, and calculates the prediction error for each prediction source. Based on the cumulative prediction error of each prediction source, the server updates the distribution parameters of each prediction source using a weighted recursive average method. The updated distribution parameters are used for the calculation of the confidence lower limit in step two of the next scheduling cycle.

[0100] Steps four through six above are executed cyclically at predetermined time intervals (every hour in this embodiment) within the current scheduling cycle until the current scheduling cycle ends, forming a closed-loop rolling optimization process.

[0101] To verify the effectiveness of the method of the present invention, this embodiment uses the seven-day scheduling cycle of the watershed as a simulation scenario, sets the same hydrological input conditions, and compares the method of the present invention with the following two traditional methods: 1) Static scheduling method: The "single prediction, single allocation" model is adopted. A fixed allocation plan is formulated based on the predicted expected value before the start of the scheduling cycle, and it is not adjusted during the cycle.

[0102] 2) Gain Allocation Only Method: Using the minimum predicted value of a single water source as a constraint, the first-stage fixed allocation is not performed; instead, gain allocation is performed only when there is a surplus.

[0103]

[0104] As shown in Table 1: (1) Regarding the water supply guarantee rate, the method of this invention achieves a domestic water supply guarantee rate of 98.5%, which is 7.3 percentage points higher than the static scheduling method and 4.8 percentage points higher than the gain allocation method alone. This indicates that using the confidence lower limit accumulation value as a fixed allocation constraint effectively ensures basic water supply security under extreme prediction deviations.

[0105] (2) In terms of comprehensive benefits. The average daily comprehensive economic benefit of the method of this invention is RMB 724,000, which is 24.2% higher than that of the static scheduling method and 15.5% higher than that of the gain allocation method alone. This shows that the second-stage gain allocation, through the dream optimization algorithm, accurately allocates the surplus water to high-efficiency terminals, which significantly improves the economic value of water resource utilization.

[0106] (3) In terms of execution accuracy and adjustment efficiency, the average deviation rate between the actual output and the commanded output of the method of this invention is only 2.1%, which is much lower than the 8.7% of the static scheduling method, indicating that the rolling feedback correction mechanism makes the allocation command more in line with the actual working conditions. The invention adjusts 4 times within the cycle, which is an orderly gain allocation, avoiding the disadvantage of static scheduling being unable to respond to hydrological changes, and also avoiding the problem of frequent adjustments in the gain allocation method due to the lack of a basic fixed scheme.

[0107] In summary, simulation results show that the method of the present invention is significantly superior to the traditional method in ensuring water supply security, improving economic efficiency, and controlling execution deviations, verifying the effectiveness and superiority of the two-stage architecture of fixed allocation in the early stage combined with dynamic regulation within the cycle.

[0108] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A multi-objective water resource optimization scheduling method based on a dream optimization algorithm, characterized in that, Includes the following steps: S1. Collect runoff forecast data messages, rainfall forecast data messages, and reclaimed water production forecast data messages. Verify and parse each data message to obtain the parsed forecast values ​​and corresponding distribution parameters. The distribution parameters include the expected values ​​of the forecast values ​​within the preset scheduling period. and the corresponding standard deviation ; S2. Based on the predicted values ​​obtained from step S1 and the corresponding distribution parameters, calculate the minimum available water volume prediction value and the expected available water volume prediction value. S3. Using the lowest predicted available water volume as the upper limit of available water volume constraint, construct and solve a multi-objective optimization allocation model to generate a first-stage fixed allocation scheme. The fixed allocation scheme remains constant during execution and includes the allocated water volume values ​​of each water terminal controller. S4. During the execution of the current scheduling cycle, based on the real-time monitoring data reports from the hydrological monitoring terminal, the short-term rolling prediction program is used to perform rolling correction prediction of the available water volume for the remaining period in the current scheduling cycle, and outputs the corrected expected available water volume prediction value. Based on the corrected expected available water volume prediction value and the minimum available water volume prediction value, the allocable surplus water volume is calculated. S5. When the allocable surplus water is greater than zero, the second stage of dynamic gain allocation is executed. Based on the actual water usage status data of each water terminal, the allocable surplus water is used as the allocable increment constraint to construct a gain allocation model. The dream optimization algorithm is used to solve the gain allocation model. S6. Collect terminal operation status messages reported by each water terminal controller at predetermined time intervals, extract actual water output data, compare the actual water output with the allocation instruction value for the corresponding time period, calculate the execution deviation rate, use the execution deviation rate as feedback input, update the model parameters of the short-term rolling prediction program online, and compare the actual water inflow data with the predicted value to update the distribution parameters of each water source. Steps S4 to S6 are executed cyclically at predetermined time intervals within the current scheduling cycle until the current scheduling cycle ends.

2. The multi-objective water resource optimization scheduling method based on the dream optimization algorithm according to claim 1, characterized in that, In step S2, the calculation process for the minimum predicted available water volume and the expected predicted available water volume is as follows: Expected values ​​of predicted values ​​for each water source As the mean, As a distribution parameter, it is assumed that the predicted available water volume of each water source follows a normal distribution. Set probability threshold The probability threshold is determined based on the inverse function of the cumulative distribution function of the standard normal distribution. quantile coefficients ; For each water source, calculate the probability threshold for that water source using the following formula. The lower confidence limit below The lower confidence limit indicates whether, under the predicted distribution of the water source, there is a certain level of confidence. The probability that the actual usable water volume is not less than this value: ； Lower confidence limit for all water sources By summing the results, we obtain the predicted minimum available water volume within the scheduling cycle. : ； The expected values ​​of each water source By summing the results, we obtain the expected available water volume forecast for the scheduling period. : 。 3. The multi-objective water resource optimization scheduling method based on the dream optimization algorithm according to claim 1, characterized in that, In step S3, a non-dominated sorting genetic algorithm is used to solve the multi-objective optimization allocation model to generate a set of Pareto optimal solutions. According to the preset decision rules, a scheme is selected from the Pareto optimal solution set as the first-stage fixed allocation scheme.

4. The multi-objective water resource optimization scheduling method based on the dream optimization algorithm according to claim 1, characterized in that, In step S4, the short-term rolling forecast program uses an autoregressive moving average model, inputting the latest measured data of each water source as the observed value into the autoregressive moving average model, and outputting the corrected expected available water quantity forecast value. .

5. A multi-objective water resource optimization scheduling method based on a dream optimization algorithm according to claim 1, characterized in that, In step S4, the allocable surplus water is the corrected predicted value of the expected available water. Compared with the minimum available water forecast The difference is calculated using the following formula: 。 6. The multi-objective water resource optimization scheduling method based on the dream optimization algorithm according to claim 1, characterized in that, In step S5, the actual water usage status data of each water terminal includes the current actual water supply of each terminal, the current water level of the reservoir, the water demand completion rate, and the output value per unit of water volume.

7. A multi-objective water resource optimization scheduling method based on a dream optimization algorithm according to claim 1, characterized in that, In step S5, the objective function of the gain allocation model is to maximize the comprehensive benefits of surplus water allocation, the decision variable is the additional allocation water volume value of each water terminal controller, and the constraints include the total additional allocation water volume not exceeding the allocable surplus water volume, the upper limit constraint of the water transmission capacity of each terminal, and the upper limit constraint of the reservoir capacity of the water storage facility.

8. A multi-objective water resource optimization scheduling method based on a dream optimization algorithm according to claim 1, characterized in that, In step S6, the model parameters of the short-term rolling prediction program are updated online using the recursive least squares method, and the distribution parameters of each water source are updated using the weighted recursive average method.

9. A water resource optimization scheduling system based on two-stage allocation, used to implement the method according to any one of claims 1-8, characterized in that, include: The dispatch center server, deployed in the water resources dispatch management center, is equipped with a processor, memory module, communication interface and persistent storage device. It is used to receive and cache multi-source prediction data, calculate the minimum available water volume prediction value and the expected available water volume prediction value, generate and issue the first-stage fixed allocation scheme, calculate the allocable surplus water volume, generate and issue the second-stage dynamic gain allocation scheme, and receive and roll over feedback data from the terminal. Multiple hydrological monitoring terminals are deployed at river monitoring sections and rainfall monitoring stations. They are equipped with water level sensors, flow sensors, communication modules and microcontrollers to collect raw monitoring data of runoff and rainfall and report them to the dispatch center server through the communication network. A weather forecast server, deployed at the weather forecast center, is connected to the dispatch center server via a communication link and is used to provide rainfall forecast data messages; The monitoring terminal for the reclaimed water treatment plant is deployed at the outlet of the reclaimed water treatment plant and is equipped with a flow sensor, a communication module and a microcontroller. It is used to collect data on the amount of reclaimed water produced and report it to the dispatch center server through the communication network. Multiple water terminal controllers are deployed in each water unit and are equipped with programmable logic controllers or embedded microcontrollers, communication modules, flow meter interfaces and valve / pump group drive interfaces. They are used to receive and execute fixed allocation instruction messages and incremental allocation instruction messages issued by the dispatch center server, and report terminal operation status messages. The communication network, including wired communication links and wireless communication links, is used for data packet transmission between the aforementioned hardware nodes.

10. A water resource optimization scheduling system based on two-stage allocation according to claim 9, characterized in that, The dispatch center server dynamically adjusts the time interval for each hydrological monitoring terminal to report real-time monitoring data through remote configuration messages: when the rate of change of the measured data continuously exceeds the preset high threshold, the reporting interval is shortened; when the rate of change of the measured data continuously falls below the preset low threshold, the reporting interval is extended.

Images