Water plant residual pressure power generation resource developable quantity accounting method and system

Abstract

The application discloses a water plant residual pressure power generation resource developable quantity accounting method and system, which comprises the following steps: collecting water conservancy and hydropower data and resource data of a preset water plant, and preprocessing the water conservancy and hydropower data and the resource data; identifying water energy resource development of the resource data to obtain a recommended development project list, and performing multi-target optimization arrangement on the recommended development project list to obtain arrangement data; identifying and determining main parameters of the arrangement data according to the water conservancy and hydropower data to obtain the main parameters, performing economic and simulation engineering evaluation on the recommended development project list according to the main parameters to obtain a feasibility degree, and performing resource calculation according to the main parameters to obtain development resource data; and constructing a water plant residual pressure power generation resource developable quantity accounting model according to the feasibility degree and the development resource data, inputting to-be-accounted data into the water plant residual pressure power generation resource developable quantity accounting model, and outputting an accounting result.

Description

Technical Field

[0001] This invention relates to the field of calculating the exploitable amount of waste pressure power generation resources, and in particular to a method and system for calculating the exploitable amount of waste pressure power generation resources in water plants. Background Technology

[0002] Guided by the "dual carbon" goals, the development and utilization of renewable energy has become a core direction for energy transformation. The residual pressure resources contained in water supply networks, as clean and low-carbon energy sources, possess enormous development potential. To ensure stable water supply pressure throughout the city, water plants need to maintain a high-pressure state in the network through pressurization equipment. However, the demand for water pressure at the end-user level is relatively low, and a large amount of residual pressure is dissipated through pressure reducing valves and other equipment, resulting not only in energy waste but also increased network operation losses.

[0003] Currently, the development of waste pressure power generation resources lacks a systematic method for calculating the exploitable amount, and existing technologies have significant limitations: First, the accuracy of resource identification is insufficient, making it difficult to locate high-potential development nodes in complex pipeline networks, and easily overlooking high-quality resources or misjudging inefficient locations; second, the calculation parameters do not fully consider the dynamic characteristics of the pipeline network, resulting in a large deviation between the predicted and actual power generation; third, the evaluation system is singular, failing to comprehensively consider the impact of economic feasibility, engineering reliability, and pipeline safety, which can easily lead to investment risks or operational hazards.

[0004] Meanwhile, traditional accounting methods have weak capabilities in integrating multi-source data, failing to effectively combine multi-dimensional information such as geospatial, hydrological, meteorological, and infrastructure data. Furthermore, they do not quantify the impact of parameter uncertainties on the results, leading to insufficient reliability. These problems severely restrict the large-scale and standardized development of waste pressure power generation resources in hydropower plants. Therefore, developing an accounting method that accurately identifies resources, scientifically quantifies potential, and comprehensively assesses feasibility is of significant practical importance for promoting the implementation of waste pressure power generation technology and improving energy utilization efficiency. Summary of the Invention

[0005] The purpose of this invention is to provide a method for calculating the exploitable amount of hydropower generation resources from residual pressure in water plants.

[0006] To achieve the above objectives, the present invention is implemented according to the following technical solution: This invention includes the following steps: Collect water conservancy and hydropower data and resource data from a pre-designated water plant, and preprocess the water conservancy and hydropower data and the resource data; the resource data includes geospatial data, hydrological and meteorological data, infrastructure data and socio-economic data; The resource data is used to identify hydropower resource development to obtain a list of recommended development projects, and the list of recommended development projects is then used for multi-objective optimization to obtain deployment data. Based on the water conservancy and hydropower data, the main parameters of the deployment data are identified and determined to obtain the main parameters. Based on the main parameters, the recommended development project list is evaluated economically and through simulated engineering to obtain the feasibility. Based on the main parameters, resource calculations are performed to obtain development resource data. Based on the feasibility level and the development resource data, a calculation model for the exploitable amount of water plant residual pressure power generation resources is constructed. The data to be calculated is input into the calculation model for the exploitable amount of water plant residual pressure power generation resources, and the calculation results are output.

[0007] Furthermore, the method for identifying hydropower resource development projects from the resource data to obtain a recommended development project list includes: By acquiring pipeline connection relationships, pipe diameter, material, length, valve / pressure reducing valve location, pressure, flow rate, historical records of annual operating hours, water density, and gravitational acceleration at each node through resource data, key locations in the pipeline network are defined as graph nodes; key locations include pressure gauge installation points, pressure reducing valve assemblies, water pump outlets, and high and low zone pipeline connection points; Treating the pipe as an edge in the graph, with the edge weight being the pressure loss coefficient, the expression is:

[0008] in This is the friction coefficient. Let be the length of the pipe from node i to node j. Let be the pipe diameter between the i-th node and the j-th node. Let be the average flow between the i-th node and the j-th node. Let be the edge weight between the i-th node and the j-th node; Node features are extracted using a graph convolutional network, including node pressure, flow rate, and number of connected pipes, to generate node embedding vectors. These vectors are then combined with PageRank to calculate node importance scores.

[0009] in This is the pipe damping coefficient. Let the initial importance of the i-th node be... The importance score is given to the i-th node. For the i-th node, For the j-th node, For nodes The set of all direct neighbor nodes, For nodes and nodes Edge weights; The nodes are sorted in descending order of importance scores to obtain a preliminary list of node priorities. For the top 30% of nodes, their inlet and outlet pressure differences are calculated. Nodes with an inlet and outlet pressure difference greater than 0.2 MPa and an annual operating time greater than 6000 hours are selected as candidate resource points. The comprehensive potential of nodes is calculated based on importance score, annual operating hours, and import / export pressure difference. Nodes are then ranked according to their comprehensive potential using a comprehensive index, expressed as follows:

[0010] in For nodes The comprehensive potential For normalized nodes Annual operating hours, For normalized nodes The difference between import and export pressures The importance score after normalization; Remove duplicate points on the same branch of the pipeline that are less than 500m apart, and output the list of candidate resource points after removal as a recommended development project list according to comprehensive indicators.

[0011] Furthermore, the method for obtaining deployment data by multi-objective optimization deployment of the recommended development project list includes: Based on the optimization objectives of maximizing power generation efficiency, minimizing investment costs, and minimizing the impact on pipeline safety, an objective function is constructed, expressed as follows:

[0012]

[0013]

[0014] in For real-time traffic, For real-time residual pressure head, For the efficiency of the water turbine, Annual operating hours For the density of water, It is the acceleration due to gravity. For the length of the pipe, For pipe diameter, For power generation efficiency, For investment costs, To ensure the safety of the pipeline network, The pressure value of the node before deployment. This represents the pressure value at the nodes after deployment. For the rated head, For rated flow rate, For installation costs, For pipeline renovation costs, For equipment procurement costs, The number of nodes; Constraints include engineering constraints, equipment constraints, and hydraulic constraints. Engineering constraints include: the installation location must meet the pipeline's load-bearing requirements; equipment constraints include: the actual flow rate of the turbine. Hydraulic constraints: Pipeline flow velocity less than or equal to 2.5 m / s; The third-generation non-dominated sorting genetic algorithm is used to design and distribute the data according to the objective function: multiple initial solutions are randomly generated. ,in The coordinates are the installation location coordinates; hydraulic simulation is performed to verify each solution, and calculations are performed. , , Values ​​are used to filter individuals that violate hard constraints; The population is sorted in a non-dominated manner, divided into different Pareto levels, the vertical distance between each individual and the dynamic reference point is calculated, high-quality individuals are selected and retained through a tournament, and adaptive crossover and mutation are performed on the selected individuals to generate the offspring population. The parent and offspring populations are merged, and the non-dominated sorting is re-established. For each reference point, the closest individual is retained, and a new generation of population is formed. This process is repeated until the maximum number of iterations is reached, and the final Pareto optimal solution set is output. Each solution corresponds to a complete set of deployment data, which includes equipment signals, number of installation locations, and pipeline parameters.

[0015] Furthermore, the method for identifying and determining the main parameters of the deployment data based on the aforementioned water conservancy and hydropower data to obtain the main parameters includes: Latin hypercube sampling was used to generate 20 initial parameter combinations from water conservancy and hydropower data and deployment data. The error between the corresponding simulation results and the actual data was calculated using hydraulic simulation software to form the initial dataset. ;in Let a be the three-dimensional optimization parameter vector for group a. The root mean square error, For flow coefficient, This is the head correction factor. This refers to the pipe damping coefficient; A Gaussian process is used as a surrogate model to model the nonlinear relationship between parameters and error; its mean function is set as a constant function, and the covariance function is a radial basis function, expressed as:

[0016] in For signal variance, For length scale, For parameter vectors and covariance value, For parameter vectors and Euclidean distance; Minimizing the error between simulation results and actual data is taken as the objective function. An expected improvement function is used to balance exploration and development, expressed as:

[0017] in The mean value predicted by the surrogate model. The standard deviation of the surrogate model's prediction. This is the current minimum error. To explore the coefficient, Let be the probability density function of the standard normal distribution. The cumulative distribution function of the standard normal distribution. For the function to be improved; Based on the current dataset Train a Gaussian process model, predict the mean and standard deviation of each point in the parameter space, maximize the expected improvement function, select the next set of parameters to be evaluated, and perform hydraulic simulation calculations. Update the dataset Iterate continuously until the error is less than the error threshold; where Let be the objective function. For the first The parameter vector for the next iteration. For the first The root mean square error of the next iteration; When the absolute value of the error change over 5 consecutive iterations is less than or equal to the error change threshold, the iteration stops and the optimal parameters are output.

[0018] Furthermore, the method for obtaining the feasibility level by conducting economic and simulated engineering assessments of the recommended development project list based on the aforementioned key parameters includes: Based on the key parameters and the recommended development project list, obtain the design flow rate, residual pressure head, equipment efficiency, annual operating hours, initial investment, annual operating cost, project life, electricity price volatility, subsidy amount, and subsidy duration. Calculate the actual power generation:

[0019] in This represents the actual annual power generation. For design traffic, For the remaining pressure head, For overall equipment efficiency, Annual operating hours For the density of water, It is the acceleration due to gravity; Calculate annual cash flow and project net present value:

[0020]

[0021] in For the net cash flow in year s, The electricity price in year s, Annual operating costs To calculate the cumulative operating period, The year that the subsidy will be disbursed will be the last year. For the amount of subsidy, For initial investment, Let be the net present value of the project in year s. The benchmark discount rate; Geometric Brownian motion is used to simulate the stochastic process of electricity prices. Multiple future electricity price paths are generated through Monte Carlo simulation, and the electricity price distribution under each path is obtained, expressed as:

[0022] in Electricity price drift rate, For electricity price volatility, For the standard Wiener process in year s; Based on a binomial tree model, assuming managers can choose to invest immediately or delay investment for one year, with the delay cost being 5% of the initial investment; a binomial tree for electricity prices is constructed, including an increase factor. Decreasing factor and risk-neutral probability Calculate the value of the deferred option based on the deferred investment. ;in After the electricity price increase Values, After the drop in electricity prices Values, The optimal value of investing after a one-year delay; among which For time step; Assuming the project has negative cash flow in a given year, it can be abandoned and its residual value can be obtained. Calculate the annual cash flow under different electricity price paths, working backward from the end of the project's lifespan, to calculate the value of continuing operation versus abandonment: , Residual value To determine the optimal value of investing after a delay of s+1 years, To determine the optimal value of the investment after a delay of s years, the average value of multiple paths simulated is taken as the value of the forfeited option. ; The net present value of a project is adjusted based on the value of the delayed option and the value of the forfeited option, expressed as follows:

[0023] in To determine the feasibility level, a weighted scoring system is constructed based on the adjusted net present value (NPV) and risk indicators. The evaluation dimensions of the weighted scoring system include economic value, technological reliability, risk control, and policy stability.

[0024] Furthermore, the method for obtaining development resource data by performing resource calculations based on the aforementioned key parameters includes: The impact of uncertainties in key parameters on development resource data is quantified through Monte Carlo simulation, generating probability distribution results: the distribution type and parameters are determined based on historical data or industry standards, including inlet and outlet flow rates, residual pressure head, and equipment efficiency; the probability distribution interval of each parameter is divided into multiple sub-intervals, and one sample point is randomly selected in each sub-interval. The selected samples are randomly combined to form multiple independent parameter combinations; development resource data for a single simulation is calculated based on the hydraulic power formula; the distribution curves of multiple annual power generation samples are fitted by kernel density estimation, key quantiles are calculated, and a 95% confidence interval is calculated using a bootstrap sampling method.

[0025] Furthermore, the method for constructing a calculation model for the exploitable amount of hydropower generation resources based on the aforementioned feasibility level and the aforementioned development resource data includes: Based on the feasibility level and development resource data, obtain spatiotemporal feature vectors, deployment parameters, and feasibility scores; use one-hot encoding for the classification features in the deployment parameters, and perform min-max normalization to the [0,1] interval for the numerical features; and weight and fuse the economic and engineering feasibility scores into a comprehensive feasibility vector. The model for calculating the exploitable amount of hydropower generation resources from waste pressure includes an input layer, a Transformer encoder, a reinforcement learning optimization module, and an output layer. Input layer: Based on the location encoding, temporal and location information is added to the spatiotemporal features. The standardized spatiotemporal features are then added to the location embedding to obtain the spatiotemporal embedding vector. The expression is:

[0026]

[0027] in For time steps, Indexed by feature dimensions, For embedded dimensions, Encode the time location vector; Spatiotemporal embedding vectors Layout parameter embedding vector Feasibility scoring embedding vector Concatenate the sequences according to the time dimension to form the input sequence X; Transformer encoder: Takes the input sequence X and generates query, key, and value matrices through a linear transformation, respectively. , , ,in To query learnable parameters, For key learnable parameters, For learnable parameters, For query, As key, The output is the value; each attention head is output, the outputs of multiple attention heads are concatenated and subjected to a linear transformation to obtain the multi-head self-attention output. Each encoder layer consists of a first layer and a second layer, and outputs a context feature vector. The first layer consists of multi-head self-attention, residual connections, and layer normalization; the second layer consists of a feedforward neural network, residual connections, and layer normalization. The expression is:

[0028]

[0029]

[0030] in It is a feedforward neural network. These are intermediate features following the multi-head self-attention module. The final output features of a single encoder layer, For layer normalization, The weight matrix is ​​the first-level linear transformation. This is the bias vector for the first-level linear transformation. This is the weight matrix for the second-level linear transformation. This is the bias vector for the second-level linear transformation; Reinforcement Learning Optimization Module: A proximal policy-based optimization reinforcement learning module is introduced to dynamically adjust the encoder. The encoder's state consists of the current context feature vector and historical prediction error, while the action is the attention weight adjustment coefficient. The reward function aims to maximize prediction accuracy. The policy network is updated by optimizing the pruning objective function, and the optimized context feature vector is output. ; Output layer: The vector is compressed into a fixed-dimensional vector using global average pooling, and then the predicted value is output through a fully connected network. The 95% confidence interval is calculated by introducing a Gaussian distribution parameterization into the output layer of the Bayesian neural network.

[0031] Secondly, a system for calculating the exploitable resources for hydropower generation from residual pressure in water plants includes: Data acquisition and processing module: used to collect water conservancy and hydropower data and resource data of the preset water plant, and to preprocess the water conservancy and hydropower data and resource data; Project identification and screening module: used to identify hydropower resource development from the resource data to obtain a list of recommended development projects, and to perform multi-objective optimization deployment of the recommended development project list to obtain deployment data; Parameter identification and feasibility assessment module: used to identify and determine the main parameters of the layout data based on the water conservancy and hydropower data, to conduct economic and simulated engineering assessments of the recommended development project list based on the main parameters to obtain the degree of feasibility, and to calculate development resource data based on the main parameters. Model building and output module: This module is used to build a calculation model for the exploitable amount of water plant residual pressure power generation resources based on the feasibility level and the development resource data. It inputs the data to be calculated into the calculation model for the exploitable amount of water plant residual pressure power generation resources and outputs the calculation results.

[0032] The beneficial effects of this invention are: This invention provides a method and system for calculating the exploitable resources for hydropower generation from residual pressure in water plants. Compared with existing technologies, this invention has the following technical advantages: This invention uses graph convolutional networks combined with Page Ranking to accurately identify high-potential waste pressure development nodes, eliminating invalid sites and improving resource identification efficiency and accuracy. A multi-dimensional evaluation system integrates economic, engineering, and pipeline safety factors, incorporating option value to correct net present value and quantify parameter uncertainty, making feasibility assessments more realistic. A computational model integrating Transformer and reinforcement learning is constructed to dynamically optimize feature extraction and output computational results with confidence intervals, significantly improving prediction accuracy. Simultaneously, the entire process achieves data-driven automated computation, balancing power generation efficiency, investment costs, and pipeline safety, providing a scientific and reliable quantitative basis for the large-scale development of waste pressure power generation resources in water plants. Attached Figure Description

[0033] Figure 1 This is a flowchart illustrating the steps of a method for calculating the exploitable amount of hydropower generation resources from residual pressure in a water plant, as described in this invention. Detailed Implementation

[0034] The present invention will be further described below through specific embodiments. The illustrative embodiments and descriptions herein are used to explain the present invention, but are not intended to limit the present invention.

[0035] The present invention provides a method and system for calculating the exploitable resources of hydropower generation from residual pressure in water plants, comprising the following steps: like Figure 1 As shown, this embodiment includes the following steps: Collect water conservancy and hydropower data and resource data from a pre-designated water plant, and preprocess the water conservancy and hydropower data and the resource data; the resource data includes geospatial data, hydrological and meteorological data, infrastructure data and socio-economic data; In the actual assessment, a water supply plant in a certain city was used as the accounting object, with a water supply capacity of 150,000 m³. 3 / d, water supply network coverage area 28km 2 There are a total of 326 pipeline nodes, 18 pressure reducing valve groups, and 3 water pump stations. The main water supply pipelines have a diameter of DN200~DN1200 and a pipeline operating pressure of 0.3~0.8MPa. Water conservancy and hydropower data and resource data for the water plant were collected over the past three years (2022-2024). Resource data includes four main categories: geospatial, hydrological and meteorological, infrastructure, and socioeconomic data. Core data includes: Water conservancy and hydropower data: average node pressure 0.45 MPa, average flow rate 120 m³ / h. 3 / h; Rated pump head 80m, efficiency 82%; Geospatial data: Water plant ground elevation 28.5m, highest node elevation of pipeline network 45.2m, pipeline burial depth 1.5~3.0m; Hydrological and meteorological data: Annual average temperature 18.6℃, annual rainfall 1250mm, no extreme hydrological impacts; Infrastructure data: Main pipeline material ductile iron, DN800 pipeline length 6.8km; Pressure reducing valve group is all pilot operated, nominal pressure 1.0MPa; Socioeconomic data: Industrial electricity price 0.65 yuan / (kW・h), annual operating labor cost 80,000 yuan / station; New energy subsidy 0.12 yuan / (kW・h), for 5 years; Preprocessing: Outlier removal, missing value completion, and data standardization are performed on the collected data; The resource data is used to identify hydropower resource development to obtain a list of recommended development projects, and the list of recommended development projects is then used for multi-objective optimization to obtain deployment data. Based on the water conservancy and hydropower data, the main parameters of the deployment data are identified and determined to obtain the main parameters. Based on the main parameters, the recommended development project list is evaluated economically and through simulated engineering to obtain the feasibility. Based on the main parameters, resource calculations are performed to obtain development resource data. Based on the feasibility level and the development resource data, a calculation model for the exploitable amount of water plant residual pressure power generation resources is constructed. The data to be calculated is input into the calculation model for the exploitable amount of water plant residual pressure power generation resources, and the calculation results are output.

[0036] In this embodiment, the method for identifying hydropower resource development projects from the resource data to obtain a recommended development project list includes: By acquiring pipeline connection relationships, pipe diameter, material, length, valve / pressure reducing valve location, pressure, flow rate, historical records of annual operating hours, water density, and gravitational acceleration at each node through resource data, key locations in the pipeline network are defined as graph nodes; key locations include pressure gauge installation points, pressure reducing valve assemblies, water pump outlets, and high and low zone pipeline connection points; Treating the pipe as an edge in the graph, with the edge weight being the pressure loss coefficient, the expression is:

[0037] in This is the friction coefficient. Let be the length of the pipe from node i to node j. Let be the pipe diameter between the i-th node and the j-th node. Let be the average flow between the i-th node and the j-th node. Let be the edge weight between the i-th node and the j-th node; Node features are extracted using a graph convolutional network, including node pressure, flow rate, and number of connected pipes, to generate node embedding vectors. These vectors are then combined with PageRank to calculate node importance scores.

[0038] in This is the pipe damping coefficient. Let the initial importance of the i-th node be... The importance score is given to the i-th node. For the i-th node, For the j-th node, For nodes The set of all direct neighbor nodes, For nodes and nodes Edge weights; The nodes are sorted in descending order of importance scores to obtain a preliminary list of node priorities. For the top 30% of nodes, their inlet and outlet pressure differences are calculated. Nodes with an inlet and outlet pressure difference greater than 0.2 MPa and an annual operating time greater than 6000 hours are selected as candidate resource points. The comprehensive potential of nodes is calculated based on importance score, annual operating hours, and import / export pressure difference. Nodes are then ranked according to their comprehensive potential using a comprehensive index, expressed as follows:

[0039] in For nodes The comprehensive potential For normalized nodes Annual operating hours, For normalized nodes The difference between import and export pressures The importance score after normalization; Remove duplicate points on the same branch of the pipeline that are less than 500m apart, and output the list of candidate resource points after removal as a recommended development project list according to comprehensive indicators; In the actual assessment, 68 key locations in the pipeline network, including pressure gauge installation points, pressure reducing valve assemblies, pump outlets, and high / low zone pipeline connection points, were defined as graph nodes. Pipelines were considered graph edges. Core parameters were extracted from resource data, and edge weights were calculated. The core parameters and calculation results are as follows: Node pair 12-15, pressure reducing valve assembly, pipeline length 320m, pipe diameter 0.8m, average flow rate 0.045m³ / h. 3 / s, friction factor is 0.021, edge weight is 0.0086; node pair 28-30, pump outlet, pipe length 180m, pipe diameter 1m, average flow rate 0.062m³ / s. 3 / s, friction factor is 0.018, edge weight is 0.0052; node pair 56-58, high-low zone connection point, pipe length 450m, pipe diameter 0.6m, average flow rate 0.028m³ / s. 3 / s, the friction factor is 0.023, and the edge weight is 0.0105; A two-layer GCN graph convolutional network is used to extract node features (node ​​pressure, flow rate, number of connecting pipes, 3D input), generate a 64-dimensional node embedding vector and perform L2 normalization. The mean of the embedding vector is used as the initial importance of the node. The node importance score is calculated by combining Page Ranking Formula. The pipe damping coefficient is 0.75. The scores of the 68 nodes were sorted in descending order, and the top 30% were selected. The importance scores of some nodes are as follows: Node 13, pressure reducing valve assembly, initial importance 0.82, importance score 0.915; Node 30, water pump outlet, initial importance 0.78, importance score 0.876; Node 58, high-low zone connection point, initial importance 0.75, importance score 0.832; Node 18, pressure reducing valve assembly, initial importance 0.72, importance score 0.798. The inlet and outlet pressure differences and annual operating hours of 20 nodes were calculated. Nodes with pressure differences > 0.2 MPa and annual operating hours > 6000 h were selected, resulting in 12 candidate resource points: Node 15, inlet and outlet pressure difference 0.32 MPa, annual operating hours 8760, node flow rate 150 m³ / h. 3 / h; Node 30, inlet and outlet pressure difference 0.28MPa, annual operating hours 8760, node flow rate 220m³ / h 3 / h; Node 58, inlet and outlet pressure difference 0.25MPa, annual operating hours 8200, node flow rate 95m³ / h. 3 / h; Node 18, inlet and outlet pressure difference 0.26MPa, annual operating hours 8500, node flow rate 110m³ / h. 3 / h; Node 42, inlet and outlet pressure difference 0.23MPa, annual operating hours 6800, node flow rate 80m³ / h. 3 / h; The comprehensive potential of 12 candidate nodes was calculated, and duplicate points on the same branch with a distance of less than 500m were removed (3 in total). Finally, a list of recommended development projects consisting of 9 nodes was obtained. The top 5 nodes in terms of comprehensive potential are: Node 15, comprehensive potential 0.926, West trunk line, meets pipeline load-bearing requirements, open site; Node 30, comprehensive potential 0.889, East trunk line, meets pipeline load-bearing requirements, near a pumping station; Node 58, comprehensive potential 0.845, South trunk line, meets pipeline load-bearing requirements, requires simple modification; Node 18, comprehensive potential 0.802, West trunk line, meets pipeline load-bearing requirements, open site; Node 42, comprehensive potential 0.768, North trunk line, meets pipeline load-bearing requirements, near a valve well.

[0040] In this embodiment, the method for obtaining deployment data by multi-objective optimization deployment of the recommended development project list includes: Based on the optimization objectives of maximizing power generation efficiency, minimizing investment costs, and minimizing the impact on pipeline safety, an objective function is constructed, expressed as follows:

[0041]

[0042]

[0043] in For real-time traffic, For real-time residual pressure head, For the efficiency of the water turbine, Annual operating hours For the density of water, It is the acceleration due to gravity. For the length of the pipe, For pipe diameter, For power generation efficiency, For investment costs, To ensure the safety of the pipeline network, The pressure value of the node before deployment. This represents the pressure value at the nodes after deployment. For the rated head, For rated flow rate, For installation costs, For pipeline renovation costs, For equipment procurement costs, The number of nodes; Constraints include engineering constraints, equipment constraints, and hydraulic constraints. Engineering constraints include: the installation location must meet the pipeline's load-bearing requirements; equipment constraints include: the actual flow rate of the turbine. Hydraulic constraints: Pipeline flow velocity less than or equal to 2.5 m / s; The third-generation non-dominated sorting genetic algorithm is used to design and distribute the data according to the objective function: multiple initial solutions are randomly generated. ,in The coordinates are the installation location coordinates; hydraulic simulation is performed to verify each solution, and calculations are performed. , , Values ​​are used to filter individuals that violate hard constraints; The population is sorted in a non-dominated manner, divided into different Pareto levels, the vertical distance between each individual and the dynamic reference point is calculated, high-quality individuals are selected and retained through a tournament, and adaptive crossover and mutation are performed on the selected individuals to generate the offspring population. The parent and offspring populations are merged, and the non-dominated sorting is re-established. For each reference point, the nearest individual is retained, and a new generation of population is formed. This process is repeated until the maximum number of iterations is reached, and the final Pareto optimal solution set is output. Each solution corresponds to a complete set of deployment data, which includes equipment signals, number of installation locations, and pipeline parameters. In actual assessment, , , This represents the actual average flow rate of the node. The pressure difference between the inlet and outlet of the node; the installation location coordinates are within 50m of the node; the pipeline length is within the range of [5m, 50m]; the pipeline diameter adopts the general nominal diameter of the water plant; the water density is 1000kg / m³. 3 The acceleration due to gravity is 9.81 m / s². 2 ; Population size: 100, maximum number of iterations: 50, crossover probability: 0.8, mutation probability: 0.05; hydraulic simulation verification was performed using EPANET software, calculating the three objective function values ​​for each solution and filtering out individuals that violated hard constraints. Core deployment data for node 15: turbine rated flow rate 160, turbine rated head 31.5, installation location coordinates (118.XX, 24.XX), modified pipeline diameter 200, modified pipeline length 2m, power generation efficiency 0.892, investment cost 485,000 yuan, and pipeline safety impact 0.021. The total investment cost for the nine recommended nodes is estimated at 3.862 million yuan, all of which meet the safety impact threshold of the pipeline network.

[0044] In this embodiment, the method for identifying and determining the main parameters of the deployment data based on the hydropower data to obtain the main parameters includes: Latin hypercube sampling was used to generate 20 initial parameter combinations from water conservancy and hydropower data and deployment data. The error between the corresponding simulation results and the actual data was calculated using hydraulic simulation software to form the initial dataset. ;in Let a be the three-dimensional optimization parameter vector for group a. The root mean square error, For flow coefficient, This is the head correction factor. This refers to the pipe damping coefficient; A Gaussian process is used as a surrogate model to model the nonlinear relationship between parameters and error; its mean function is set as a constant function, and the covariance function is a radial basis function, expressed as:

[0045] in For signal variance, For length scale, For parameter vectors and covariance value, For parameter vectors and Euclidean distance; Minimizing the error between simulation results and actual data is taken as the objective function. An expected improvement function is used to balance exploration and development, expressed as:

[0046] in The mean value predicted by the surrogate model. The standard deviation of the surrogate model's prediction. This is the current minimum error. To explore the coefficient, Let be the probability density function of the standard normal distribution. The cumulative distribution function of the standard normal distribution. For the function to be improved; Based on the current dataset Train a Gaussian process model, predict the mean and standard deviation of each point in the parameter space, maximize the expected improvement function, select the next set of parameters to be evaluated, and perform hydraulic simulation calculations. Update the dataset Iterate continuously until the error is less than the error threshold; where Let be the objective function. For the first The parameter vector for the next iteration. For the first The root mean square error of the next iteration; When the absolute value of the error change over 5 consecutive iterations is less than or equal to the error change threshold, the iteration stops and the optimal parameters are output. In the actual assessment, the error threshold was 5%, and the error variation threshold was 0.001. Based on the water plant's water conservancy and hydropower data and layout data, 20 sets of initial parameter combinations were generated, of which 3 sets were: Group 1, flow coefficient 0.95, head correction coefficient 0.98, and pipeline damping coefficient 0.72; Group 2, flow coefficient 0.92, head correction coefficient 0.96, and pipeline damping coefficient 0.75; Group 3, flow coefficient 0.97, head correction coefficient 0.99, and pipeline damping coefficient 0.7. With a signal variance of 0.01, a length scale of 0.5, and an exploration coefficient of 0.05, the root mean square error (RMSE) of the simulation results and actual data for 20 sets of parameters was calculated using EPANET software to form an initial dataset, where the initial RMSE ranged from 8.2% to 15.6%. The optimal parameters are: flow coefficient 0.96, head correction coefficient 0.98, and pipe damping coefficient 0.74. The final mean square error is 4.2%, which meets the accuracy requirements of the calculation.

[0047] In this embodiment, the method for obtaining the feasibility level of the recommended development project list through economic and simulated engineering evaluation based on the main parameters includes: Based on the key parameters and the recommended development project list, obtain the design flow rate, residual pressure head, equipment efficiency, annual operating hours, initial investment, annual operating cost, project life, electricity price volatility, subsidy amount, and subsidy duration. Calculate the actual power generation:

[0048] in This represents the actual annual power generation. For design traffic, For the remaining pressure head, For overall equipment efficiency, Annual operating hours For the density of water, It is the acceleration due to gravity; Calculate annual cash flow and project net present value:

[0049]

[0050] in For the net cash flow in year s, The electricity price in year s, Annual operating costs To calculate the cumulative operating period, The year that the subsidy will be disbursed will be the last year. For the amount of subsidy, For initial investment, Let be the net present value of the project in year s. The benchmark discount rate; Geometric Brownian motion is used to simulate the stochastic process of electricity prices. Multiple future electricity price paths are generated through Monte Carlo simulation, and the electricity price distribution under each path is obtained, expressed as:

[0051] in Electricity price drift rate, For electricity price volatility, For the standard Wiener process in year s; Based on a binomial tree model, assuming managers can choose to invest immediately or delay investment for one year, with the delay cost being 5% of the initial investment; a binomial tree for electricity prices is constructed, including an increase factor. Decreasing factor and risk-neutral probability Calculate the value of the deferred option based on the deferred investment. ;in After the electricity price increase Values, After the drop in electricity prices Values, The optimal value of investing after a one-year delay; among which For time step; Assuming the project has negative cash flow in a given year, it can be abandoned and its residual value can be obtained. Calculate the annual cash flow under different electricity price paths, working backward from the end of the project's lifespan, to calculate the value of continuing operation versus abandonment: , Residual value To determine the optimal value of investing after a delay of s+1 years, To determine the optimal value of the investment after a delay of s years, the average value of multiple paths simulated is taken as the value of the forfeited option. ; The net present value of a project is adjusted based on the value of the delayed option and the value of the forfeited option, expressed as follows:

[0052] in To determine the feasibility level, a weighted scoring system is constructed based on the adjusted net present value (NPV) and risk indicators. The evaluation dimensions of the weighted scoring system include economic value, technological reliability, risk control, and policy stability. In the actual assessment, multiple paths were simulated, resulting in 1000 simulated paths; the weights for economic value, technological reliability, risk control, and policy stability were 0.4, 0.2, 0.2, and 0.2, respectively; among which, economic value... Technical reliability Risk control Policy stability ; The net present value threshold for the project. For operational reliability coefficient, This represents the probability of loss. For the probability of continued subsidies, This represents the percentage of paths in the simulation where the project's net present value is less than 0. Based on the optimal key parameters and the recommended development project list, unified evaluation parameters (average values ​​for each node) were extracted for 9 nodes. These parameters, combined with local policies and industry standards, determined the design flow rate to be 115 m³ / s. 3 / h, residual pressure head 26.8m, overall equipment efficiency 0.85, annual operating hours 8200h, initial investment 429,000 yuan, annual operating cost 85,000 yuan, project life 20 years, benchmark discount rate 8%, electricity price volatility 5%, subsidy amount 0.12 yuan / kW·h, subsidy duration 5 years, equipment residual value 43,000 yuan / node; The annual actual power generation of a single node is 592,000 yuan / kW·h, the total annual actual power generation of the nine nodes is 5,328,000 kWh, the traditional net present value of a single node is 685,000 yuan, and the total traditional net present value of the nine nodes is 6,165,000 yuan. With a time step of 1 year, an increase factor of 1.0513, a decrease factor of 0.9509, and a risk-neutral probability of 0.56, the calculated value of a single-node delay option is 52,000 yuan. Working backward from the end of the project's lifespan, the value of continuing operation versus abandonment under 1,000 electricity price paths is calculated, and the average value is 38,000 yuan for the abandonment option of a single node. The adjusted net present value is 6.975 million yuan, with economic value of 40 points, technical reliability of 19.6 points, risk control of 19.36 points, and policy stability of 19.8 points. The total feasibility score for a single node is 98.76 points, and the average feasibility score of the 9 recommended nodes is 96.2 points, all of which are excellent, indicating extremely high development feasibility.

[0053] In this embodiment, the method for obtaining development resource data by calculating resources based on the main parameters includes: The impact of uncertainties in key parameters on development resource data is quantified through Monte Carlo simulation, generating probability distribution results: the distribution type and parameters are determined based on historical data or industry standards, including inlet and outlet flow rates, residual pressure head, and equipment efficiency; the probability distribution interval of each parameter is divided into multiple sub-intervals, and one sample point is randomly selected in each sub-interval. The selected samples are then randomly combined to form multiple independent parameter combinations; development resource data for a single simulation is calculated based on the hydraulic power formula; the distribution curves of multiple annual power generation samples are fitted using kernel density estimation, key quantiles are calculated, and a 95% confidence interval is calculated using bootstrap sampling. In actual assessments, the data on developed resources include annual exploitable power generation, installed capacity, utilization efficiency of surplus resources, annual equivalent savings in standard coal, and annual reduction in carbon dioxide emissions. Monte Carlo simulation was used to quantify parameter uncertainties (inlet and outlet flow rates, residual head, and equipment efficiency). The distribution types of each parameter were set according to industry standards: Inlet and outlet flow rates: normal distribution N(115,5)m 3 / h; Residual pressure head: normal distribution N(26.8,1.2)m; Equipment efficiency: normal distribution N(0.85,0.02); 1000 sets of parameter combinations were randomly selected, and the development resource data were calculated based on the hydraulic power formula. The distribution curve was fitted by kernel density estimation, and the 95% confidence interval was calculated by bootstrap sampling. Finally, the total development resource data (mean, 95% confidence interval) for 9 nodes were obtained: annual developable power generation of 5.328 million kWh, 95% confidence interval (518.5~547.2); total installed capacity of 720 kW, 95% confidence interval (695~745); surplus pressure resource utilization efficiency of 86.5%, 95% confidence interval (84.2~88.8); annual equivalent standard coal saving of 170.5 t, 95% confidence interval (165.8~175.2); annual carbon dioxide emission reduction of 442.3 t, 95% confidence interval (430.5~454.1).

[0054] In this embodiment, the method for constructing a calculation model for the exploitable amount of hydropower generation resources based on the feasibility level and the development resource data includes: Based on the feasibility level and development resource data, obtain spatiotemporal feature vectors, deployment parameters, and feasibility scores; use one-hot encoding for the classification features in the deployment parameters, and perform min-max normalization to the [0,1] interval for the numerical features; and weight and fuse the economic and engineering feasibility scores into a comprehensive feasibility vector. The model for calculating the exploitable amount of hydropower generation resources from waste pressure includes an input layer, a Transformer encoder, a reinforcement learning optimization module, and an output layer. Input layer: Based on the location encoding, temporal and location information is added to the spatiotemporal features. The standardized spatiotemporal features are then added to the location embedding to obtain the spatiotemporal embedding vector. The expression is:

[0055]

[0056] in For time steps, Indexed by feature dimensions, For embedded dimensions, Encode the time location vector; Spatiotemporal embedding vectors Layout parameter embedding vector Feasibility scoring embedding vector Concatenate the sequences according to the time dimension to form the input sequence X; Transformer encoder: Takes the input sequence X and generates query, key, and value matrices through a linear transformation, respectively. , , ,in To query learnable parameters, For key learnable parameters, For learnable parameters, For query, As key, The output is the value; each attention head is output, the outputs of multiple attention heads are concatenated and subjected to a linear transformation to obtain the multi-head self-attention output. Each encoder layer consists of a first layer and a second layer, and outputs a context feature vector. The first layer consists of multi-head self-attention, residual connections, and layer normalization; the second layer consists of a feedforward neural network, residual connections, and layer normalization. The expression is:

[0057]

[0058]

[0059] in It is a feedforward neural network. These are intermediate features following the multi-head self-attention module. The final output features of a single encoder layer, For layer normalization, The weight matrix is ​​the first-level linear transformation. This is the bias vector for the first-level linear transformation. This is the weight matrix for the second-level linear transformation. This is the bias vector for the second-level linear transformation; Reinforcement Learning Optimization Module: A proximal policy-based optimization reinforcement learning module is introduced to dynamically adjust the encoder. The encoder's state consists of the current context feature vector and historical prediction error, while the action is the attention weight adjustment coefficient. The reward function aims to maximize prediction accuracy. The policy network is updated by optimizing the pruning objective function, and the optimized context feature vector is output. ; Output layer: The vector is compressed into a fixed-dimensional vector by global average pooling, and then the predicted value is output through a fully connected network. The 95% confidence interval is calculated by introducing Gaussian distribution parameterization into the output layer of the Bayesian neural network. In practical evaluation, the graph convolutional network adopts a 2-layer GCN structure. The input features are node pressure, flow rate, and number of connecting pipes (3D), and the output is a 64-dimensional node embedding vector. The output embedding vector is then L2 normalized to obtain the normalized node feature vector. The mean of the normalized embedding vectors is used as the initial importance of the nodes. ; Input layer: Spatiotemporal feature vector (24-dimensional), deployment parameter vector (12-dimensional), feasibility score vector (4-dimensional), one-hot encoding of classification features, and min-max normalization of numerical features to [0,1]; Transformer encoder: 6-layer encoder, 8 attention heads, embedding dimension d_model=128, feedforward neural network hidden layer dimension 512; PPO reinforcement learning module: state is context feature vector + historical prediction error, action is attention weight adjustment coefficient, reward function is to maximize prediction accuracy; Output layer: Bayesian neural network with Gaussian distribution parameterization, calculating 95% confidence interval; Using the water plant's historical data from the past three years as the training set (70%), validation set (20%), and test set (10%), the model was trained until the loss function converged (loss value < 0.01), and the prediction accuracy on the test set reached 94.8%. Input the data to be calculated (deployment data, feasibility level, and parameter data of the 9 recommended nodes) into the calculation model, and output the final exploitable amount of the water plant's residual pressure power generation resources (95% confidence interval): Recommended total installed capacity: 715kW, 95% confidence interval (690~740); Annual exploitable power generation: 5.302 million kWh, 95% confidence interval (515.8~544.6); Total revenue over the entire life cycle of the project: 28.65 million yuan, 95% confidence interval (2780~2950); Total return on investment over the entire life cycle: 742%, 95% confidence interval (720~765).

[0060] Secondly, a system for calculating the exploitable resources for hydropower generation from residual pressure in water plants includes: Data acquisition and processing module: used to collect water conservancy and hydropower data and resource data of the preset water plant, and to preprocess the water conservancy and hydropower data and resource data; Project identification and screening module: used to identify hydropower resource development from the resource data to obtain a list of recommended development projects, and to perform multi-objective optimization deployment of the recommended development project list to obtain deployment data; Parameter identification and feasibility assessment module: used to identify and determine the main parameters of the layout data based on the water conservancy and hydropower data, to conduct economic and simulated engineering assessments of the recommended development project list based on the main parameters to obtain the degree of feasibility, and to calculate development resource data based on the main parameters. Model building and output module: This module is used to build a calculation model for the exploitable amount of water plant residual pressure power generation resources based on the feasibility level and the development resource data. It inputs the data to be calculated into the calculation model for the exploitable amount of water plant residual pressure power generation resources and outputs the calculation results.

[0061] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for calculating the exploitable amount of hydropower generation resources from residual pressure in a water plant, characterized in that, Includes the following steps: Collect water conservancy and hydropower data and resource data from a pre-designated water plant, and preprocess the water conservancy and hydropower data and the resource data; the resource data includes geospatial data, hydrological and meteorological data, infrastructure data and socio-economic data; The resource data is used to identify hydropower resource development to obtain a list of recommended development projects, and the list of recommended development projects is then used for multi-objective optimization to obtain deployment data. Based on the water conservancy and hydropower data, the main parameters of the deployment data are identified and determined to obtain the main parameters. Based on the main parameters, the recommended development project list is evaluated economically and through simulated engineering to obtain the feasibility. Based on the main parameters, resource calculations are performed to obtain development resource data. Based on the feasibility level and the development resource data, a calculation model for the exploitable amount of water plant residual pressure power generation resources is constructed. The data to be calculated is input into the calculation model for the exploitable amount of water plant residual pressure power generation resources, and the calculation results are output.

2. The method for calculating the exploitable amount of hydropower generation resources from residual pressure in a water plant according to claim 1, characterized in that, A method for identifying recommended development projects from the aforementioned resource data, including: By acquiring pipeline connection relationships, pipe diameter, material, length, valve / pressure reducing valve location, pressure, flow rate, historical records of annual operating hours, water density, and gravitational acceleration at each node through resource data, key locations in the pipeline network are defined as graph nodes; key locations include pressure gauge installation points, pressure reducing valve assemblies, water pump outlets, and high and low zone pipeline connection points; Treating the pipe as an edge in the graph, with the edge weight being the pressure loss coefficient, the expression is: ； in This is the friction coefficient. Let be the length of the pipe from node i to node j. Let be the pipe diameter between the i-th node and the j-th node. Let be the average flow between the i-th node and the j-th node. Let be the edge weight between the i-th node and the j-th node; Node features are extracted using a graph convolutional network, including node pressure, flow rate, and number of connected pipes, to generate node embedding vectors. These vectors are then combined with PageRank to calculate node importance scores. ； in This is the pipe damping coefficient. Let the initial importance of the i-th node be... The importance score is given to the i-th node. For the i-th node, For the j-th node, For nodes The set of all direct neighbor nodes, For nodes and nodes Edge weights; The nodes are sorted in descending order of importance scores to obtain a preliminary list of node priorities. For the top 30% of nodes, their inlet and outlet pressure differences are calculated. Nodes with an inlet and outlet pressure difference greater than 0.2 MPa and an annual operating time greater than 6000 hours are selected as candidate resource points. The comprehensive potential of nodes is calculated based on importance score, annual operating hours, and import / export pressure difference. Nodes are then ranked according to their comprehensive potential using a comprehensive index, expressed as follows: ； in For nodes The comprehensive potential For normalized nodes Annual operating hours, For normalized nodes The difference between import and export pressures The importance score after normalization; Remove duplicate points on the same branch of the pipeline that are less than 500m apart, and output the list of candidate resource points after removal as a recommended development project list according to comprehensive indicators.

3. The method for calculating the exploitable amount of hydropower resources generated from residual pressure in a water plant according to claim 1, characterized in that, A method for obtaining deployment data by performing multi-objective optimization deployment of the recommended development project list includes: Based on the optimization objectives of maximizing power generation efficiency, minimizing investment costs, and minimizing the impact on pipeline safety, an objective function is constructed, expressed as follows: ； ； ； in For real-time traffic, For real-time residual pressure head, For the efficiency of the water turbine, Annual operating hours For the density of water, It is the acceleration due to gravity. For the length of the pipe, For pipe diameter, For power generation efficiency, For investment costs, To ensure the safety of the pipeline network, The pressure value of the node before deployment. This represents the pressure value at the nodes after deployment. For the rated head, For rated flow rate, For installation costs, For pipeline renovation costs, For equipment procurement costs, The number of nodes; Constraints include engineering constraints, equipment constraints, and hydraulic constraints. Engineering constraints include: the installation location must meet the pipeline's load-bearing requirements; equipment constraints include: the actual flow rate of the turbine. Hydraulic constraints: Pipeline flow velocity less than or equal to 2.5 m / s; The third-generation non-dominated sorting genetic algorithm is used to design and distribute the data according to the objective function: multiple initial solutions are randomly generated. ,in The coordinates are the installation location coordinates; hydraulic simulation is performed to verify each solution, and calculations are performed. , , Values ​​are used to filter individuals that violate hard constraints; The population is sorted in a non-dominated manner, divided into different Pareto levels, the vertical distance between each individual and the dynamic reference point is calculated, high-quality individuals are selected and retained through a tournament, and adaptive crossover and mutation are performed on the selected individuals to generate the offspring population. The parent and offspring populations are merged, and the non-dominated sorting is re-established. For each reference point, the closest individual is retained, and a new generation of population is formed. This process is repeated until the maximum number of iterations is reached, and the final Pareto optimal solution set is output. Each solution corresponds to a complete set of deployment data, which includes equipment signals, number of installation locations, and pipeline parameters.

4. The method for calculating the exploitable amount of hydropower generation resources based on residual pressure in a water plant according to claim 1, characterized in that, A method for identifying and determining key parameters of the deployment data based on the aforementioned hydropower data to obtain key parameters includes: Latin hypercube sampling was used to generate 20 initial parameter combinations from water conservancy and hydropower data and deployment data. The error between the corresponding simulation results and the actual data was calculated using hydraulic simulation software to form the initial dataset. ;in Let a be the three-dimensional optimization parameter vector for group a. The root mean square error, For flow coefficient, This is the head correction factor. This refers to the pipe damping coefficient; A Gaussian process is used as a surrogate model to model the nonlinear relationship between parameters and error; its mean function is set as a constant function, and the covariance function is a radial basis function, expressed as: ； in For signal variance, For length scale, For parameter vectors and covariance value, For parameter vectors and Euclidean distance; Minimizing the error between simulation results and actual data is taken as the objective function. An expected improvement function is used to balance exploration and development, expressed as: ； in The mean value predicted by the surrogate model. The standard deviation of the surrogate model's prediction. This is the current minimum error. To explore the coefficient, Let be the probability density function of the standard normal distribution. The cumulative distribution function of the standard normal distribution. For the function to be improved; Based on the current dataset Train a Gaussian process model, predict the mean and standard deviation of each point in the parameter space, maximize the expected improvement function, select the next set of parameters to be evaluated, and perform hydraulic simulation calculations. Update the dataset Iterate continuously until the error is less than the error threshold; where Let be the objective function. For the first The parameter vector for the next iteration. For the first The root mean square error of the next iteration; When the absolute value of the error change over 5 consecutive iterations is less than or equal to the error change threshold, the iteration stops and the optimal parameters are output.

5. The method for calculating the exploitable amount of hydropower generation resources based on residual pressure in a water plant according to claim 1, characterized in that, A method for obtaining the feasibility of the recommended development project list by conducting economic and simulated engineering assessments based on the aforementioned key parameters includes: Based on the key parameters and the recommended development project list, obtain the design flow rate, residual pressure head, equipment efficiency, annual operating hours, initial investment, annual operating cost, project life, electricity price volatility, subsidy amount, and subsidy duration. Calculate the actual power generation: ； in This represents the actual annual power generation. For design traffic, For the remaining pressure head, For overall equipment efficiency, Annual operating hours For the density of water, It is the acceleration due to gravity; Calculate annual cash flow and project net present value: ； ； in For the net cash flow in year s, The electricity price in year s, Annual operating costs To calculate the cumulative operating period, The year that the subsidy will be disbursed will be the last year. For the amount of subsidy, For initial investment, Let be the net present value of the project in year s. The benchmark discount rate; Geometric Brownian motion is used to simulate the stochastic process of electricity prices. Multiple future electricity price paths are generated through Monte Carlo simulation, and the electricity price distribution under each path is obtained, expressed as: ； in Electricity price drift rate, For electricity price volatility, For the standard Wiener process in year s; Based on a binomial tree model, assuming managers can choose to invest immediately or delay investment for one year, with the delay cost being 5% of the initial investment; a binomial tree for electricity prices is constructed, including an increase factor. Decreasing factor and risk-neutral probability Calculate the value of the deferred option based on the deferred investment. ;in After the electricity price increase Values, After the drop in electricity prices Values, The optimal value of investing after a one-year delay; among which For time step; Assuming the project has negative cash flow in a given year, it can be abandoned and its residual value can be obtained. Calculate the annual cash flow under different electricity price paths, working backward from the end of the project's lifespan, to calculate the value of continuing operation versus abandonment: , Residual value To determine the optimal value of investing after a delay of s+1 years, To determine the optimal value of the investment after a delay of s years, the average value of multiple paths simulated is taken as the value of the forfeited option. ; The net present value of a project is adjusted based on the value of the delayed option and the value of the forfeited option, expressed as follows: ； in To determine the feasibility level, a weighted scoring system is constructed based on the adjusted net present value (NPV) and risk indicators. The evaluation dimensions of the weighted scoring system include economic value, technological reliability, risk control, and policy stability.

6. The method for calculating the exploitable amount of hydropower generation resources based on residual pressure in a water plant according to claim 1, characterized in that, A method for constructing a calculation model for the exploitable amount of hydropower generation resources based on the aforementioned feasibility level and the aforementioned development resource data includes: Based on the feasibility level and development resource data, obtain spatiotemporal feature vectors, deployment parameters, and feasibility scores; use one-hot encoding for the classification features in the deployment parameters, and perform min-max normalization to the [0,1] interval for the numerical features; and weight and fuse the economic and engineering feasibility scores into a comprehensive feasibility vector. The model for calculating the exploitable amount of hydropower generation resources from waste pressure includes an input layer, a Transformer encoder, a reinforcement learning optimization module, and an output layer. Input layer: Based on the location encoding, temporal and location information is added to the spatiotemporal features. The standardized spatiotemporal features are then added to the location embedding to obtain the spatiotemporal embedding vector. The expression is: ； ； in For time steps, Indexed by feature dimensions, For embedded dimensions, Encode the time location vector; Spatiotemporal embedding vectors Layout parameter embedding vector Feasibility scoring embedding vector Concatenate the sequences according to the time dimension to form the input sequence X; Transformer encoder: Takes the input sequence X and generates query, key, and value matrices through a linear transformation, respectively. , , ,in To query learnable parameters, For key learnable parameters, For learnable parameters, For query, As key, The output is the value; each attention head is output, the outputs of multiple attention heads are concatenated and subjected to a linear transformation to obtain the multi-head self-attention output. Each encoder layer consists of a first layer and a second layer, and outputs a context feature vector. The first layer consists of multi-head self-attention, residual connections, and layer normalization; the second layer consists of a feedforward neural network, residual connections, and layer normalization. The expression is: ； ； ； in It is a feedforward neural network. These are intermediate features following the multi-head self-attention module. The final output features of a single encoder layer, For layer normalization, The weight matrix is ​​the first-level linear transformation. This is the bias vector for the first-level linear transformation. This is the weight matrix for the second-level linear transformation. This is the bias vector for the second-level linear transformation; Reinforcement Learning Optimization Module: A proximal policy-based optimization reinforcement learning module is introduced to dynamically adjust the encoder. The encoder's state consists of the current context feature vector and historical prediction error, while the action is the attention weight adjustment coefficient. The reward function aims to maximize prediction accuracy. The policy network is updated by optimizing the pruning objective function, and the optimized context feature vector is output. ; Output layer: The vector is compressed into a fixed-dimensional vector using global average pooling, and then the predicted value is output through a fully connected network. The 95% confidence interval is calculated by introducing a Gaussian distribution parameterization into the output layer of the Bayesian neural network.

7. A system for calculating the exploitable resources for hydropower generation from residual pressure in a water plant, used to execute the method described in any one of claims 1-6, characterized in that, include: Data acquisition and processing module: used to collect water conservancy and hydropower data and resource data of the preset water plant, and to preprocess the water conservancy and hydropower data and resource data; Project identification and screening module: used to identify hydropower resource development from the resource data to obtain a list of recommended development projects, and to perform multi-objective optimization deployment of the recommended development project list to obtain deployment data; Parameter identification and feasibility assessment module: used to identify and determine the main parameters of the layout data based on the water conservancy and hydropower data, to conduct economic and simulated engineering assessments of the recommended development project list based on the main parameters to obtain the degree of feasibility, and to calculate development resource data based on the main parameters. Model building and output module: This module is used to build a calculation model for the exploitable amount of water plant residual pressure power generation resources based on the feasibility level and the development resource data. It inputs the data to be calculated into the calculation model for the exploitable amount of water plant residual pressure power generation resources and outputs the calculation results.

Images