Site selection method for underground gas storage chamber of compressed air energy storage system

Abstract

The invention provides an underground gas storage chamber site selection method for a compressed air energy storage system, and the method comprises the following steps: integrating multi-source data of a target region, building a digital twinborn body of the target region, and building a multi-field coupling model based on the digital twinborn body; a Pareto optimal library address set is generated through rapid non-dominated sorting and congestion degree calculation by adopting multi-target particle swarm optimization and taking carbon sink cooperation coefficient maximization, construction operation cost minimization and geological risk minimization as targets; outputting a dynamic risk threshold value corresponding to the target area through a deep belief network, quantifying a risk probability under parameter uncertainty based on a Bayesian network, and rejecting library addresses with risks exceeding the standard in a library address set according to the risk probability to obtain an optimized library address set; the method comprises the following steps: constructing a self-adaptive correction model based on a DQN network, automatically correcting simulation parameters of a digital twinborn body, constructing a comprehensive energy efficiency evaluation index system based on the digital twinborn body after parameter correction, calculating a comprehensive energy efficiency value, and determining an optimal site selection result.

Description

Technical Field

[0001] This invention belongs to the field of compressed air energy storage technology, and particularly relates to a method for selecting the site of underground gas storage chambers for compressed air energy storage systems. Background Technology

[0002] Among related technologies, compressed air energy storage (CAES) is a large-scale, long-term energy storage technology. The location of its underground gas storage chamber is crucial, as it directly affects the power plant's safety, operating efficiency, service life, and environmental impact.

[0003] However, existing site selection technologies have significant limitations. In terms of model building, the parameters of digital twins used to assist decision-making largely rely on human experience, leading to insufficient model accuracy and significant deviations from reality. Furthermore, regarding decision-making criteria, the evaluation index systems used often only cover geological conditions and ground infrastructure, failing to comprehensively reflect the overall effectiveness and long-term impact of the energy storage system. Traditional evaluation systems are incomplete and their processing methods are crude, resulting in insufficient adaptability and robustness of the final site selection results in complex and ever-changing real-world engineering scenarios, making it difficult to meet the high standards required for site selection in compressed air energy storage power plants. Summary of the Invention

[0004] The purpose of this invention is to provide a method for selecting the location of underground gas storage chambers for compressed air energy storage systems, aiming to solve the technical problems mentioned in the background art.

[0005] To achieve the above objectives, the embodiments of the present invention provide the following technical solutions.

[0006] According to an embodiment of the present invention, a method for site selection of underground gas storage chambers for compressed air energy storage systems is provided, comprising the following steps: Integrate multi-source data of the target area and establish a digital twin of the target area. Based on the digital twin, construct a multi-field coupling model. In the coupling model, use a hybrid temporal prediction network to output the parameter prediction sequence of the coupling model. Combine Sobol global sensitivity analysis to screen the parameters of the coupling model. A multi-objective particle swarm optimization approach is adopted, with the objectives of maximizing the carbon sink synergy coefficient, minimizing construction and operation costs, and minimizing geological risks. Pareto optimal site set is generated through fast non-dominated sorting and congestion calculation. The system outputs a dynamic risk threshold corresponding to the target area through a deep belief network, and quantifies the risk probability under parameter uncertainty based on a Bayesian network. Based on the risk probability, the system removes the storage sites in the storage site set that exceed the risk limit, thereby obtaining an optimized storage site set. An adaptive correction model based on the DQN network is constructed to automatically correct the simulation parameters of the digital twin. Based on the parameter-corrected digital twin, a comprehensive energy efficiency evaluation index system is constructed. The weights of each index in the system are calculated by weight allocation. Combined with the measured data of each site in the optimized site set and the simulation parameters of the corrected twin, the comprehensive energy efficiency value is calculated and the optimal site selection result is determined.

[0007] Furthermore, the multi-source data includes geological data, ecological data, and carbon sink data; The steps for establishing a digital twin of the target region include: Extract GIS data of macroscopic areas and BIM data of microscopic components from multi-source data; Based on GIS data, construct a macroscopic model covering the target area; Based on BIM data, a parametric method was used to construct the chamber model of the candidate gas storage chamber; The chamber model is embedded into the corresponding spatial position of the macro model in a geometrically aligned manner to form a fused model. The fused model is then modified to obtain a digital twin of the target area.

[0008] Furthermore, the step of constructing a multi-field coupling model based on a digital twin includes: Geological parameters, chamber structure parameters, ecological carbon sink parameters, and real-time monitoring parameters are extracted from the digital twin. Establish the mapping relationship between each parameter and the target physical field, wherein the target physical field includes at least the gas storage pressure field, the geological stress field, the groundwater seepage field, and the ecological carbon sink field; Based on the parameters mapped to each physical field, corresponding physical field sub-models are constructed to obtain a multi-field coupling model.

[0009] Furthermore, the steps of using a hybrid temporal prediction network to output the parameter prediction sequence of the coupled model in the coupled model, combined with Sobol global sensitivity analysis to screen the parameters of the coupled model, include: A hybrid neural network containing a Long Short-Term Memory (LSTM) layer and a Gated Recurrent Unit (GRU) layer is constructed. Historical monitoring data and future scene parameters from the digital twin are used as input features. The hybrid neural network is trained using the mean squared error loss function and the Adam optimizer to output the future time series of key parameters of the coupled model. The mean squared error loss function is expressed as: The parameter update of the Adam optimizer is represented as: In the formula, N represents the sample size. This represents the predicted value of the i-th sample. This represents the true value of the i-th sample. Denotes the network parameters in the t-th iteration. express Updated network parameters Indicates the learning rate. , Represents the momentum parameter. Let t represent the first-order momentum of the t-th cycle. Let t represent the second momentum of the t-th round. This indicates the prevention of the minimum value where the denominator is 0; Based on the digital twin, the variation range of each parameter in the coupled model is determined to construct a parameter space. A sampling method is used to generate a sample set within the parameter space. The overall order sensitivity index of each parameter to the output of the coupled model is expressed as follows: In the formula, V(Y) represents the total order sensitivity index of the i-th parameter, and V(Y) represents the total variance of the coupled model output. This represents the conditional variance of the model output after removing the i-th parameter; Based on the calculated total order sensitivity index, key parameters that significantly affect the output of the coupled model are selected.

[0010] Furthermore, the step of generating the Pareto optimal library address set includes: Establish a multi-objective optimization function that includes at least maximizing the carbon sink synergy coefficient, minimizing construction and operation costs, and minimizing geological risks; Initialize the particle swarm based on candidate library addresses that satisfy basic constraints, using digital twins as a basis; During the iteration process, the particle velocity and position are updated using dynamically adjusted inertia weights and learning factors. In the formula, This represents the velocity of the i-th particle in the t-th round. This represents the position of the i-th particle in the t-th round. This represents the inertia weight in round t; , Represents the learning factor. , Represents a random number. Let represent the optimal position of the i-th particle. The global optimal position of the particle swarm is represented by T, and the total number of iterations is represented by T. A fast sorting mechanism that includes non-dominated sorting and objective function contribution evaluation is used to screen particles; Calculate the crowding degree of particles in the target space and weight the crowding degree of particles located in the critical site selection region; after the iteration is completed, select non-dominated solutions with high crowding degree values ​​to form the Pareto optimal site set.

[0011] Furthermore, the step of outputting a dynamic risk threshold corresponding to the target area through a deep belief network includes: Multi-dimensional features related to risk are selected from the target area to form an input feature set; the multi-dimensional features include at least regional geological features, climate and hydrological features, energy storage operation features and ecological constraint features. A deep belief network model is constructed and trained using historical site selection case data. During the pre-training phase, the reconstruction error is minimized using contrastive divergence, as shown below: In the formula, This represents the original value of the i-th node in the input layer. This represents the reconstruction value at the i-th stage of the input layer, where n represents the number of nodes in the input layer. The training objective is to establish a mapping relationship from the input feature set to various risk thresholds; The input feature set of the target region is input into the trained deep belief network model, which outputs a basic dynamic risk threshold. Based on the specific scene conditions of the target region, the basic dynamic risk threshold is corrected to obtain the final adapted dynamic risk threshold.

[0012] Furthermore, the steps for quantifying risk probabilities under parameter uncertainty based on Bayesian networks include: Construct a multi-level Bayesian network comprising an input layer, an intermediate layer, and an output layer; wherein the input layer nodes represent key parameters with uncertainty, the intermediate layer nodes represent the response of a multi-field coupled model, and the output layer nodes represent the risk to be evaluated; Based on multi-field coupling mechanisms and historical case data, the dependencies between nodes are determined to construct a directed acyclic graph. Maximum likelihood estimation is then used to construct a data-driven conditional probability table for each node, as follows: In the formula, This represents the conditional probability of a node combination. This indicates the number of historical observations for this node combination; Based on the parameter probability distribution of the input layer nodes, the probability of each risk node occurring in the output layer, as well as the joint probability of multiple risks occurring together, are calculated using a probabilistic inference algorithm, and expressed as: In the formula, Indicates risk and The probability of their joint occurrence, , As an intermediate response node, This represents the uncertainty parameters of the input layer. Indicates parameters The prior probability distribution.

[0013] Further steps for automatically correcting the simulation parameters of the digital twin include: The simulation parameters of the digital twin are used as the action space, the deviation between the simulation values ​​of the coupled model and the measured values ​​of the optimized address set are used as the state space, and the minimization of the deviation is used as the reward function. Construct a deep Q-network, taking the state space as input and outputting the Q-values ​​of each action in the action space as the objective; train the network using historical and real-time data, updating the Q-values ​​using temporal difference learning during training, as shown below: In the formula, This represents the Q value of performing action a in state s. Let r represent the learning rate and r represent the immediate reward. Indicates the discount factor. This indicates the next state after action a is performed. For state The optimal action under the given conditions; The current state is input into the trained DQN network, the action with the highest Q value is selected to correct the simulation parameters of the digital twin, and the state is updated based on the corrected simulation results. This process is repeated until the simulation deviation meets the preset requirements.

[0014] Furthermore, based on the parameter-corrected digital twin, a comprehensive energy efficiency evaluation index system is constructed. The steps include calculating the weights of each index in the system through weight allocation, combining measured data from each site in the optimized site set with the simulated parameters of the corrected digital twin, calculating the comprehensive energy efficiency value, and determining the optimal site selection result. Establish a hierarchical evaluation index system with multiple dimensions, including at least geological safety, energy storage operation, ecological carbon sink and full-cycle economic efficiency, with each dimension containing several quantifiable secondary evaluation indicators; For each site in the optimized site set, based on measured data and the simulation parameters of the corrected digital twin, the quantified values ​​of each secondary indicator are calculated; after normalizing the quantified values, and combining them with the combined weights, the comprehensive energy efficiency value of each site is calculated, expressed as: In the formula, This represents the overall energy efficiency value of the i-th storage site. Indicates the scene adaptation coefficient. This represents the normalized value of the j-th index for the i-th database address; Based on the comprehensive energy efficiency value, the final preferred storage site is determined from the optimized storage site set.

[0015] Furthermore, the combined weights are calculated using a combination of subjective and objective weighting methods to determine the combined weights of each secondary indicator in the indicator system, expressed as follows: In the formula, This represents the combined weight of the j-th secondary indicator. Represents the balance coefficient. Indicates subjective weighting. Indicates objective weighting. Let m represent the information entropy of the j-th indicator, and m represent the number of optimized database addresses. This represents the percentage of the normalized value of the j-th index for the i-th database address.

[0016] Compared with existing technologies, the advantages of the underground gas storage chamber site selection method for the compressed air energy storage system of the present invention are as follows: This invention constructs a digital twin and a multi-field coupling model by integrating multi-source data. Based on the digital twin and the multi-field coupling model, a hybrid time-series prediction network is introduced to predict future operating parameters. Combined with Sobol global sensitivity analysis, key parameters are accurately screened. An adaptive correction model based on DQN is used to optimize the simulation parameters, thereby realizing parameter correction of the digital twin. In terms of site search and screening, this invention employs an improved multi-objective particle swarm optimization, aiming to maximize the carbon sink synergy coefficient, minimize construction and operation costs, and minimize geological risks. Through rapid non-dominated sorting and crowding calculation, it automatically generates a uniformly distributed Pareto optimal site set. Then, it outputs a dynamic risk threshold bound to regional characteristics through a deep belief network and uses a Bayesian network to accurately quantify the risk probability under parameter uncertainty, eliminating sites with excessive risk and obtaining a safe and controllable optimized site set. Based on a digital twin with dynamically corrected parameters, this invention constructs a comprehensive energy efficiency evaluation index system that includes geological safety, energy storage operation, ecological carbon sequestration, and full-cycle economic efficiency. This system allocates weights through a combination of subjective and objective methods, and calculates the comprehensive energy efficiency value of each site by combining measured data of the optimized site set with the twin simulation parameters, ultimately obtaining the optimal site selection scheme. Attached Figure Description

[0017] The accompanying drawings, which form part of this application, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0018] In the attached diagram: Figure 1 This is a flowchart illustrating the implementation of the underground gas storage chamber site selection method for the compressed air energy storage system of the present invention. Figure 2 This is a sub-flowchart of the underground gas storage chamber site selection method for the compressed air energy storage system of the present invention; Figure 3 This is another sub-flowchart of the method for selecting the location of underground gas storage chambers in the compressed air energy storage system of the present invention; Figure 4 This is a hardware structure block diagram of a computer terminal provided in another embodiment of the present invention. Detailed Implementation

[0019] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0020] According to an embodiment of this application, a method embodiment for selecting the location of an underground gas storage chamber for a compressed air energy storage system is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0021] Please refer to Figure 1 According to an embodiment of the present invention, a method for selecting the location of an underground gas storage chamber for a compressed air energy storage system is provided, comprising the following steps: Step S101: Integrate multi-source data of the target area and establish a digital twin of the target area. Based on the digital twin, construct a multi-field coupling model. In the coupling model, use a hybrid temporal prediction network to output the parameter prediction sequence of the coupling model. Combine Sobol global sensitivity analysis to screen the parameters of the coupling model. In step S101 of the present invention, the multi-source data includes geological data, ecological data and carbon sink data; Specifically, in this embodiment of the invention, the step of establishing a digital twin of the target area includes extracting GIS data of the macro-area and BIM data of the micro-components from multi-source data; Based on GIS data, construct a macroscopic model covering the target area; Based on BIM data, a parametric method was used to construct the chamber model of the candidate gas storage chamber; The chamber model is embedded into the corresponding spatial position of the macro model in a geometrically aligned manner to form a fused model. The fused model is then modified to obtain a digital twin of the target area.

[0022] Furthermore, such as Figure 2 As shown, the steps for constructing a multi-field coupling model based on a digital twin include: Step S201: Extract geological parameters, chamber structure parameters, ecological carbon sink parameters, and real-time monitoring parameters from the digital twin; Step S202: Establish the mapping relationship between each parameter and the target physical field, wherein the target physical field includes at least the gas storage pressure field, the geological stress field, the groundwater seepage field, and the ecological carbon sink field; Step S203: Based on the parameters mapped to each physical field, construct the corresponding physical field sub-models to obtain the multi-field coupling model.

[0023] Furthermore, step S203 of this embodiment of the invention further includes: determining the coupling relationship and coupling coefficient between each physical field, and constructing a coupling iterative algorithm based on this, wherein the iterative algorithm includes: The output of the previous physical field sub-model at a specific time is used as the boundary condition of the next physical field sub-model at that time for iterative calculation, thereby realizing the dynamic collaborative simulation of multiple physics fields. The output results of the multi-field collaborative simulation are compared with the real-time monitoring data obtained from the digital twin; Based on the comparison error, the corresponding parameters extracted by the digital twin are adjusted until the error between the simulation output result and the real-time monitoring data meets the preset tolerance, thus completing the iterative optimization of the multi-field coupling model.

[0024] Furthermore, in step S101 of this embodiment of the invention, the step of using a hybrid temporal prediction network to output the parameter prediction sequence of the coupled model in the coupled model, and combining it with Sobol global sensitivity analysis to filter the parameters of the coupled model, includes: A hybrid neural network containing a Long Short-Term Memory (LSTM) layer and a Gated Recurrent Unit (GRU) layer is constructed. Historical monitoring data and future scene parameters from the digital twin are used as input features. The hybrid neural network is trained using the mean squared error loss function and the Adam optimizer to output the future time series of key parameters of the coupled model. The mean squared error loss function is expressed as: The parameter update of the Adam optimizer is represented as: In the formula, N represents the sample size. This represents the predicted value of the i-th sample. This represents the true value of the i-th sample. Denotes the network parameters in the t-th iteration. express Updated network parameters Indicates the learning rate. , Represents the momentum parameter. Let t represent the first-order momentum of the t-th cycle. Let t represent the second momentum of the t-th round. This indicates the prevention of the minimum value where the denominator is 0; Based on the digital twin, the variation range of each parameter in the coupled model is determined to construct a parameter space. A sampling method is used to generate a sample set within the parameter space. The overall order sensitivity index of each parameter to the output of the coupled model is expressed as follows: In the formula, V(Y) represents the total order sensitivity index of the i-th parameter, and V(Y) represents the total variance of the coupled model output. This represents the conditional variance of the model output after removing the i-th parameter; Based on the calculated total order sensitivity index, key parameters that significantly affect the output of the coupled model are selected.

[0025] Historical monitoring data includes rock mass stress, groundwater seepage flow, and carbon sink density. Future scenario parameters include the daily power output of wind and solar power plants and the annual rainfall in the region; The key parameters of the coupled model include at least one of the following: rock mass elastic modulus, permeability coefficient, carbon sink density, and chamber sealing failure probability.

[0026] As can be seen, this invention constructs a digital twin and a multi-field coupling model by integrating multi-source data, and on the basis of the digital twin and the multi-field coupling model, introduces a hybrid time series prediction network to predict future operating parameters, combines Sobol global sensitivity analysis to accurately screen key parameters, and uses an adaptive correction model based on DQN to optimize the simulation parameters, thereby realizing parameter correction of the digital twin. Please continue to refer to Figure 1 The method for selecting the location of underground gas storage chambers for compressed air energy storage systems according to embodiments of the present invention further includes: Step S102: Using multi-objective particle swarm optimization, with the objectives of maximizing the carbon sink synergy coefficient, minimizing construction and operation costs, and minimizing geological risks, a Pareto optimal site set is generated through fast non-dominated sorting and congestion calculation. Specifically, step S102, the step of generating the Pareto optimal library address set, includes: Establish a multi-objective optimization function that includes at least maximizing the carbon sink synergy coefficient, minimizing construction and operation costs, and minimizing geological risks; wherein: The carbon sink synergy coefficient is calculated based on the net increase in carbon sink output by the digital twin and the carbon emissions from construction and operation. The construction and operation costs are calculated based on maintenance costs predicted by BIM component data and multi-field coupling model. The geological risk is calculated by weighting the rock mass stability coefficient, sealing failure probability, and groundwater influence coefficient output by the multi-field coupling model. Initialize the particle swarm based on candidate library addresses that satisfy basic constraints, using digital twins as a basis; During the iteration process, the particle velocity and position are updated using dynamically adjusted inertia weights and learning factors. In the formula, This represents the velocity of the i-th particle in the t-th round. This represents the position of the i-th particle in the t-th round. This represents the inertia weight in round t; , Represents the learning factor. , Represents a random number. Let represent the optimal position of the i-th particle. The global optimal position of the particle swarm is represented by T, and the total number of iterations is represented by T. A fast sorting mechanism that includes non-dominated sorting and objective function contribution evaluation is used to screen particles; Calculate the crowding degree of particles in the target space and weight the crowding degree of particles located in the critical site selection region; after the iteration is completed, select non-dominated solutions with high crowding degree values ​​to form the Pareto optimal site set.

[0027] Please continue to refer to Figure 1 The method for selecting the location of underground gas storage chambers for compressed air energy storage systems according to embodiments of the present invention further includes: Step S103: Output the dynamic risk threshold corresponding to the target area through the deep belief network, and quantify the risk probability under parameter uncertainty based on the Bayesian network. Eliminate the storage addresses with excessive risk in the storage address set according to the risk probability to obtain the optimized storage address set. Furthermore, the step of outputting a dynamic risk threshold corresponding to the target area through a deep belief network includes: Multi-dimensional features related to risk are selected from the target area to form an input feature set; the multi-dimensional features include at least regional geological features, climate and hydrological features, energy storage operation features and ecological constraint features. Construct a deep belief network model, wherein the deep belief network model contains a multi-layer restricted Boltzmann machine; The model is trained using historical site selection case data. During the pre-training phase, the reconstruction error is minimized using contrastive divergence, as shown below: In the formula, This represents the original value of the i-th node in the input layer. This represents the reconstruction value at the i-th stage of the input layer, where n represents the number of nodes in the input layer. The training objective is to establish a mapping relationship from the input feature set to various risk thresholds; The input feature set of the target region is input into the trained deep belief network model, which outputs a basic dynamic risk threshold. Based on the specific scene conditions of the target region, the basic dynamic risk threshold is corrected to obtain the final adapted dynamic risk threshold.

[0028] Furthermore, the specific scenario conditions include at least one of earthquake intensity, annual rainfall, or carbon sink density; When the seismic intensity of the target area is higher than the preset level, the risk threshold of the rock mass stability coefficient is increased accordingly; when the annual rainfall of the target area is higher than the preset value, the risk threshold of the groundwater influence coefficient is increased accordingly. When the carbon sink density in the target area is higher than the preset value, the risk threshold for the probability of seal failure is reduced accordingly.

[0029] For further details, please refer to Figure 3 In one implementation of the present invention, the step of quantifying the risk probability under parameter uncertainty based on a Bayesian network includes: Step S301: Construct a multi-level Bayesian network containing an input layer, an intermediate layer, and an output layer; wherein, the input layer nodes represent key parameters with uncertainty, the intermediate layer nodes represent the response of the multi-field coupled model, and the output layer nodes represent the risk to be evaluated; Step S302: Based on the multi-field coupling mechanism and historical case data, determine the dependencies between nodes to construct a directed acyclic graph, and use maximum likelihood estimation to construct a data-driven conditional probability table for each node, represented as: In the formula, This represents the conditional probability of a node combination. This indicates the number of historical observations for this node combination; Step S303: Based on the parameter probability distribution of the input layer nodes, calculate the probability of each risk node occurring in the output layer, as well as the joint probability of multiple risks occurring together, using a probabilistic inference algorithm, expressed as: In the formula, Indicates risk and The probability of their joint occurrence, , As an intermediate response node, This represents the uncertainty parameters of the input layer. Indicates parameters The prior probability distribution.

[0030] In one implementation of the present invention, the key parameters of the input layer include at least the rock mass permeability coefficient, the aging rate of the chamber sealing material, the carbon sink growth rate, the groundwater recharge, and the rock mass elastic modulus, and each parameter is assigned a probability distribution type determined based on measured and simulated data. The data sources for constructing the conditional probability table include: response results obtained by random sampling simulation of input parameters through a multi-field coupling model, and statistical relationships of historical risk cases of compressed air energy storage caverns.

[0031] In the step of calculating the probability of occurrence of each risk node in the output layer using a probabilistic reasoning algorithm, the Markov chain Monte Carlo algorithm is used to perform the probabilistic reasoning, and the probability of occurrence of each risk, the joint probability, and the contribution of each uncertainty parameter to the risk are output.

[0032] Furthermore, the step of automatically correcting the simulation parameters of the digital twin specifically includes: using the simulation parameters of the digital twin as the action space, using the deviation between the simulated values ​​of the coupled model and the measured values ​​of the optimized address set as the state space, using the minimization of the deviation as the reward function, the reward function being a composite function whose value is jointly determined by the basic reward, the accuracy reward, and the oscillation penalty; the basic reward is calculated based on the weighted sum of multiple simulation deviations; the accuracy reward is granted when a specific deviation reaches a high accuracy standard; and the oscillation penalty is triggered when the parameters are detected to be repeatedly corrected in the opposite direction. Construct a deep Q-network, taking the state space as input and outputting the Q-values ​​of each action in the action space as the objective; train the network using historical and real-time data, updating the Q-values ​​using temporal difference learning during training, as shown below: In the formula, This represents the Q value of performing action a in state s. Let r represent the learning rate and r represent the immediate reward. Indicates the discount factor. This indicates the next state after action a is performed. For state The optimal action under the given conditions; Furthermore, in this embodiment of the invention, the current state is input into the trained DQN network, the action with the highest Q value is selected to correct the simulation parameters of the digital twin, and the state is updated based on the corrected simulation results. This process is iterated until the simulation deviation meets the preset requirements.

[0033] In terms of site search and selection, this invention employs an improved multi-objective particle swarm optimization, aiming to maximize the carbon sink synergy coefficient, minimize construction and operation costs, and minimize geological risks. Through rapid non-dominated sorting and crowding calculation, it automatically generates a uniformly distributed Pareto optimal site set. Then, it outputs a dynamic risk threshold bound to regional characteristics through a deep belief network and uses a Bayesian network to accurately quantify the risk probability under parameter uncertainty, eliminating sites with excessive risks and obtaining a safe and controllable optimized site set.

[0034] Please continue to refer to Figure 1 The method for selecting the location of underground gas storage chambers for compressed air energy storage systems according to embodiments of the present invention further includes: Step S104: Construct an adaptive correction model based on the DQN network to automatically correct the simulation parameters of the digital twin. Based on the digital twin with corrected parameters, construct a comprehensive energy efficiency evaluation index system. Calculate the weights of each index in the system by weight allocation. Combine the measured data of each site in the optimized site set with the simulation parameters of the corrected twin to calculate the comprehensive energy efficiency value and determine the optimal site selection result.

[0035] In step S104, based on the parameter-corrected digital twin, a comprehensive energy efficiency evaluation index system is constructed. The weights of each index in the system are calculated through weight allocation. Combined with the measured data of each site in the optimized site set and the simulated parameters of the corrected digital twin, the comprehensive energy efficiency value is calculated, and the optimal site selection result is determined. This includes the following steps: Establish a hierarchical evaluation index system with multiple dimensions, including at least geological safety, energy storage operation, ecological carbon sink and full-cycle economic efficiency, with each dimension containing several quantifiable secondary evaluation indicators; The combined weights of each secondary indicator in the indicator system are calculated using a method that integrates subjective and objective weighting, and are expressed as follows: In the formula, This represents the combined weight of the j-th secondary indicator. Represents the balance coefficient. Indicates subjective weighting. Indicates objective weighting. Let m represent the information entropy of the j-th indicator, and m represent the number of optimized database addresses. This represents the proportion of the normalized value of the j-th indicator for the i-th database address; For each site in the optimized site set, based on its measured data and the simulation parameters of the corrected digital twin, the quantified values ​​of each secondary indicator are calculated; after normalizing the quantified values, and combining them with the combined weights, the comprehensive energy efficiency value of each site is calculated, expressed as: In the formula, This represents the overall energy efficiency value of the i-th storage site. Indicates the scene adaptation coefficient. This represents the normalized value of the j-th index for the i-th database address; Based on the comprehensive energy efficiency value, the final preferred storage site is determined from the optimized storage site set.

[0036] In a preferred embodiment of the present invention, the secondary indicators under the geological safety dimension include at least one of the following: long-term stability coefficient of rock mass, probability of failure of chamber sealing, and risk coefficient of groundwater intrusion. In a preferred embodiment of the present invention, the secondary indicators under the energy storage operation dimension include at least one of the following: chamber volume utilization rate, compression heat recovery efficiency, and charging / discharging pressure fluctuation coefficient; In a preferred embodiment of the present invention, the secondary indicators under the ecological carbon sink dimension include at least one of the following: net increase in carbon sink over the entire life cycle, ecological restoration compliance rate, and carbon sink synergy coefficient. In a preferred embodiment of the present invention, the secondary indicators under the full-cycle economic dimension include at least one of unit gas storage cost, annual operation and maintenance cost ratio, and investment payback period.

[0037] Based on a digital twin with dynamically corrected parameters, this invention constructs a comprehensive energy efficiency evaluation index system that includes geological safety, energy storage operation, ecological carbon sequestration, and full-cycle economic efficiency. This system allocates weights through a combination of subjective and objective methods and calculates the comprehensive energy efficiency value of each site by combining measured data from the optimized site set with the twin simulation parameters, thereby eliminating the subjective arbitrariness of decision-making and ultimately obtaining the site selection scheme.

[0038] like Figure 4As shown, according to another embodiment of this application, a computer device is provided, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the underground gas storage chamber site selection method for compressed air energy storage system as described in any of the above embodiments.

[0039] The computer equipment can be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal equipment may include, but is not limited to, a processor and memory.

[0040] The processor can be a central processing unit, or other general-purpose processors, digital signal processors, application-specific integrated circuits, off-the-shelf programmable gate arrays or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting various parts of the terminal device via various interfaces and lines.

[0041] The memory can be used to store the computer program. The processor implements various functions of the terminal device by running or executing the computer program stored in the memory and calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, applications required for at least one function, etc.; the data storage area may store data created based on the use of the mobile phone, etc.

[0042] In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart memory card, secure digital card, flash memory card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0043] Another preferred embodiment of the present invention provides a storage medium, which is a computer-readable storage medium, and a computer program is stored in the computer-readable storage medium. When the computer program is executed by a processor, it can implement the steps of the underground gas storage chamber site selection method of the compressed air energy storage system described in the above embodiments.

[0044] The computer program includes computer program code, which may be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory, random access memory, electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0045] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for selecting the location of underground gas storage chambers for compressed air energy storage systems, characterized in that, Includes the following steps: Integrate multi-source data of the target area and establish a digital twin of the target area. Based on the digital twin, construct a multi-field coupling model. In the coupling model, use a hybrid temporal prediction network to output the parameter prediction sequence of the coupling model. Combine Sobol global sensitivity analysis to screen the parameters of the coupling model. A multi-objective particle swarm optimization approach is adopted, with the objectives of maximizing the carbon sink synergy coefficient, minimizing construction and operation costs, and minimizing geological risks. Pareto optimal site set is generated through fast non-dominated sorting and congestion calculation. The system outputs a dynamic risk threshold corresponding to the target area through a deep belief network, and quantifies the risk probability under parameter uncertainty based on a Bayesian network. Based on the risk probability, the system removes the storage sites in the storage site set that exceed the risk limit, thereby obtaining an optimized storage site set. An adaptive correction model based on the DQN network is constructed to automatically correct the simulation parameters of the digital twin. Based on the parameter-corrected digital twin, a comprehensive energy efficiency evaluation index system is constructed. The weights of each index in the system are calculated by weight allocation. Combined with the measured data of each site in the optimized site set and the simulation parameters of the corrected twin, the comprehensive energy efficiency value is calculated and the optimal site selection result is determined.

2. The method for selecting the location of underground gas storage chambers for compressed air energy storage systems according to claim 1, characterized in that, The multi-source data includes geological data, ecological data, and carbon sink data; The steps for establishing a digital twin of the target region include: Extract GIS data of macroscopic areas and BIM data of microscopic components from multi-source data; Based on GIS data, construct a macroscopic model covering the target area; Based on BIM data, a parametric method was used to construct the chamber model of the candidate gas storage chamber; The chamber model is embedded into the corresponding spatial position of the macro model in a geometrically aligned manner to form a fused model. The fused model is then modified to obtain a digital twin of the target area.

3. The method for selecting the location of underground gas storage chambers for compressed air energy storage systems according to claim 2, characterized in that, The steps for constructing a multi-field coupling model based on a digital twin include: Geological parameters, chamber structure parameters, ecological carbon sink parameters, and real-time monitoring parameters are extracted from the digital twin. Establish the mapping relationship between each parameter and the target physical field, wherein the target physical field includes at least the gas storage pressure field, the geological stress field, the groundwater seepage field, and the ecological carbon sink field; Based on the parameters mapped to each physical field, corresponding physical field sub-models are constructed to obtain a multi-field coupling model.

4. The method for selecting the location of underground gas storage chambers for compressed air energy storage systems according to claim 3, characterized in that, The steps for using a hybrid temporal prediction network to output the parameter prediction sequence of the coupled model in the coupled model, combined with Sobol global sensitivity analysis to screen the parameters of the coupled model, include: A hybrid neural network containing a Long Short-Term Memory (LSTM) layer and a Gated Recurrent Unit (GRU) layer is constructed. Historical monitoring data and future scene parameters from the digital twin are used as input features. The hybrid neural network is trained using the mean squared error loss function and the Adam optimizer to output the future time series of key parameters of the coupled model. Based on the digital twin, the variation range of each parameter of the coupled model is determined to construct the parameter space, and a sampling method is used to generate a sample set in the parameter space. The total order sensitivity index of each parameter to the output result of the coupled model is based on the Sobol method. Based on the calculated total order sensitivity index, key parameters that significantly affect the output of the coupled model are selected.

5. The method for selecting the location of underground gas storage chambers for compressed air energy storage systems according to claim 4, characterized in that, The steps for generating the Pareto optimal library address set include: Establish a multi-objective optimization function that includes at least maximizing the carbon sink synergy coefficient, minimizing construction and operation costs, and minimizing geological risks; Initialize the particle swarm based on candidate library addresses that satisfy basic constraints, using digital twins; During the iteration process, the particle velocity and position are updated using dynamically adjusted inertia weights and learning factors. In the formula, This represents the velocity of the i-th particle in the t-th round. This represents the position of the i-th particle in the t-th round. This represents the inertia weight in round t; , Represents the learning factor. , Represents a random number. Let represent the optimal position of the i-th particle. The global optimal position of the particle swarm is represented by T, and the total number of iterations is represented by T. A fast sorting mechanism that includes non-dominated sorting and objective function contribution evaluation is used to screen particles; Calculate the crowding degree of particles in the target space and weight the crowding degree of particles located in the critical site selection region; after the iteration is completed, select non-dominated solutions with high crowding degree values ​​to form the Pareto optimal site set.

6. The method for selecting the location of underground gas storage chambers for compressed air energy storage systems according to claim 5, characterized in that, The steps of outputting a dynamic risk threshold corresponding to the target area using a deep belief network include: Multi-dimensional features related to risk are selected from the target area to form an input feature set; the multi-dimensional features include at least regional geological features, climate and hydrological features, energy storage operation features and ecological constraint features. A deep belief network model is constructed and trained using historical site selection case data. During the pre-training phase, the reconstruction error is minimized using contrastive divergence, as shown below: In the formula, This represents the original value of the i-th node in the input layer. This represents the reconstruction value at the i-th stage of the input layer, where n represents the number of nodes in the input layer. The training objective is to establish a mapping relationship from the input feature set to various risk thresholds; The input feature set of the target region is input into the trained deep belief network model, which outputs a basic dynamic risk threshold. Based on the specific scene conditions of the target region, the basic dynamic risk threshold is corrected to obtain the final adapted dynamic risk threshold.

7. The method for selecting the location of underground gas storage chambers for compressed air energy storage systems according to claim 6, characterized in that, The steps for quantifying risk probability under parameter uncertainty based on Bayesian networks include: Construct a multi-level Bayesian network comprising an input layer, an intermediate layer, and an output layer; wherein the input layer nodes represent key parameters with uncertainty, the intermediate layer nodes represent the response of a multi-field coupled model, and the output layer nodes represent the risk to be evaluated; Based on multi-field coupling mechanisms and historical case data, the dependencies between nodes are determined to construct a directed acyclic graph. Maximum likelihood estimation is then used to construct a data-driven conditional probability table for each node, as follows: In the formula, This represents the conditional probability of a node combination. This indicates the number of historical observations for this node combination; Based on the parameter probability distribution of the input layer nodes, the probability of each risk node occurring in the output layer, as well as the joint probability of multiple risks occurring together, are calculated using a probabilistic inference algorithm, and expressed as: In the formula, Indicates risk and The probability of their joint occurrence, , As an intermediate response node, This represents the uncertainty parameters of the input layer. Indicates parameters The prior probability distribution.

8. The method for selecting the location of underground gas storage chambers for compressed air energy storage systems according to claim 7, characterized in that, The steps for automatically correcting the simulation parameters of a digital twin specifically include: The simulation parameters of the digital twin are used as the action space, the deviation between the simulation values ​​of the coupled model and the measured values ​​of the optimized address set are used as the state space, and the minimization of the deviation is used as the reward function. Construct a deep Q-network, taking the state space as input and outputting the Q-values ​​of each action in the action space as the objective; train the network using historical and real-time data, updating the Q-values ​​using temporal difference learning during training, as shown below: In the formula, This represents the Q value of performing action a in state s. Let r represent the learning rate and r represent the immediate reward. Indicates the discount factor. This indicates the next state after action a is performed. For state The optimal action under the given conditions; The current state is input into the trained DQN network, the action with the highest Q value is selected to correct the simulation parameters of the digital twin, and the state is updated based on the corrected simulation results. This process is repeated until the simulation deviation meets the preset requirements.

9. The method for selecting the location of underground gas storage chambers for compressed air energy storage systems according to claim 8, characterized in that, Based on the parameter-corrected digital twin, a comprehensive energy efficiency evaluation index system is constructed. The steps include calculating the weights of each index in the system through weight allocation, combining measured data from each site in the optimized site set with the simulated parameters of the corrected digital twin, calculating the comprehensive energy efficiency value, and determining the optimal site selection result. Establish a hierarchical evaluation index system with multiple dimensions, including at least geological safety, energy storage operation, ecological carbon sink and full-cycle economic efficiency, with each dimension containing several quantifiable secondary evaluation indicators; For each site in the optimized site set, based on measured data and the simulation parameters of the corrected digital twin, the quantified values ​​of each secondary indicator are calculated; after normalizing the quantified values, and combining them with the combined weights, the comprehensive energy efficiency value of each site is calculated, expressed as: In the formula, This represents the overall energy efficiency value of the i-th storage site. Indicates the scene adaptation coefficient. This represents the normalized value of the j-th index for the i-th database address; Based on the comprehensive energy efficiency value, the final preferred storage site is determined from the optimized storage site set.

10. The method for selecting the location of underground gas storage chambers for a compressed air energy storage system according to claim 9, characterized in that, The combined weights are calculated using a combination of subjective and objective weighting methods, and are expressed as follows: In the formula, This represents the combined weight of the j-th secondary indicator. Represents the balance coefficient. Indicates subjective weighting. Indicates objective weighting. Let m represent the information entropy of the j-th indicator, and m represent the number of optimized database addresses. This represents the percentage of the normalized value of the j-th index for the i-th database address.

Images