101results about How to "Describe well" patented technology

The invention discloses a learning plan recommendation method and system, and an education robot, and relates to the technical field of education robots, demand identification is carried out by using a trained personalized learning demand identification model, and personalized learning demands are output and identified; matching learning resources and learning plans according to personalized learning requirements of the target students, outputting an optimal learning path, collecting feedback data for comprehensive evaluation, generating a state coefficient according to evaluation data, and if the state coefficient does not exceed expectation, optimizing the personalized optimized learning path of the target students by using a multi-target optimization algorithm, so as to improve the learning efficiency of the target students. Classifying the target students based on a pre-trained density clustering algorithm; and after the abnormal learning behavior is identified, starting a multi-level alarm mechanism, and outputting a corresponding response strategy by an emergency response strategy library according to the abnormal learning behavior. And the current learning path is adjusted and optimized, so that the learning path of the student is more matched with the current learning resource and learning plan.

The invention discloses an equipment maintenance knowledge representation method combining knowledge distillation and semantic analysis. The method comprises the steps of collecting structured data, unstructured text data and multi-modal data, constructing a semantic knowledge graph by using a natural language processing technology, extracting visual knowledge by using a computer vision technology, and fusing the visual knowledge with semantic knowledge to form a multi-modal knowledge base. And performing interactive knowledge distillation on the text data by using a large language model, generating optimized knowledge key points, and fusing the optimized knowledge key points with the multi-modal knowledge base to generate a comprehensive knowledge base. Knowledge is hierarchically organized according to the domain ontology and stored in a graph database, a natural language query interface is constructed, intelligent answering is performed by integrating technologies, question and answer corpus training is automatically generated, and automatic question and answer services are provided for users. According to the method, heterogeneous data integration, multi-modal knowledge fusion, hierarchical knowledge organization, natural language interaction and man-machine collaborative optimization are realized, and an intelligent equipment maintenance knowledge base can be efficiently constructed.

The invention relates to the field of battery health state prediction, in particular to a lithium battery health state evaluation method based on an EKF optimization algorithm considering capacity and cycle index. According to the technical scheme, when voltage, current and time are known, a second-order equivalent circuit model of the lithium battery is firstly established, the influence of the number of cycles on the maximum available capacity is considered, and a model between SOC-OCV-Cycs is established; the capacity is dynamically adjusted after one-time circulation is finished; when the Kalman filtering algorithm is used for correcting errors, the influence of aging is considered, and the estimation precision of the SOH can be achieved by comprehensively considering the influence of the number of cycles.

A water leakage monitoring method for a digital twinning system of a coal mine shaft comprises the steps that an RFID antenna and a Twinin-Tag sensor are installed on the wall of the shaft, a monitoring terminal is arranged on the ground, and an RFID reader is arranged at an outlet of the shaft; a shaft monitoring signal is collected in real time through a Twindow-Tag sensor, the shaft monitoring signal is read through an RFID reader, and then shaft monitoring data are obtained by analyzing and processing the shaft monitoring signal; collecting RSSI reading and phase data by using a short time window, and performing state identification by using a decision tree; carrying out coarse segmentation and fine segmentation on the leakage signal, and determining the location of a water leakage point in combination with a change point detection algorithm; and the leakage volume and the leakage rate are estimated. According to the method, the water leakage point can be accurately positioned, the volume and flow velocity information of the leaked water in the shaft can be obtained, so that a reliable basis is provided for subsequent maintenance decision making, and the method is suitable for long-term monitoring operation of the coal mine shaft.

The invention discloses an image retrieval system and method based on color moments and deep learning features. The system comprises an image feature extraction module, an index construction module, a retrieval trap door generation module and an image retrieval module. The image feature extraction module extracts color moment features and deep learning features of images in the image data set to obtain image multi-level features; the index construction module is used for constructing an index tree from bottom to top for color moment clustering; the trap door generation module extracts hierarchical features of the query image to generate a retrieval image trap door; the image retrieval module traverses the index tree from top to bottom according to the retrieval trap door to obtain a similar image candidate set, calculates the similarity between the depth feature of the image in the candidate similar image set and the depth feature of the query image, and outputs top-k images similar to the query image; according to the method, the color moments of the images are extracted and clustered, and the index tree is constructed, so that the similarity between the comparison images and the query images does not need to be calculated one by one during retrieval, and the image retrieval efficiency is improved.

The invention discloses a numerical analysis method for predicting speed distribution of downwash airflow of an unmanned aerial vehicle in a rice canopy based on a fluid-solid coupling method. The numerical analysis method comprises the following steps: measuring physical parameters of rice plants based on physiological characteristics of rice; establishing a three-dimensional rice plant model based on three-dimensional geometric modeling software; dividing a three-dimensional rice plant canopy grid based on grid division software; establishing a fluid-solid coupling model of the lower washing flow field of the unmanned aerial vehicle and the rice canopy; setting an initial boundary condition and performing spatial discretization; full-coupling finite element numerical simulation of the fluid-solid coupling model is carried out; performing post-processing on the numerical simulation result based on flow field post-processing software to obtain an evolution process and spatial distribution of the downward washing flow field of the unmanned aerial vehicle and dynamic behaviors of rice plants; and analyzing the influence of the dynamic behavior of the rice plant on the downwash airflow of the unmanned aerial vehicle, and determining the distribution rule of the downwash airflow speed in the rice canopy. The problems of high time consumption, influence of external conditions, difficulty in copying, limitation and the like are effectively solved.

The invention relates to the technical field of medical images, in particular to a neural CT map segmentation method based on high-precision registration. The method comprises the following steps: acquiring a nerve CT image; performing coarse registration based on mutual information according to the CT image, and performing multi-scale fusion based on wavelet to obtain a fused image; performing standard map preprocessing on a preset standard nerve CT map database according to the fused image to obtain a preprocessed standard map set; adding a multi-scale feature fusion module to a pre-established basic U-Net architecture, and embedding a space attention module to obtain a to-be-trained U-Net network; and performing deep registration network training on the to-be-trained U-Net network by using the fused image and the preprocessed standard atlas set to obtain a trained U-Net network. According to the invention, through technologies of multi-modal image fusion, deep learning registration and conditional random field optimization, a segmentation effect with higher precision and higher robustness is realized.

The invention provides a small-clear-distance tunnel surrounding rock pressure calculation method, and aims to solve the problem that the existing small-clear-distance tunnel surrounding rock pressure calculation is complicated and improve the existing small-clear-distance tunnel surrounding rock pressure calculation method. According to Hp and hq, dividing the burial depth condition of the tunnel into deep burial, shallow burial and ultra-shallow burial; when the tunnel is deeply buried, a method for calculating the surrounding rock pressure of the deeply-buried small-clear-distance tunnel is adopted; when the tunnel is shallowly buried, judging whether bias voltage exists in the tunnel or not; if the unsymmetrical pressure does not exist, a calculation method of the non-unsymmetrical pressure surrounding rock pressure of the shallow-buried small-clear-distance tunnel is adopted; if the bias voltage exists, adopting a calculation method of the bias voltage surrounding rock pressure of the shallow-buried small-clear-distance tunnel; when ultra-shallow burying is carried out, whether bias voltage exists or not is judged; if the bias voltage does not exist, a calculation method of the non-bias voltage surrounding rock pressure of the ultra-shallow buried small-clear-distance tunnel is adopted; and if the bias voltage exists, a calculation method of the bias voltage surrounding rock pressure of the ultra-shallow-buried small-clear-distance tunnel is adopted.

The invention provides an isogeometric topology optimization method and device based on thermoelastic metamaterial design, and relates to the field of isogeometric topology optimization. The method comprises the following steps: constructing a multi-material interpolation model corresponding to a thermoelastic metamaterial, and obtaining multiple groups of density distributions corresponding to multiple solid materials and one density distribution corresponding to one solid material in a thermoelastic metamaterial design domain according to the multi-material interpolation model; obtaining a relation interpolation among the multiple groups of density distributions, and obtaining an elastic modulus and a thermal expansion coefficient according to the relation interpolation; calculating an equivalent elastic matrix corresponding to the thermoelastic metamaterial according to the elastic modulus, and calculating a thermal expansion coefficient matrix corresponding to the thermoelastic metamaterial according to the thermal expansion coefficient; performing discretization form processing on the calculation equations of the equivalent elastic matrix and the thermal expansion coefficient matrix through an isogeometric grid, constructing a constraint function according to the equivalent elastic matrix, constructing a target function according to the thermal expansion coefficient matrix, and calculating sensitivity information corresponding to the first design variable through the target function and the constraint function; updating the first design variable according to the sensitivity information until the thermal expansion coefficient meets a convergence condition, and acquiring a second design variable meeting the convergence condition through geometric topology design models such as a negative thermal expansion metamaterial; and performing optimization design on the thermoelastic metamaterial according to the second design variable. The problem that when a complex geometric structure is processed, a finite element method is difficult to guarantee efficient and high-precision design of the thermoelastic metamaterial is solved.

The invention discloses a multi-section cohesion unit analysis method considering a composite material resin enrichment area and a fiber bridging influence mechanism, and belongs to the technical field of cohesion unit analysis methods.The multi-section cohesion unit analysis method comprises the steps that a multi-direction laying layer angle DCB sample is designed and prepared, and a DCB experiment is carried out; s2, calculating data obtained in the experiment in S1 to obtain I-type interlayer fracture toughness-crack length data of the DCB samples with different interlayer laying angles, and drawing an R curve; establishing a numerical analysis model of the DCB sample; analyzing the parameters of the four CZM units according to a J integral method in combination with data in an experiment to obtain cohesion model parameters of each unit; and a sectional cohesion model is adopted to simulate the layering damage behavior of the composite material laminated plate. By adopting the multi-section cohesion unit analysis method considering the resin enrichment area and the fiber bridging influence mechanism of the composite material, a more accurate I-type layering initial strain energy release rate can be obtained, and the performance evaluation of the composite material can be more accurate.

The invention relates to the technical field of metal surface defect detection, and discloses a metal surface defect detection method which comprises the following steps: detecting two circular surfaces of a circular metal sheet, and marking specific points of a first detection surface and a second detection surface; primary coating operation is conducted on the first detection face, a first workpiece is obtained, rubbing is conducted, new points are marked, and a first rubbing graph is obtained; performing secondary coating operation on the second detection surface to obtain a second workpiece, rubbing and marking new points to obtain a second rubbing graph; acquiring images of the two rubbing graphs, and processing the images to form a first image and a second image; normal and abnormal areas on the detection surface are obtained according to the image, and the to-be-detected workpieces are classified after comparison; according to the number of the concave positions and the convex positions of the detection surfaces and the corresponding relation, the workpieces are classified, different defect types of the workpieces are distinguished, a corresponding processing scheme is better selected in the production process, the production cost is reduced, and the utilization rate of the workpieces is improved.

The invention belongs to the field of bioinformatics, and particularly provides a traditional Chinese medicine target relation prediction method based on multi-attention graph neural network learning. The method comprises the following steps: firstly, constructing a heterogeneous network comprising a plurality of entities such as traditional Chinese medicines, targets, diseases and prescriptions and a plurality of edge types so as to better describe the relationship between the traditional Chinese medicines and the targets; then, a method based on a multi-attention graph neural network is provided to learn fusion feature representation of traditional Chinese medicine and targets based on a heterogeneous network; and finally, realizing the prediction of the traditional Chinese medicine-target relationship based on a deep neural network DNN. Based on the multi-source heterogeneous data and the heterogeneous relationship, the traditional Chinese medicine and the target are subjected to fusion feature representation by using the multi-attention graph neural network, more comprehensive information is considered, and the prediction accuracy is effectively improved while the traditional Chinese medicine target relationship prediction is realized.

This invention discloses a method and storage medium for reconstructing the energy surface of vacancy-helium-hydrogen complexes. The main steps include: S1, calculating the energy of small complex clusters; S2, analyzing the formation energies of vacancy-helium and vacancy-hydrogen complexes with respect to p2 and p3, respectively; S3, selecting a suitable univariate function to fit the relationship between cluster energy and p2 and p3 at a given p1; S4, fitting the relationship between the univariate function parameters and p1; and S5, calculating the energy of arbitrary clusters and defect energy levels. This method reconstructs the energy surface of vacancy-helium-hydrogen complexes based on a class of univariate functions, enabling high-precision reproduction of the energy and energy levels of defect clusters. This reconstruction method not only accurately describes the energy of defect clusters within a known computational set but also provides a good description of the energy and energy level trends of clusters outside the computational set, exhibiting excellent extrapolation properties.

This invention relates to the field of social network analysis technology and proposes a social network sentiment evolution model that integrates user influence and activity. To address the shortcomings of existing technologies, such as inaccurate simulations and inflexible predictions in current network sentiment analysis and sentiment evolution processes, this invention provides a social network sentiment evolution model that integrates user influence and activity. The model analysis method involves: collecting text sentiment data from target users; collecting user influence data from each user; collecting the activity level of the target social network at time t; collecting the forgetting probability of each user at time t as a measure of sentiment evolution capability; establishing a dynamic model for the sentiment evolution of the social network; optimizing the evolution probability of the information propagation model based on text sentiment; and analyzing the evolution process of the dynamic model for the sentiment evolution of the social network based on the optimized information propagation model. This model is suitable for application in social network analysis.

The invention discloses a liquid hydrogen superconducting co-melting energy storage semi-physical simulation system which comprises a liquid hydrogen superconducting co-melting energy storage physical system, a real-time simulation machine and an upper computer. The liquid hydrogen superconducting co-melting energy storage physical system comprises a fuel cell device, a water electrolysis hydrogen production device and a liquid hydrogen cooling superconducting magnetic energy storage device. The fuel cell device is used for providing electric power supply for the electric power system; the water electrolysis hydrogen production device is used for converting electric energy into chemical energy to be stored in hydrogen; the liquid hydrogen cooling superconducting magnetic energy storage device is used for carrying out liquid hydrogen energy storage and superconducting magnetic energy storage; the real-time simulation machine is used for generating a control signal of the liquid hydrogen superconducting co-melting energy storage physical system according to the operation parameters of the liquid hydrogen superconducting co-melting energy storage physical system and the operation condition of the simulation renewable energy power generation system; and the upper computer is used for monitoring and managing the operation conditions of the real-time simulation machine and the liquid hydrogen cooling superconducting magnetic energy storage real object device. According to the invention, the accuracy and reliability of the simulation result can be improved.

A stable pharmaceutical composition belongs to the field of pharmaceutical technology. The pharmaceutical composition, whose active ingredients are edaravone and dextrorotatory borneol, can unexpectedly control the contents of its unique impurities SCR-756 and impurity SCR-757, thereby solving the problems of impurities and product quality being difficult to control and short storage period from the source.

The invention relates to the field of tight gas reservoir technology development, and provides a gas reservoir residual gas quantitative description method which comprises the following steps: generating a productivity parameter of a tight gas reservoir according to a preset productivity equation; constructing an initial production and storage coefficient definition formula of the tight gas reservoir according to the production capacity parameter and a pre-obtained energy storage parameter; performing target parameter conversion on the initial production and storage coefficient definition formula to obtain a target production and storage coefficient definition formula of the initial production and storage coefficient definition formula; and obtaining physical property parameters of the tight gas reservoir, and generating a production and storage coefficient distribution diagram of the tight gas reservoir according to the physical property parameters and the target production and storage coefficient definition formula. The invention further provides a gas reservoir residual gas quantitative description device, equipment, a medium and a program. The accuracy of quantitative description of the residual gas of the gas reservoir can be improved.

The application discloses a kind of wheat scab monitoring methods based on satellite remote sensing, belong to remote sensing monitoring technical field, first obtain the remote sensing data and meteorological precipitation data of wheat planting area in predicted year and N years before predicted year, the remote sensing data include multispectral data and thermal infrared remote sensing data;Then the relative vigor index is calculated using the multispectral data;Using the multispectral data and thermal infrared remote sensing data obtains normalized surface temperature index;Using the meteorological precipitation data obtains normalized precipitation index;Finally, the relative vigor index, normalized surface temperature index and normalized precipitation index are used to calculate the wheat scab incidence index of predicted year;The application can better reduce the interference caused by the difference of wheat vigor in different regions in large-scale monitoring on scab monitoring model, improve the objectivity and practicability of scab remote sensing monitoring.

The invention relates to a construction and control method of an input-output control system for industry. The construction and control method comprises the steps that an input-output model between the equipment manufacturing industry and the petrochemical industry in a single region is established based on an input-output model; expanding the established input-output model to a multi-region input-output model; establishing a multi-region input-output diffusion model and an input-output control system; designing a pulse controller of the actuator fault; and analyzing the index tolerance of the input-output control system to obtain the gain of the pulse observer and the gain of the pulse controller for the designed pulse observer and the pulse controller to complete the control of the input-output control system. Compared with a continuous time observation and control method, the method provided by the invention can realize the balance between supply and demand of industrial production by regulating and controlling the product demands of the equipment manufacturing industry and the petrochemical industry with lower cost and time consumption.

The present application provides a method and device for determining the steady-state frequency response of an MHD angular velocity sensor element. The method comprises: obtaining structural dimension parameters and material property parameters of the MHD angular velocity sensor element in response to a modeling instruction; generating an average magnetic induction intensity of a radial magnetic circuit in a conductive fluid ring in the MHD angular velocity sensor element based on the Maxwell equations according to the structural dimension parameters and the material property parameters; generating an average circumferential velocity of the conductive fluid relative to the magnetic field based on a low magnetic Reynolds number incompressible equation group according to the structural dimension parameters and the material property parameters; generating a frequency response characteristic function of the MHD angular velocity sensor element based on the average magnetic induction intensity and the circumferential average velocity based on the Faraday motional electromotive force method; and solving the frequency response characteristic function to obtain amplitude-frequency response data and phase-frequency response data of the MHD angular velocity sensor element.

The invention relates to the technical field of scheduling optimization problem solving, in particular to an accelerated solving method and system for a continuous time scheduling problem based on integer variable pattern analysis. According to the method provided by the invention, the solution time of the continuous time scheduling problem can be effectively shortened by analyzing and mining the mode information hidden in the historical optimal solution of the scheduling problem. The invention provides a method for effectively extracting the integer variable pattern information in the continuous time scheduling problem and expressing the integer variable pattern information as reinforced inequality constraints, and the solving efficiency of the continuous time scheduling problem is remarkably improved by utilizing the pattern information in the optimal solution of the same type of scheduling problems. Compared with a traditional genetic algorithm or direct solving by using a commercial solver, the method has the advantages that the solving speed of a class of mixed integer programming problems can be remarkably improved by reducing the feasible region of the optimization problem to be solved, and meanwhile, the optimality of a scheduling scheme is well ensured.

The present invention claims a method for spatial downscaling of land surface temperature based on the Geographically Weighted Durbin Model (GWDM). Land surface temperature with high temporal and spatial resolution is crucial in the study of environmental ecology and climate systems. However, due to technical limitations, satellite thermal sensors cannot provide thermal infrared images with high temporal resolution and high spatial resolution at the same time. The spatial downscaling method combines low spatial resolution LST and auxiliary data with high spatial resolution from other satellite sensors to improve the temporal and spatial resolution of temperature images under the premise of assuming scale invariance. However, spatial non-stationarity and spatial autocorrelation coexist in most spatial variables, and the spatial characteristics of land surface temperature should be fully considered as spatial variables in the downscaling process. However, previous studies on downscaling of land surface temperature only considered spatial non-stationarity and ignored the spatial autocorrelation between independent variables and dependent variables.

The invention relates to the technical field of block chains, in particular to an export trade account receivable financing digital asset management method capable of increasing credit, and the method comprises the steps: digital asset generation: a financing party stores a hash value of a file generated in export trade on a block chain system, and the corresponding relation is file 1.. n to hash1.. n; and performing digital asset transaction: querying related information of the digital assets by other parties, obtaining the original text of the evidence storage file from other ways, comparing the original text with the hash value of the evidence storage so as to ensure that the information of the file is consistent with the evidence storage and no tampering behavior exists, and purchasing a certain share of digital assets according to prices agreed by the two parties. According to the method, the digital assets with different qualities are flexibly distinguished, the expressive force and the transaction capability of the digital assets are enhanced, and the tampering resistance of the digital assets is improved by combining the certificate of the digital assets with the generation process of the digital assets.

The invention discloses an analog circuit performance index obtaining method and device based on a convolutional neural network, and relates to the technical field of integrated circuits, and the method comprises the steps: obtaining a to-be-recognized analog circuit, and determining the design parameters of the to-be-recognized analog circuit; according to the structure of the analog circuit to be identified, mapping the design parameters into the two-dimensional space matrix to obtain a two-dimensional sparse matrix corresponding to the design parameters; inputting the two-dimensional sparse matrix into a trained convolutional neural network model for processing to obtain a performance index corresponding to the design parameter; wherein the trained convolutional neural network model takes data of a preset category as a training data set, the initial convolutional neural network model is trained to obtain the trained convolutional neural network model, and the data of the preset category comprises design parameters and calibration performance indexes of the analog circuit of which the type is the same as that of the analog circuit to be identified; the design parameters are in one-to-one correspondence with the calibration performance indexes. The design efficiency of the integrated circuit can be improved.

The invention discloses a learning plan recommendation method, system and educational robot, which relates to the technical field of educational robots. The method uses a trained personalized learning demand identification model to identify the demand, outputs the identified personalized learning demand; matches learning resources and learning plans according to the personalized learning needs of target students, and outputs the optimal learning path; collects feedback data for comprehensive evaluation and generates a state coefficient from the evaluation data; if the state coefficient does not exceed the expectation, uses a multi-objective optimization algorithm to optimize the personalized optimization learning path of the target student, and classifies the target students based on a pre-trained density clustering algorithm; after identifying abnormal learning behavior, a multi-level alarm mechanism is enabled, and according to the abnormal learning behavior, the corresponding response strategy is output from the emergency response strategy library. The current learning path is adjusted and optimized to make the student's learning path more matched with the current learning resources and learning plan.

The invention belongs to the technical field of artificial intelligence, and provides a system operation process description generation method and device based on artificial intelligence. According to the technical scheme, key value pair matching is conducted on element names and display names; interaction information of a user and a service system is analyzed, analysis and feature extraction are carried out on the interaction information, key information and intention of user interaction are obtained, all service process information is stored in the constructed database, and the service process information comprises the interaction information, an analysis result, the key information and intention information; performing feature extraction, feature selection and feature coding on the business process information, and generating a sequence description of the business process by using the trained operation process description generation model; the key value pair matching result is performed according to the element name and the display name, and the sequence description of the business process is converted into the display result to be output, so that the problems of use cost and information security risk are solved, and accurate requirements are provided for RPA business process design.

The invention belongs to the field of petroleum and natural gas development, and particularly relates to a fatigue life prediction method based on drill string dynamic stress, and the method comprises the following steps: inputting basic parameters: the basic parameters comprise a well track, a bottom drilling tool combination, a drill string structure and drilling parameters; according to a Lagrange equation, a drill string system dynamic model is established; node velocity, displacement and acceleration are obtained through a drill string system dynamics model, and drill string dynamic equivalent stress is calculated according to the node velocity, displacement and acceleration; counting a stress amplitude and a stress mean value by adopting a rain flow counting method; correcting the stress amplitude and the stress mean value by adopting a goodman formula; substituting into an S-N curve and calculating to obtain the fatigue life. According to the method, the fatigue life of the whole-well drill string can be analyzed, the fatigue service life of the whole-well drill string under the working condition is predicted through data analysis and processing, the presented result is the specific service time of the drill string, and the method has a very strong guiding effect on field engineering practice.

The present invention discloses a material for an organic electroluminescent element, an organic electroluminescent element, and a consumer product. The organic electroluminescent element comprises: an anode, a cathode, and a hole injection layer disposed between the anode and the cathode, wherein the hole injection layer comprises a compound represented by the general formula (I). The hole injection layer and the charge generation layer disclosed in the present invention comprise a naphthyridine-based compound or consist of a naphthyridine-based compound, which can reduce the voltage of an OLED element and improve the efficiency and life of the element.

The invention discloses a social network emotion evolution model fusing user influence and activeness, and relates to the technical field of social network analysis. In order to solve the technical defects of inaccurate simulation and inflexible prediction in the existing network sentiment analysis and sentiment evolution analysis process in the prior art, the invention provides a social network sentiment evolution model fusing user influence and activeness. The model analysis method comprises the steps of collecting a text emotion of a target user; collecting the user influence of each user; collecting the activeness of the target social network at the t moment; the forgetting probability of each user at the t moment is collected as the emotional evolution ability; an emotional evolution dynamics model of the social network; optimizing the evolution probability of the information spreading model according to the text emotion; and analyzing the evolution process of the emotional evolution dynamics model of the social network according to the optimized information propagation model. The method is suitable for social network analysis work.

The application discloses a compressed air energy storage power generation system containing dehydration, and relates to the field of compressed air energy storage power generation. The compressed air energy storage power generation system containing dehydration comprises an energy releasing subsystem and a controller. The energy releasing subsystem comprises a dehydration device, a temperature recovery device and an expansion power generation device. The working medium taken from a gas storage device is dehydrated through the dehydration device, then is heated through the temperature recovery device, and finally enters the expansion power generation device to generate power. The dehydration device comprises a first expander and a first separator. The working medium after passing through the first expander comprises liquid working medium and gaseous working medium. The liquid working medium is separated out through the first separator, and the gaseous working medium flows out of the outlet of the first separator and then flows into the temperature recovery device to be heated. The application solves the technical problem that the water content of the working medium in the compressed air energy storage system in the prior art is difficult to control, has the advantage that the working medium gas is always in a gaseous state in the expansion power generation process, and improves the efficiency and safety of the system.