32results about How to "Realization reservation" patented technology

The invention discloses an infrared and visible light fusion method based on multi-feature extraction. A multi-feature extraction network is composed of a fusion network, a semantic segmentation network, a content loss function and a semantic loss function. The fusion network comprises a feature extraction part and a feature reconstruction part, and the feature extraction part comprises a convolutional layer with an activation function, an adaptive gradient feature extraction module and a space channel reconstruction convolution and residual gradient extraction module; the feature reconstruction part is composed of an attention mechanism and four convolution layers connected in series; the method comprises the following steps: S1, inputting an infrared image and a visible light image into a fusion network, and generating a fusion image through feature extraction and feature reconstruction; s2, transmitting the fused image to a segmentation network for segmentation; and S3, comparing the segmentation result with a semantic tag, and evaluating richness of semantic information contained in the fused image. The contour of the fused image is clearer, the target is clearer, and the index is remarkably improved.

The invention discloses a universal self-supervised learning Raman spectrum noise reduction method, which comprises the following steps of: 1, acquiring Raman data; step 2, preprocessing the Raman data, removing a baseline from the Raman data by using an adaptive iterative reweighted penalty least square method, and standardizing a spectrum after baseline removal to obtain preprocessed spectrum data; step 3, processing the preprocessed spectral data by using a down-sampling strategy H (.) to obtain two sub-spectrum sets h1 (.) and h2 (.) with independent noise, each sub-spectrum set comprising 10 sub-spectrums; and step 4, inputting the preprocessed spectrum data and the corresponding two sub-spectrum sets into a U-Net network f theta (.) to obtain a spectrum after noise reduction processing. The noise reduction method can carry out good noise reduction processing on various Raman data, and the self-supervised learning method does not need a complex long-time learning process.

The invention discloses a model fine tuning method and system based on task priority and elastic weight solidification, belongs to the technical field of machine learning, and aims to solve the technical problem of how to provide a stable, efficient and low-power silk screen manufacturing model which can adapt to various complex production scenes. Comprising the following steps: collecting sample data, and dividing the sample data into a plurality of subsets; service weights and priority coefficients of all new tasks in the silk screen manufacturing process are analyzed based on the sorted subsets; constructing a knowledge graph based on the entities and the relationship between the entities; calculating the importance degree of the model parameters to the old task based on the Fisher information matrix, and adjusting the priority level of the model parameters based on the knowledge graph; and on the basis of the Fisher information matrix, updating a loss function of the silk screen manufacturing model under the new task through EWC, and correcting the current silk screen manufacturing model on the basis of entities and relationships in the knowledge graph, so that multi-task model fine tuning is realized.

The invention discloses a desertification region soil ecology improvement and restoration system which comprises a multifunctional biochar preparation module used for collecting desertification region crop waste and preparing biochar through high-temperature pyrolysis to serve as a soil conditioner; the microbial remediation module is used for selecting drought-enduring plant species and culturing soil microorganisms; the resource comprehensive utilization module is used for collecting rainwater and establishing a water storage system; and the soil remediation device comprises a fertilization unit, an irrigation unit, a soil softening unit and a vegetation planting unit. Soil improvement is achieved through fertilization, irrigation and soil softening, the survival rate of later plants is increased, power is supplied to the soil remediation device in cooperation with the comprehensive resource utilization module, the utilization rate of renewable resources is increased, and therefore the ecological relation of mutual promotion among soil, vegetation and microorganisms is constructed; the resource comprehensive utilization module and the soil remediation device can guarantee sustainable development of the ecological relationship, and soil ecology improvement and remediation in the desertification area are achieved.

The invention relates to a mainstream anaerobic ammonia oxidation quick start and efficient stable operation method, which uses an immersed anaerobic ammonia oxidation granular sludge membrane biological composite reactor comprising an upflow anaerobic sludge blanket (UASB), and an MBR is immersed in the upper part of the upflow anaerobic sludge blanket (UASB); the height-diameter ratio of the UASB is 4-6, and the overall length of the MBR is 1 / 2-1 / 3 of the height of the UASB. The granulation of the sludge is realized through the UASB, and the enrichment capacity, the sedimentation performance and the stability of the sludge are improved; in the operation process of the mainstream Anammox, the complete retention of sludge functional bacteria is realized by adopting the operation of gradually shortening the hydraulic retention time and improving the nitrogen load and through the biological complete interception effect of the immersed MBR. In the granulation process of the sludge, the settling performance of the sludge is improved, a large amount of flocculent sludge is prevented from being in direct contact with membrane filaments, the membrane pollution problem in the membrane assembly is relieved, meanwhile, water flow shearing force generated around the membrane assembly can effectively wash pollutants on the surface of the membrane assembly, and the membrane pollution problem is further controlled.

The invention relates to the technical field of solid waste treatment, and provides a closed-loop dechlorination and desalination method for incineration fly ash. Soluble and insoluble chlorine salts in the incineration fly ash are removed through alkali washing, and then heavy metals and chlorine salts in alkali washing waste liquid are removed through heavy metal removal, primary dechlorination and secondary dechlorination. The Ca-Al layered double hydroxides obtained in the process can be circularly used for removing heavy metals, and the recovered alkali solution can be circularly used for alkali washing. According to the method, the removal efficiency of chlorine salts, especially insoluble chlorine salts, in the incineration fly ash is remarkably improved, meanwhile, removal of sulfate is promoted, calcium in the fly ash is reserved, and resource utilization is facilitated. And meanwhile, secondary environmental pollution caused by wastewater is effectively prevented. The byproduct Ca-Al layered bimetal hydroxide has a wide application prospect in the fields of gas adsorption, wastewater purification, catalysis and the like due to unique interlayer ion exchange property and excellent adsorption capacity, and has a good economic value.

The invention discloses a problem processing method and device and electronic equipment. The method relates to the field of artificial intelligence, and comprises the following steps: under the condition that a target question of a user is received, determining at least one target document block required for answering the target question from a plurality of document blocks contained in a pre-constructed database; generating a prompt statement based on the at least one target document block and the target question, the prompt statement being used for guiding the target large model to determine a text answer corresponding to the target question according to the at least one target document block, under the condition that image link information associated with the text answer exists in at least one target document block, the target large model is guided to output the text answer and an image corresponding to the image link information; and answering the target question according to the prompt statement through the target large model. The technical problem of low answering accuracy due to the fact that a large model cannot answer user questions in combination with pictures in the prior art is solved.

The invention provides a wurtzite boron nitride / diamond composite material and a preparation method and application thereof, the composite material is composed of two phases of wurtzite boron nitride and diamond, and has a stacked sandwich microstructure, the sandwich microstructure comprises at least one first wurtzite boron nitride layer and at least one second wurtzite boron nitride layer, wherein the first wurtzite boron nitride layer is formed by wurtzite boron nitride; at least one second wurtzite boron nitride layer composed of wurtzite boron nitride; the composite material comprises a first wurtzite boron nitride layer, a second wurtzite boron nitride layer and a diamond layer, the diamond layer is arranged between the first wurtzite boron nitride layer and the second wurtzite boron nitride layer and is made of diamond, wurtzite boron nitride is a metastable phase stably existing under normal pressure, and the Vickers hardness of the composite material is larger than 50 GPa. The novel wurtzite boron nitride / diamond composite material disclosed by the invention has high hardness, high toughness and high thermal stability, and the preparation method disclosed by the invention is low in initial raw material price and simple in preparation process.

The invention provides a margin retaining device and a margin retaining method used in a sealing plate, and the margin retaining device comprises a folding fixing module, a cable surrounding wall, a guide plate, and a fixing cavity pedestal. The folding fixing module comprises an X-shaped folding rod, a telescopic rod and a T-shaped fixing sliding rod; a group of X-shaped folding rods are connected with a telescopic rod and then are connected with a T-shaped fixed sliding rod; the guide plate is provided with a group of cable placing grooves for placing cables and is fixed on the fixed cavity base; and the cable surrounding wall is fixed on the X-shaped folding rod, and the fixed cavity base is in sliding connection with the T-shaped fixed sliding rod. According to the technical scheme, the allowance in the sealing plate can be reserved, the error reworking phenomenon is reduced, and the cable winding and bending radian requirements are met.

The invention discloses a high-toughness plastic for a medical needle tube and a preparation method of the high-toughness plastic, and belongs to the technical field of polymer composite materials. The product comprises the following components: polycaprolactone, maleic anhydride grafted polylactic acid, a chitosan modified hydroxypropyl methyl cellulose derived carbon dot bacteriostatic agent and the like. The preparation method comprises the steps of preparation of the carbon dot bacteriostatic agent, pretreatment of raw materials, preparation of the functionalized master batch, segmented temperature control twin-screw extrusion melt blending and the like. The composite carbon dot bacteriostatic agent is constructed by hydrothermal carbonization-two-step amidation grafting, and is matched with the functionalized master batch preparation and interface regulation and control technology, so that the problems that the bacteriostatic agent of the existing medical needle tube material is easy to migrate, poor in compatibility and insufficient in toughness are solved, the synergistic improvement of long-term bacteriostasis and high toughness is realized, the biocompatibility is guaranteed, and the medical needle tube material has a good application prospect. The device is suitable for intrusive medical instruments such as medical needle tubes.

The invention relates to the technical field of traditional Chinese medicine preparations, and discloses a plaster for treating pain in neck, shoulders, waist and legs, which comprises a volatile and heat-sensitive medicinal material group, a heat-resistant medicinal material group with toxic components, a floating medicinal material group, sesame oil, a composite micro-coating matrix and a physical activator. During preparation, volatile and heat-sensitive medicinal materials are pretreated by adopting a composite micro-coating matrix, and toxic medicinal materials are subjected to micro-wetting treatment by adopting a physical exciting agent; and putting the mixed powder to be fed into high-temperature sesame oil at one time for reaction, then transferring to a closed mechanical forming cavity, and forming through a warm-pressure coupling process of medium-temperature normal pressure, high-temperature positive-pressure intense permeation and medium-temperature negative-pressure sudden drop. According to the invention, the degradation of toxic diester-type alkaloid and the retention of thermosensitive components are synchronously realized by utilizing the phase change heat insulation of the coating substrate and the vaporization and micro-boiling mechanism of the exciting agent, and the physical topological structure of the plaster is remolded through temperature-pressure coupling, so that the leaching amount of effective components, the transdermal release efficiency and the product forming stability are improved.

The application provides a wurtzite boron nitride / diamond composite material and a preparation method and application thereof. The composite material is composed of two phases of wurtzite boron nitride and diamond, and has a stacked sandwich microstructure, which comprises at least one first wurtzite boron nitride layer composed of wurtzite boron nitride, at least one second wurtzite boron nitride layer composed of wurtzite boron nitride, and a diamond layer composed of diamond and sandwiched between the first wurtzite boron nitride layer and the second wurtzite boron nitride layer. The wurtzite boron nitride is a metastable phase stably existing under normal pressure, and the Vickers hardness of the composite material is greater than 50 GPa. The novel wurtzite boron nitride / diamond composite material has high hardness, high toughness and high thermal stability, and the preparation method has low starting material price and simple preparation process.

The present invention provides a time series data completion method and system, the method comprising: collecting time series related operation data of a DC converter transformer, constructing an original time series defect data set X 0; for the original time series defective data set X 0 performs time correlation preprocessing to obtain the time correlation preprocessing matrix X 1; The time-dependent preprocessing matrix X 1 Input the CNN network to integrate the data sequence to obtain the integration matrix X 2. Integrate the matrix X 2. Send it to the bidirectional long short-term memory neural network Bi-LSTM for time correlation prediction learning to obtain the time correlation prediction result X 3. Predict the time correlation results X 3. Send it to the anti-attention layer to reduce the attention of the pre-filled data to obtain the reconstructed attention matrix X 4; Reconstruct the attention matrix X 4 is put into the fully connected output layer to obtain the data prediction completion result #imgabs0#. The present invention solves the technical problems of gradient disappearance, overfitting and low accuracy of completed data.

This utility model provides a sample-keeping anti-counterfeiting incense packaging tube, belonging to the field of incense packaging technology. It includes a top cover, a connecting tube, and an incense tube. The connecting tube has a reserved channel with a limiting component. The limiting component contains multiple radially distributed tooth-shaped elastic limiting pieces. In its natural state, the tips of the elastic limiting pieces form a conical protrusion, enclosing a closed limiting hole. This utility model, through the deformation and reset characteristics of the elastic limiting pieces, strictly restricts the incense to only moving from the incense tube towards the top cover for sample incense sampling and inspection. Once all incense is removed, it cannot be reversed, preventing incense inconsistent with the incense already in the tube from being mixed back into the tube. This avoids confusion between different batches and qualities of incense, ensuring the consistency between the incense quality in the packaging tube and the seller's quality commitment. This helps quality-conscious and reputable companies protect their products from counterfeiting or malicious complaints and claims.

The invention discloses a track line segmentation method for high-precision detection of the tail end of a track. The method comprises the following steps: constructing to-be-segmented track line image data; inputting the to-be-segmented track line image data into a track line segmentation network RTEW-Net for processing to obtain a train operation track line segmentation result; wherein the rail line segmentation network RTEW-Net is used for extracting rail line tail end preliminary features from the train operation rail line feature graph, performing feature detail maintenance on the preliminary features, and retaining the rail line tail end features in combination with the train operation rail line feature graph; the track line segmentation network RTEW-Net is obtained through training of a training set, and the training set is a track line image and a corresponding label. According to the invention, accurate segmentation of the track line is realized, and the running safety of the train is guaranteed.

The invention discloses a battery aging trajectory adaptive extrapolation method fusing physical time sequence prior, which comprises the following steps: collecting full life cycle constant current charging data of a battery, extracting an incremental capacity curve, and constructing characteristic peak evolution to form multi-dimensional physical time sequence prior; a state space backbone network is introduced, and trend perception is achieved through a priori initialization implicit state; constructing an attention anomaly detection module to discriminate real relaxation and noise; monotonic constraint is released through a micro-relaxation layer for a relaxation event, and data is corrected through Kalman filtering for noise; and constructing a dynamic trade-off loss function to adaptively adjust the physical constraint strength, and iteratively training the model based on the corrected data to realize stable extrapolation prediction. The method effectively solves the problems that in the prior art, real physical relaxation and measurement noise cannot be distinguished, and prediction deviation is caused by physical constraint rigidity.

The invention discloses a short-term photovoltaic power prediction method based on characteristic decomposition reconstruction, and the method comprises the steps: decomposing historical photovoltaic output power data into a plurality of intrinsic mode function sequences, screening out IMF components highly related to a prediction target through a Pearson's correlation coefficient method, and effectively eliminating noise interference; meanwhile, the screened IMF components are added and then denoised new features are added into the data set, the original data set is further enriched, higher-quality input is provided for the model, the capturing capacity of the model on the internal characteristics of the data in the training stage is remarkably enhanced, and then the prediction stability and accuracy are improved; besides, the Mamba model is adopted to predict the photovoltaic power generation power, input characteristics can be efficiently processed, intelligent screening and reservation of information are achieved, an accurate short-term photovoltaic power generation power prediction result is output, and powerful support is provided for operation and scheduling of a photovoltaic power generation system.

This invention discloses a risk classification method for power grid emergency retrieval generation based on counterfactual probing, belonging to the fields of power system emergency management and artificial intelligence technology. Addressing the problem that retrieval enhancement generation models in power grid emergency scenarios are prone to generating outdated procedures, non-authoritative information, and knowledge conflicts, this invention controls the generated content through domain-specific counterfactual probing, semantic stability judgment, and a hierarchical risk fusion mechanism. First, the initial answer generated by the retrieval enhancement generation model system based on the user query is obtained. Then, counterfactual prompts based on three dimensions—timeliness, authority, and conflict—are used to pressure the model and generate corresponding reconstructed answers. Next, the semantic stability of the initial and reconstructed answers is compared by a hierarchical judgment module, and risk identification is performed in conjunction with conflict review results. Finally, the risk fusion module classifies risks into four levels—high confidence, conflict risk, timeliness / authoritative risk, and no answer—based on priority rules, outputting the final risk label and explanatory basis. The method proposed in this invention can effectively intercept false information with high confidence, ensuring the security of power grid emergency decision-making while providing fine-grained risk explanations, avoiding the loss of key decision information caused by traditional binary risk control.

The application relates to a road network adjustment method and device. The method comprises the following steps: acquiring road network data of a target area; constructing a road network graph corresponding to the road network data according to the road network data; obtaining node embedding of all nodes and edge embedding of all edges of the road network graph according to the road network graph; obtaining an action probability distribution according to the node embedding of all nodes and the edge embedding of all edges; selecting a corresponding action according to the action probability distribution, wherein the action is an atomic operation on the road network graph; and adjusting the road network graph according to the selected action. The application abstracts a city road network into a graph structure for modeling, realizes deep understanding and reservation of spatial topological relations of the road network, understands and analyzes the road network from the perspective of overall correlation, solves the bottleneck problem that a traditional model cannot effectively process non-Euclidean space data such as the road network, and provides accurate and structured data basis for subsequent intelligent decision-making.

The invention discloses a method for measuring residual quantity of residual impurities DMAP and EDU in synthesis of antitumor drugs, and belongs to the technical field of drug analysis. According to the method, high performance liquid chromatography is adopted for detection; dimethyl sulfoxide is used as a solvent to dissolve the sample for detection; detection conditions of the high performance liquid chromatography are as follows: a chromatographic column is an octadecyl bonded silica gel column; the mobile phase A is a sodium perchlorate buffer solution or a potassium hexafluorophosphate buffer solution; the mobile phase B is acetonitrile; the detector is an ultraviolet detector; the detection wavelength for detecting the DMAP is 278 to 282 nm, and the detection wavelength for detecting the EDU is 198 to 202 nm; a gradient elution mode is adopted. The method is used for detecting DMAP and EDU residues in the raw material medicine, is high in sensitivity and accuracy, does not need a special column, can meet the control limit of the ICH guidance principle on genotoxic impurities and general impurities of the raw material medicine, provides reliable guarantee for impurity control of the raw material medicine, and further guarantees the safety of clinical medication.

The utility model provides a molecular film supporting device which comprises a plurality of micro-supporting structure units arranged on a substrate in an array mode, and each micro-supporting structure unit comprises a supporting cavity structure arranged on an upper material layer and a micro-well structure arranged on a lower material layer and corresponding to the supporting cavity structure; the supporting cavity structure comprises a pore channel structure and a plurality of fluid buffering structures which are arranged on the peripheral edge of the pore channel structure and communicated with the pore channel structure, the pore channel structure penetrates through the upper material layer, and the fluid buffering structures are arranged on the upper surface of the upper material layer but do not penetrate through the upper material layer; the molecular film support device has the technical effects of uniform film formation, higher stability and small film formation difference among the micro support structures, and solves the technical problems of poor uniformity and stability of molecular film formation caused by unreasonable structure of the support device in the prior art.

This invention relates to the field of speech data processing technology, and discloses a data processing method and system based on a language function detection system. The method includes: acquiring raw detection audio streams and task event logs through the language function detection system; constructing a task-level unified time index sequence; constructing a dynamic silence threshold; performing task-level coarse segmentation on the raw detection audio stream; constructing effective response scores; identifying effective response segments and secondary correction segments; reconstructing task-level structured results; and outputting aphasia language ability assessment data. Compared to existing technologies that primarily process entire recordings uniformly, especially in aphasia rehabilitation assessment scenarios with long pauses, repetitive pronunciations, and intrusive prompts, this invention addresses the technical problem of accurately extracting effective response segments. By using task event-driven temporal alignment and dynamic silence threshold segmentation, this method improves the reliability of aphasia language function assessment data.

The invention provides a time sequence data completion method and system, and the method comprises the steps: collecting time sequence correlation operation data of a DC converter transformer, and constructing an original time sequence defect data set X0; performing time correlation preprocessing on the original time sequence defect data set X0 to obtain a time correlation preprocessing matrix X1; inputting the time correlation preprocessing matrix X1 into a CNN network for data sequence integration to obtain an integrated matrix X2; the integrated matrix X2 is sent to a bidirectional long-short-term memory neural network Bi-LSTM for time correlation prediction learning, and a time correlation prediction result X3 is obtained; the time correlation prediction result X3 is sent to an anti-attention layer anti-attention to reduce the attention of the pre-filled data to obtain a reconstructed attention matrix X4; and putting the reconstructed attention matrix X4 into a full-connection output layer to obtain a data prediction completion result # imgabs0 #. The technical problems of gradient disappearance, overfitting and low accuracy of complemented data are solved.

The application discloses a controllable stripping and plating method for a diamond-like carbon coating on a workpiece surface, and is aimed at processing a workpiece with a diamond-like carbon coating on the surface, and comprises the following steps: adopting oxygen plasma to bombard the surface of the workpiece after pretreatment, completely stripping carbon substances on the surface layer of the workpiece, and converting part or all of a transition layer deposited in gas phase into a metal carbon / oxide compound; then immersing the workpiece after the oxygen plasma bombardment in an alkaline immersion liquid A and an alkaline immersion liquid B in sequence, completely stripping the metal carbon / oxide compound remaining on the surface of the workpiece, and performing chemical polishing on exposed metal substances. The application does not damage the water plating metal layer, the stripping and plating quality is easy to control, and is beneficial to industrial production.

The invention discloses a bone tissue regeneration and repair module based on a direct-connection electrode cell-loaded biological material and an in-vitro electrical stimulation sterile culture mold and a preparation method of the bone tissue regeneration and repair module. The preparation method comprises the following steps: (1) obtaining the size parameters of the initial bone tissue regeneration and repair module and the in-vitro electrical stimulation culture sterile mold as required; (2) preparing an in-vitro electrical stimulation culture sterile mold based on the parameters, and respectively arranging double gates on two sides of the mold for connecting an electrode slice with an electrical stimulation power supply and ensuring an internal sterile environment; the upper end of the mold is open and is provided with a cover plate for separation. And 3) preparing a cell-carrying biological material directly connected with the conductive slice electrode, and externally connecting an electric field applying device through the conductive slice electrode. And 4) applying a bionic pulse direct-current electric field to the bone tissue, and performing in-vitro culture to obtain the bone tissue regeneration and repair module. The in-vitro incubation of the bone tissue regeneration and repair module is realized through the cooperation of the direct-connection electrode cell-loaded biological material and the electrical stimulation culture mold.

This application provides a data management method, apparatus, electronic device, and storage medium. The data management method includes: initializing N browser windows to cache target data from a target storage table into a first variable in each browser window; creating communication links between the N browser windows, where N is greater than or equal to 2; the target storage table is a storage table in the IndexedDB database corresponding to the browser; and the target data is an object structure including data keys and data values. When the target browser window triggers an update of the target data value corresponding to the target data key in the target storage table, the content in the first variable corresponding to the N browser windows is updated. When the array set includes the target data key, an update is triggered in the browser window corresponding to the target data key. This application can achieve unified data management across multiple browser windows, updating the corresponding browser window based on changes in the target data value.

The present disclosure discloses an electric-quantity-based path planning method for an electric vehicle compatible with an energy storage charging pile. The method includes the following steps: 1) charging piles in a whole network submitting state information of the piles to a dispatching system platform, and before departure, an electric vehicle applying for a reservation from the dispatching system platform; 2) finding a shortest path from a departure place to a destination; 3) the dispatching system platform judging whether the reserved allocated electric quantity of the electric vehicle at one or more charging piles along the path meets an endurance requirement for driving to the destination; 4) if so, generating a whole-journey reservation scheme; and 5) if not, excluding the shortest path, and finding the shortest path from the remaining paths, and performing cyclic execution from Step 3) until all paths are traversed. By means of the method of the present disclosure, whole-journey electric-quantity-based path planning for an electric vehicle can be realized, the fused use of an energy storage charging pile and a power-grid direct-supply charging pile is realized using arrival time matching, and charging scheduling is also performed on the vehicle through the dispatching system platform, such that the electric quantity of the charging pile resources, especially an energy storage pile, in a whole network is fully utilized.

The utility model discloses a wooden door convenient to assemble, which comprises an upper door body and a lower door body, the upper door body and the lower door body are connected through a mounting mechanism, two ends of the mounting mechanism are provided with limiting mechanisms, the lower door body is provided with a door lock device, and the bottom end of the lower door body is provided with a plugging assembly through an adjusting assembly. The gap at the bottom end of the door body can be adjusted according to needs, the gap can be reserved and blocked, the sound insulation effect can be improved during blocking, privacy is high, meanwhile, heat exchange with the outside can be reduced when an air conditioner is turned on, the refrigerating or heating effect is ensured, the door body can be conveniently spliced and assembled, and the service life of the door body is prolonged. The door body can be split during carrying and carrying, carrying operation is convenient, splicing and fixing can be conducted during assembling, normal use of the door body is not affected, and flexibility and convenience are achieved.

The invention discloses a preparation method of soluble soybean polysaccharide. The method comprises the following steps: dissolving soybean phospholipid in n-hexane, adding 50-60% ethyl acetate-water solution, stirring, extracting, standing for layering, taking the lower ethyl acetate-water phase, concentrating and recovering ethyl acetate, concentrating the residual water phase through a ceramic membrane, intercepting macromolecular impurities, diluting the intercepted phase with water, continuously concentrating, and concentrating the permeate through a nanofiltration membrane to remove small molecular impurities, thereby obtaining the soybean phospholipid. And carrying out spray drying on a trapped phase of the nanofiltration membrane to obtain a product. The preparation method has the following beneficial effects: soybean lecithin is used as a raw material, so that the source of soluble soybean polysaccharide is expanded; through a specific membrane separation process, the purity of the soluble soybean polysaccharide is greater than or equal to 85%, and the net extraction rate is greater than or equal to 95%; the production process is simple, and large-scale industrial production is easy to realize.

The invention relates to a photovoltaic power generation power prediction method based on a stacked state frequency memory network and transfer learning. Compared with the prior art, the photovoltaic power generation power prediction method overcomes the defects that a photovoltaic power generation power prediction method is insufficient in frequency domain feature utilization, insufficient in robustness under data uncertainty and low in prediction precision when new site data is insufficient. The method comprises the following steps: data acquisition and preprocessing; constructing a three-layer stacked state frequency memory network; training a source site model and adjusting and optimizing parameters; target site migration and multi-strategy fine tuning; and predicting photovoltaic power generation power. According to the invention, controllable retention and forgetting of multi-frequency components are realized through time domain-frequency domain coupling and gating memory updating at a model structure level, and multi-scale characterization capability is enhanced through three-layer stacking and residual fusion at a network depth level. And the influence of insufficient new site data on the prediction precision is reduced through source site pre-training and target site multi-strategy fine tuning on the engineering application level, so that the short-term photovoltaic power prediction precision and the system robustness are improved.