37results about How to "Solve the problem of inaccurate forecasts" patented technology

The invention provides a continental facies clay shale reservoir stratum gas content prediction method and device and relates to the technical field of natural gas exploration. The method comprises the steps that logging data and a core sample of the continental facies clay shale reservoir stratum are obtained, and relevant parameters of the continental facies clay shale reservoir stratum are determined; then the total organic carbon volume fraction, the effective porosity, the water saturation, the continental facies clay shale reservoir stratum temperature and the pore pressure of the continental facies clay shale reservoir stratum are determined; then the adsorbed gas quantity of the continental facies clay shale reservoir stratum is determined according to an adsorbed gas quantity prediction model of the continental facies clay shale reservoir stratum; the free gas quantity of the continental facies clay shale reservoir stratum is determined according to a free gas quantity prediction model of the continental facies clay shale reservoir stratum; and accordingly the total gas content of the continental facies clay shale reservoir stratum is determined according to the adsorbed gas quantity of the continental facies clay shale reservoir stratum and the free gas quantity of the continental facies clay shale reservoir stratum. By the adoption of the continental facies clay shale reservoir stratum gas content prediction method and device, the problem that in the prior art, various factors affecting the adsorption capacity of shale are not sufficiently considered, and accordingly the current continental facies clay shale reservoir stratum gas content prediction is inaccurate can be solved.

The invention provides a data classification method and device, equipment and a computer readable storage medium. According to the data classification method and device, the equipment and the computerreadable storage medium provided by the invention, all minority sample sets are mixed to form a new sample set which serves as training samples to perform machine learning together with majority sample sets so as to obtain a first classification model, so that a problem of imbalance in sample quantity of the minority sample sets and the majority sample sets is solved; and all of the minority sample sets serve as new samples to perform training again so as to obtain a second classification model, the type of the data is predicted through combining the first classification model and the secondclassification model, and a problem of inaccurate prediction for the minority data is solved.

A trajectory prediction method, applied to automatic driving in the field of artificial intelligence, includes obtaining posture information and historical motion trajectory information of a target object (201); obtaining target information representing the behavioral intention of the target object based on the posture information (202) using the target information and the historical trajectory information as input to acquire the predicted trajectory of the target object by a trajectory prediction model (203). The method takes the behavior intention of the pedestrian as the input feature of the pedestrian trajectory prediction, which can reduce the prediction error of the pedestrian trajectory and improve the prediction performance of the pedestrian trajectory.

The invention discloses a loop instruction processing method, device and equipment and a storage medium, and relates to the field of voice and chips. According to the specific implementation scheme, acomputer program is obtained, the computer program comprises a first loop body, the first loop body is generated according to a second loop body in to-be-compiled software codes, the first loop bodycomprises a plurality of first loop instructions, and the first loop instructions can be recognized by a hardware structure of computer equipment; under the condition that the first loop body is detected, loop parameters of the first loop body are determined according to the multiple first loop instructions; a plurality of first loop instructions are obtained according to the loop parameters of the first loop body; and the plurality of first loop instructions are executed. According to the embodiment of the invention, the first loop body is generated according to the second loop body in the to-be-compiled software code, and the first loop instruction included in the first loop body can be identified by the hardware structure of the computer equipment, so that the hardware structure can directly and accurately determine the loop parameter according to the first loop instruction.

The invention provides a flow prediction model training method and device, a flow prediction method and device, equipment and a medium. The flow prediction model training method comprises the steps: obtaining a flow data set, intercepting the flow data set through a sliding window operation, and obtaining the sample data features of the flow of the size of a sliding window and the sample labels corresponding to the sample data features of the flow; obtaining a next flow value after the sample data characteristics of the sample label flow; obtaining sample attribute characteristics corresponding to the sample label; processing the sample data characteristics of the flow by using a gated recurrent neural network of the model, and inputting the output memory state and the sample attribute characteristics into a full-connection network of the model for regression prediction; and according to the sample label and the prediction result, updating the parameter weight through a back propagation algorithm to obtain a trained flow prediction model. Therefore, a prediction result is associated with the attributes of different flows, and the problems of hysteresis quality and low accuracy of flow peak prediction are solved.

The embodiment of the invention provides a rock mass instability stage prediction method and device based on multi-feature fusion, and belongs to the technical field of rock mass instability prediction. The method comprises the following steps: acquiring an acoustic emission signal of a rock mass; carrying out mixed domain feature extraction on the obtained acoustic emission signal; obtaining a prediction error according to each characteristic parameter and a preset BP neural network model; establishing a D-S evidence theory identification framework, wherein the identification framework comprises a plurality of identification sub-frameworks in one-to-one correspondence with the rock mass instability stages; constructing a basic probability distribution function of the characteristic parameters according to the prediction error, and calculating trust values of the characteristic parameters and each stage of rock mass instability; and according to the obtained trust degree value, fusingall the feature parameters under the same recognition sub-framework through a D-S evidence theory, and performing rock mass instability stage prediction based on the fused feature parameters. Accurateprediction of each stage of rock mass instability is realized, and the problem of inaccurate prediction of single characteristic parameters is effectively solved.

The invention discloses a transformer noise calculation method and system based on a multi-line sound source model. The method comprises the steps that the relative position of a point to be predictedand a transformer is determined; according to the relative positions of the to-be-predicted points, selecting one or more multi-line sound source equivalent models corresponding to the side surfacesof the transformer having influence on the to-be-predicted points; inputting the position of a to-be-predicted point into the one or more multi-line sound source equivalent models to obtain noise of one or more affected transformer side surfaces at the position of the predicted point; superposing the one or more noises at the position of the prediction point to obtain comprehensive noise at the position of the point to be predicted; the method and the system are divided into a plurality of small cube modules through a finite element method, the small cube modules are equivalent to a pluralityof corresponding line sound sources, and a multi-line sound source model is established through a finite element-boundary element coupling method; noise generated by the transformer can be accuratelypredicted at different prediction positions, and the problem that noise prediction in a near-field area is inaccurate is solved.

The invention relates to an evaluation method, in particular to an evaluation method of single coal blending coking efficiency. First, single coal coke forming tests are performed as follows: performing coke oven testing on each kind of single coal, and simulating the actual coking production state; after coke is refined, determining experimental results including the mechanical strength and the post-reaction strength of the single coal coke; establishing a same coal test database, and comparing and analyzing data; then a database of the mechanical strength and the post-reaction strength of the single coal coke and blended coal refining coke is established as follows: testing or producing blended coal, and establishing a machine account; establishing a single coal and blended coal refiningcoke corresponding database; and calculating the mass of the single coal and data of the blended coal refining coke; and coke quality is predicted as follows: calculating the mechanical strength andthe post-reaction strength of the coke by use of a coal blending scheme and the mechanical strength and the post-reaction strength of the single coal for addition. The single coal blending efficiencyis improved, stable coke quality is ensured, and economic efficiency is improved.

The invention belongs to the technical field of iron and steel production, in particular to a method for predicting quality indexes of metallurgical coke, which includes: establishing a coke quality data collecting production data; establishing a data correspondence; screening database data; establishing a regression equation of coke quality, and obtaining the partial regression coefficientsm 1, m 2, m 3,... mn and constant b with coke quality as dependent variable y, coal quality index and coal test index, and single coal quality as independent variable x. The application of the prediction equation of coke quality index, the blended coal quality index corresponding to the coke quality index, the single coal quality index and the coal test index can be added or updated. If the data area selected by the LINEST function formula is adjusted, the formula will be updated. According to the actual production situation, the coke quality reacted strength CS or M40 and its related data will be selected. The accuracy of coke quality index prediction is greatly improved due to its fast calculation, convenient update and flexible use.

The invention provides a method and a system for predicting the condensation performance of an umbrella-shaped flexible tensioned membrane condenser. The prediction method comprises the following steps: establishing a triangular mesh model of the umbrella-shaped flexible tensioned membrane condenser; carrying out shape finding design on a reflective film of the umbrella-shaped flexible tensioned membrane condenser to obtain geometric position coordinates of each triangular grid in the reflective film of the umbrella-shaped flexible tensioned membrane condenser after shape finding; randomly generating incident points of light rays on the reflective film, calculating a reflection equation of the incident light rays after being reflected by the film by utilizing a Fresnel principle and combining sunlight non-parallelism so as to obtain reflected light rays, and obtaining the position of the light rays falling on a receiving surface after being reflected by the reflective film of the umbrella-shaped flexible tensioned membrane condenser; and predicting the condensation performance of the umbrella-shaped flexible tensioned membrane condenser according to the reflected light received by the receiving surface. According to the technical scheme provided by the invention, the problem of inaccurate evaluation of the condensation performance of the umbrella-shaped flexible tensioned membrane condenser can be solved.

The invention provides a prediction method for original rock temperatures of a high-ground-temperature tunnel under the unexcavated condition. According to the prediction method, a heating conductingeffect of different elevations and different ground surface temperatures, which are formed by geological evolution, a ground temperature gradient of stratum rocks and a rock temperature difference ofdifferent points is utilized; the original rock temperature of each measured point in a corresponding range of the high-ground-temperature tunnel is relatively accurately calculated under the assistance of two independent development formulas; a problem of inaccurate prediction of a conventional method is solved; by an original rock temperature distribution diagram, heat damage prevention measurescan be taken in advance; the labor hygiene requirement of an operator is met; technical support is provided for heat damage governance and scientific construction of the high-ground-temperature tunnel; and safe and rapid construction of the high-ground-temperature tunnel is ensured.

The invention provides a time advance prediction method and device and a communication system, and relates to the technical field of communication, and the method comprises the steps: obtaining a plurality of pieces of measurement report (MR) data sent by at least one terminal in a terminal set to a base station; grouping the multiple pieces of MR data according to the user session ID to obtain multiple groups of MR data; sorting the multiple pieces of MR data in each group of MR data according to the sending time of each piece of MR data; after sorting, for one piece of MR data missing the timing advance in one group of MR data, determining whether another piece of MR data and another piece of MR data exist before and after the piece of MR data respectively, wherein the other piece of MR data and the another piece of MR data do not missing the timing advance; and if the other piece of MR data and the other piece of MR data exist, predicting the missing time advance of one piece of MR data by adopting a linear interpolation method.

The invention discloses a crack detection method based on Alford rotation. The method includes: step 1, obtaining radial components and tangential components of multi-component seismic data, wherein the radial component is a component along the direction connecting the shot point and the receiver point, and the tangential component is a component perpendicular to the direction along the line The component of the direction of the line between the shot point and the receiver point; step 2, based on the radial component and the tangential component, using Alford rotation to scan according to the rotation angle and time difference to obtain the correlation coefficient spectrum of the fast and slow waves; step 3, based on the The correlation spectrum determines the development information of subsurface fractures. The application of the present invention can effectively solve the problem of inaccurate prediction in the case of large angle intervals in conventional fracture prediction, and has positive effects on subsequent elimination of anisotropic influence and improvement of imaging quality.

The invention discloses a behavior perception recommendation method, a terminal and a computer readable storage medium, and the method comprises the steps: obtaining attention scores of a next behavior and a current behavior of a target object, and adding the attention scores as heterogeneous behavior information into an input layer of a perception recommendation model; introducing a target behavior at the output side of the perception recommendation model, and training the perception recommendation model according to the target behavior to obtain a trained perception recommendation model; and predicting a target behavior of the target object according to the trained perception recommendation model to obtain a next behavior prediction result corresponding to the target behavior, and recommending the obtained prediction result to a client where the target object is located. According to the method, the attention mechanism is applied to the input layer, and the attention score between the next behavior and the current behavior is used as heterogeneous behavior information, so that the problem that the isomerism cannot be distinguished is solved; meanwhile, a real next behavior is introduced to the output side, so that the problem of inaccurate isomerism prediction is solved.

The invention provides a CT image data processing method and device and a storage medium. The method includes: acquiring multiple CT images each including a group of layer images of a human body part;selecting a preset number of layer images arranged at intervals from the layer images of the CT images as training images of the CT images, and synthesizing the training images; inputting the synthesized image into multiple pre-established network models separately for training; fusing the respective trained network models to obtain a fused network model; training the fused network model; inputting the target CT image into the trained fused network model, and outputting a pathological prediction result of the target CT image. The method selects multiple training images from different positions in the respective CT images, synthesizes the training images to obtain a synthesized image, trains the basic network models by the synthesized image, improves the accuracy of network training, and integrates the trained network models to obtain a fused network models, thereby improving robustness and prediction accuracy.

The invention discloses a predicting method of reserve fund and a device, wherein the method comprises the following steps of: predicting the loan amount of inventory business of each subsidiary autodealer of an auto financing company, and acquiring the fund predicting amount of the inventory business of the auto financing company by statistics of the predicted loan amount of inventory business needed by each subsidiary auto dealer; predicting the loan amount of retail business needed by each auto dealer, and acquiring the fund predicting amount of the retail business of the auto financing company by statistics of the predicted loan amount of retail business needed by each subsidiary auto dealer; and acquiring the reserve fund predicting amount of the auto financing company according to the fund predicting amount of the inventory business and the fund predicting amount of the retail business of the auto financing company. The technical scheme solves the problem that the predication ofreserve fund in the prior art is not accurate, and realizes the purpose of carrying out prediction to the reserve fund so as to lead the fund to be allocated reasonably by taking reliable and large amount of historical data as a basis.

The invention discloses a flue gas oxygen content load prediction method and device based on sample migration. The method comprises the following steps: acquiring equipment data of a first participant and a second participant under a joint learning architecture; wherein the first participant is a participant who puts forward a prediction demand, and the second participant is other participants except the first participant; training a prediction classifier by using the equipment data of the first participant and the equipment data of the second participant; according to the prediction classifier, determining weight data of the equipment data of the first participant about the equipment data of the second participant; based on the equipment data and the weight data of the second participant, training a predictive gradient lifting model; and utilizing the predictive gradient lifting model to predict the flue gas oxygen content load of the first participant equipment. According to the method, the problem of inaccurate flue gas oxygen content load prediction caused by data distribution difference of energy equipment generated under different processes is solved, and the cost of an energy equipment sensor is saved.