31 results about "Ensemble prediction" patented technology

Methods and systems for generating a multi-model ensemble of global climate simulation data from a plurality of global climate simulation models (GCMs), to be used in training a neural network (NN)-based climate forecasting model, are disclosed. The methods and systems perform steps of computing a GCM validation measure for each GCM; selecting a validated subset of the GCMs, by comparing each computed GCM validation measure to a validation threshold determined based on observational historical climate data; computing a forecast skill score for each validated GCM, based on a first forecast function; selecting a validated and skillful subset of GCMs; generating one or more candidate ensembles by combining simulation data from at least two validated and skillful GCMs; computing an ensemble forecast skill score for each candidate ensemble, based on a second forecast function; and selecting a best-scored candidate ensemble. Embodiments of the present invention enable accurate climate forecasting without the need to run new dynamical global climate simulations on supercomputers.

Methods and systems for generating a multi-model ensemble of global climate simulation data from a plurality of global climate simulation models (GCMs), to be used in training a neural network (NN)-based climate forecasting model, are disclosed. The methods and systems perform steps of computing a GCM validation measure for each GCM; selecting a validated subset of the GCMs, by comparing each computed GCM validation measure to a validation threshold determined based on observational historical climate data; computing a forecast skill score for each validated GCM, based on a first forecast function; selecting a validated and skillful subset of GCMs; generating one or more candidate ensembles by combining simulation data from at least two validated and skillful GCMs; computing an ensemble forecast skill score for each candidate ensemble, based on a second forecast function; and selecting a best-scored candidate ensemble. Embodiments of the present invention enable accurate climate forecasting without the need to run new dynamical global climate simulations on supercomputers.

A method for computing a human-machine hybrid ensemble prediction includes: receiving an individual forecasting question (IFF); classifying the IFF into one of a plurality of canonical question topics; identifying machine models associated with the canonical question topic; for each of the machine models: receiving, from one of a plurality of human participants: a first task input including a selection of sets of training data; a second task input including selections of portions of the selected sets of training data; and a third task input including model parameters to configure the machine model; training the machine model in accordance with the first, second, and third task inputs; and computing a machine model forecast based on the trained machine model; computing an aggregated forecast from machine model forecasts computed by the machine models; and sending an alert in response to determining that the aggregated forecast satisfies a threshold condition.

Systems and methods are described for generating an enhanced prediction from a 2D and 3D image-based ensemble model. In various embodiments, a computing device can be configured to obtain one or more sets of 2D and 3D images and to standardize each of the 2D and 3D images to allow for comparison and interoperability. Corresponding 2D3D image pairs can be determined from the standardized 2D and 3D pairs where the 2D and 3D images correspond based on a common attribute, such as a similar timestamp or time value. The enhanced prediction can use separate underlying 2D and 3D prediction models where the 2D and 3D images of a 2D3D pair are each input to the respective underlying 2D and 3D prediction models to generate respective 2D and 3D predict actions.

The invention discloses a wave element prediction method and system. The method comprises the following steps: acquiring wind field information; performing pre-prediction according to the wind field information to perform optimization verification on the initial model to obtain a set prediction model; performing formal prediction according to the set prediction model to obtain a prediction sequence value; checking the prediction sequence value, and determining a prediction result meeting a verification requirement; performing prediction result classification processing on the prediction results through SOM machine learning to obtain prediction model sets with different precisions; selecting a target precision prediction model from the precision prediction model set, and adding a correction value to a prediction result of the target precision prediction model as a final prediction result of the output wave element; wherein the wave element prediction result comprises but is not limited to effective wave height data, wave crest period data and wave direction data. The method is small in error and good in prediction effect, and can be widely applied to the technical field of data processing.

The invention provides a reservoir flood control risk rate prediction method based on runoff ensemble forecasting. The reservoir flood control risk rate prediction method based on runoff ensemble forecasting comprises the steps that (1) a plurality of sets of runoff forecasting processes are obtained according to runoff ensemble forecasting results obtained on the basis of a plurality of forecasting schemes; (2) a reservoir outflow threshold and a reservoir water level threshold are set, and a reservoir flood control risk event is defined; (3) the reservoir upstream flood control risk rate and the reservoir downstream flood control risk rate are predicated on the basis of the runoff forecasting processes, the reservoir outflow threshold, the reservoir water level threshold and a current reservoir flood control scheduling scheme. According to the reservoir flood control risk rate prediction method based on runoff ensemble forecasting, the reservoir flood control risk rates can be analyzed in a systemized and complete mode, the reservoir flood control risk rate prediction method can be widely applied to reservoir flood control scheduling, and the basis is provided for scientific decision making of reservoir flood control scheduling.

The invention relates to a storm scale ensemble forecast perturbation method. Observation data analysis, assimilation and numerical simulation are used as major measures; an ensemble forecast method is combined with severe convection weather forecast; and by aiming at the essential differences of the global medium-range ensemble forecast and the storm scale severe convection ensemble forecast, the final goal of building a storm scale ensemble forecast system which is applicable to various kinds of storm systems and constructs the perturbation scheme in a self-adaptive way according to the real-time developed severe convection system features is achieved. The method has the beneficial effects that through a variational assimilation ensemble, the ensemble perturbation realizes the physical and power harmony; the ensemble dispersion degree of the boundary layer mode variable is improved by a random physical harmony method; through self-adaptive selection on perturbation variables sensitive to the real-time developed storm system, the variable with the most obvious influence and the highest sensitivity on the storm development is selected for perturbation; high pertinence is realized; and the ensemble dispersion degree is also improved.

The present invention proposes a global seamless typhoon power set prediction method and system, which assimilates multivariate observation data into the pre-established climate system model through the method of growth analysis and updating IAU, so as to provide the initial field, and then perform parallel calculation on the supercomputer Generate 36-year back calculation data sets and real-time forecast data sets; according to the back calculation data sets and real-time forecast data sets, use the method of direct recognition of typhoon signals to select typhoon signals from the pre-established climate system model prediction results out; the selected typhoon signals are post-processed to generate forecast pictures and released; the global seamless typhoon power set prediction method and system of the present invention can realize global typhoon multi-time scale prediction, and the climate system model coupling process takes into account Ocean and sea ice modules to improve forecast accuracy.

The invention discloses a short-term load forecasting method of a distribution network based on multi-mode fusion, the main steps of which are as follows: 1) collecting historical load time series data X of a power network. 2) Perform STL time series decomposition on historical load time series data X. 3) Get the trend item sequence X trend LSTM neural network model with N structures, residual item sequence X remainder LSTM neural network model with N structures and integrated prediction model. 4) Obtain the period item prediction result O s . 5) Obtain forecast samples. 6) Input the forecast sample into the forecast model to obtain the trend item forecast result O t and the prediction result O of the remaining items r . 7) Integrating the prediction results of periodic items O s , Trend item prediction result O t and the prediction result O of the remaining items r , and use the integrated prediction to obtain the final prediction result. The present invention helps to improve the prediction accuracy of the model while improving the robustness and generalization ability of the load prediction model.

The invention provides a maritime distress target drift set prediction method considering feedback information. The method comprises the steps of establishing a maritime search and rescue emergency guarantee database; establishing a maritime distress target drift trajectory prediction model which comprehensively considers the combined action of wind, wave and current; performing full-dimensional disturbance on the wind field, the flow field, the wave field, the distress place, the distress time, the wind-induced drift coefficient, the wave-induced drift coefficient and the flow-induced drift coefficient, constructing set members, and performing offshore distress target drift track set prediction; on the basis of the maritime distress target drift trajectory prediction model, maritime distress target drift trajectory set prediction is carried out, a search and rescue range is further calculated, and a maritime distress target drift set prediction result is obtained; in the process of searching the distress target, performing matching analysis on the set prediction result according to feedback information received in real time; optimizing and adjusting the disturbance vector according to an analysis result, and re-carrying out offshore distress target drift trajectory set prediction until a distress target is found.

The invention provides an LSSVM pulsating wind speed prediction method based on the integration of ant colony and particle swarm, which includes the following steps: performing normalization processing; calculating the pheromone concentration of each ant; moving other ants in the ant colony to the position of the head ant. Global search; update the ant pheromone concentration at each position during the iterative process, check whether the iteration termination condition is satisfied, if not, return to the third step; otherwise, the algorithm ends and outputs the optimal parameter combination; initializes the relevant parameters of the particle swarm; The fitness value of each particle's own optimal position is compared with the fitness value of the group's optimal position; the predicted fluctuating wind speed time history spectrum is obtained. The invention has the characteristics of high optimization precision, high convergence precision, few iteration times, high success rate and the like.

The invention discloses a subseasonal-seasonal-interannual scale integrated climate model ensemble prediction system, including an initialization module, a high-resolution climate model module, an ensemble prediction module and a post-prediction processing module. The initialization module is used for downloading, extracting and processing the included Multi-source data of the atmosphere, land surface, ocean, and sea ice, using external parameters to control and realize data selection, inspection, horizontal and vertical interpolation of meteorological elements and other pre-processing, suitable for data pre-processing of all climate model operations, high resolution The climate model module is used to objectively and quantitatively predict the coupled and coordinated changes of the atmosphere, land surface, ocean, and sea ice. The ensemble prediction module is used to generate any number of ensembles that combine random disturbances of physical parameters and initial value time lag methods. Predict sample membership. The invention provides a multi-time-scale and multi-space-scale climate prediction of the atmosphere, land surface, ocean and sea ice with high resolution, high prediction accuracy and good maneuverability.

The invention relates to an integration method of multiple prediction results of electric load probability density, belonging to the technical field of power system analysis. The present invention obtains a plurality of probability densities or quantile probability prediction models through the training of three types of regression models set by multiple sets of different hyperparameters, and converts the output of the quantile prediction models into obedience A probability density model for a Gaussian distribution. Using the integrated method of probability density prediction, the optimal integrated model of probability density prediction is constructed based on the trained probability density prediction model and results, and the weights of different probability density prediction methods are determined, so that the continuous level probability loss of the final integrated prediction model is minimized. This method is finally transformed into a quadratic programming problem, and then the global optimal integration weight is quickly searched by mature commercial software, which improves the accuracy of short-term load forecasting based on probability density and reduces the operating cost of power system dispatching.

Methods, systems, and computer readable storage media are disclosed for generating joint-probabilistic ensemble forecasts for future events based on a plurality of different prediction models for the future events. For example, in one or more embodiments the disclosed system determines error values for various predictions from a plurality of different prediction models (i.e., “forecasters”) for previous events. Moreover, in one or more embodiments the system generates an error probability density function by mapping the error values to an error space and applying a kernel density estimation. Furthermore, the system can apply the error probability density function(s) to a plurality of predictions from the forecasters for a future event to generate a likelihood function and a new prediction for the future event.

The invention discloses a semi-supervised integrated real-time learning method for soft measurement of Mooney viscosity of industrial mixing rubber. The present invention aims at the problem of poor prediction performance of the traditional soft-sensing method caused by the lack of marked samples and sufficient non-marked samples in the process of industrial rubber mixing. Based on the Gaussian process regression model, combined with the real-time learning method, a diverse JITGPR sub-model is constructed. Adaptive ensemble prediction is performed on selected unlabeled samples, and high-confidence pseudo-labels are selected to augment the training sample set. Finally, the final predicted value of Mooney viscosity is obtained through the fusion of the expanded training set, diverse JITGPR sub-models and limited mixing mechanism. The invention overcomes the problems of less marked samples, sufficient non-marked samples, increased cost, and difficulty in improving product quality due to the lag in obtaining the Mooney viscosity value during the rubber mixing process, realizes the online real-time prediction of the Mooney viscosity, and effectively improves the traditional Predictive performance of soft-sensing modeling of compound Mooney viscosity.