117 results about "Stochastic differential equation" patented technology

The invention relates to a new energy power generation prediction error correction method based on artificial intelligence. The method comprises the following steps: data collection and processing; predicting by using a plurality of machine learning models; performing model integration through variational Bayesian inference to generate a final integrated prediction result; carrying out dynamic error correction by adopting stochastic differential equation modeling and extended Kalman filtering; modeling relevance between power station points by using a graph neural network, and taking the relevance as auxiliary information for error correction; self-adaptive parameter adjustment based on a feedback mechanism; global optimization and final prediction result output; according to the method, by integrating various machine learning models, implementing dynamic error correction, modeling site relevance by using the graph neural network, adaptively adjusting model parameters and executing global optimization, the accuracy, stability and adaptability of new energy power generation prediction are remarkably improved; the method has the advantages that multiple models are integrated, the dynamic error correction capability is achieved, and the prediction accuracy is improved.

The invention discloses an electric power information operation violation risk supervision system based on a knowledge graph, which relates to the field of violation risk supervision and comprises a dynamic graph construction module, a causal analysis module, a strategy analysis module, a strategy modeling module and a supervision decision module. The method comprises the following steps: obtaining multi-type electric power operation field sensing data and information system data, and carrying out data processing and knowledge graph dynamic construction to obtain a dynamic knowledge graph; performing risk situation quantification and risk decision point inference based on the dynamic knowledge graph to obtain a key causal decision point set; based on the key causal decision point set, through strategy logic analysis, obtaining a logic rule set which can be directly deployed and executed; according to the method, a logic rule set which can be directly deployed and executed is subjected to dynamic strategy evolution of a stochastic differential equation to obtain a dynamic strategy model, and the dynamic strategy model is subjected to decision optimal screening of measurement transformation to obtain an optimal supervision decision set, so that a risk value can be accurately calculated, and the supervision response speed can be increased.

The invention relates to the technical field of multi-modal sensing, and discloses a multi-modal large model detection and recognition robot recognition system for a complex scene, the system constructs a dynamic manifold modeling module, realizes cross-modal joint denoising through a stochastic differential equation and depth score matching, constructs a drift term and an anisotropic diffusion term by using an optical flow field, and realizes multi-modal detection and recognition of a multi-modal large model. Dynamic noise interference such as rain fog and motion blur is eliminated; on the basis, designing an information geometric alignment module, and based on Riemannian manifold optimization and orthogonal projection matrix calculation, realizing geometric equidistant mapping of vision-Li DAR features through multi-scale measurement tensor fusion; a dynamic external parameter calibration module is further provided, SE (3) manifold Kalman filtering is combined with a noise self-adaptive scaling technology, and external parameter offset is tracked and compensated in real time. Compared with a traditional method, the method has the advantages that the core problems of cross-modal data geometric mismatch, external parameter drift accumulation, low semantic fusion efficiency and the like are solved, and the sensing precision and robustness of the automatic driving system in a complex dynamic scene are remarkably improved.

The invention belongs to the technical field of speech enhancement, and discloses a Schrodinger bridge-based diffusion model speech enhancement method and system, and the method comprises the steps: converting a diffusion process of a diffusion model into a solving process of a stochastic differential equation, and determining the stochastic differential equation according to the theoretical principle of a Schrodinger bridge, and the complex spectrum is directly used as the input of the diffusion model, so that the tedious work of extracting a phase spectrum and an amplitude spectrum from the complex spectrum and the memory overhead caused by inverse transformation are omitted, and meanwhile, the alignment problem between the phase spectrum and the amplitude spectrum is also avoided. According to the speech enhancement method, unique features of time sequence signals are captured through a Transformer module, multi-scale information is fused through a U-Net module, a loss function covering a time domain, a frequency domain and a time-frequency domain is matched, the difference between a prediction sample and a clean sample is gradually reduced, the nonlinear diffusion process from a noisy sample to the clean sample can be directly learned, and the speech enhancement accuracy is improved. Therefore, the structure information of more initial samples is reserved.

The invention relates to the field of transaction anti-fraud, and discloses a real-time transaction anti-fraud system based on a multi-modal behavior atlas, and the system comprises an acquisition processing module which is used for obtaining time sequence data of equipment, transaction and geographic modals and carrying out the preprocessing of the time sequence data, and obtaining the processed data; the multi-mode construction module is used for calculating the time-varying fluctuation rate of a transaction mode, the local dispersion degree of an equipment mode and the trajectory bending degree of a geographic mode, and constructing the processing data into random manifold features containing a Riemannian metric tensor; and the cross-modal interaction module is used for modeling a time-varying driving relation between random manifold characteristics based on a stochastic differential equation and a Poisson process. The method comprises the following steps: converting equipment, transaction and geographic data into random manifold features containing Riemannian metric tensor, reserving nonlinear time sequence association of multi-modal data, and describing a random drive of'equipment motion-geographic trajectory 'and a time-varying trigger relationship of'transaction-equipment activity' through a stochastic differential equation and a Poisson process.

The invention relates to the technical field of intelligent decision making, can be applied to business system platforms of financial science and technology, medical health and the like, and discloses a speech synthesis method, device, equipment and medium based on an optimization strategy algorithm, comprising: extracting a time sequence processing network unit and a data sampling scheduling unit in a speech synthesis model; mapping a denoising function in the time sequence processing network unit into a multi-step Markov decision function, and converting an ordinary differential equation in the data sampling scheduling unit into a multi-source stochastic differential equation; sampling multiple groups of independent audio tracks corresponding to the input text based on a multi-source stochastic differential equation; calculating a strategy gradient modulation factor by using an optimization strategy algorithm and a multi-step Markov decision function; optimizing strategy parameters of the speech synthesis model according to the strategy gradient modulation factor to obtain an optimized speech synthesis model; and obtaining a to-be-converted text, and synthesizing voice corresponding to the to-be-converted text by using the optimized voice synthesis model. And the speech synthesis accuracy is improved.

The invention discloses an engine state estimation and system modeling correction method based on double-layer variational inference, which relates to the field of engine state estimation and comprises a variational inference stage aiming at component performance states and kinetic model parameters; a system output prediction stage based on an observation equation; and a solving stage of performing objective function optimization through an evidence lower bound. The structure clearly presents information flow and key calculation links of the proposed algorithm in state estimation and model learning. According to the method, combined reasoning of state variables and model parameters is achieved by building a probability modeling structure, the modeling problem when system dynamics is partially or completely unknown is solved by combining a modeling method of a stochastic differential equation, and while the state variables and the model parameters are optimized, the modeling efficiency is improved. Precise inference of component states and reliable identification of fault features are achieved, the fault detection accuracy of sudden gas circuit abnormity reaches the standard, and meanwhile the performance is better in the aspect of tracking long-term performance degradation.

The invention discloses a fault identification method for generating vibration data based on frequency domain transformation under a finite sample. The method comprises the following steps: collecting multiple types of fault vibration signals and extracting frequency domain spectrum characteristics; a forward diffusion stochastic differential equation based on mirror image Brownian motion is constructed, and original spectral distribution is converted into Gaussian prior distribution through progressive noise injection; training a gradient network by adopting a fractional matching criterion, and realizing spectrum feature reconstruction by combining a reverse stochastic differential equation and a Langevin dynamic sampling strategy; and the generated sample and original data are spliced and then input into a spectrum feature lexical element Transform model, local resonance frequency band feature extraction is enhanced through a multi-head attention mechanism, and finally fault identification is realized through coding, decoding and linear classification. According to the method, the problem of low fault identification precision of the rotating machinery in a data scarce scene is effectively solved, and a high-reliability data enhancement and feature decoupling scheme is provided for an industrial equipment intelligent diagnosis system.

The invention discloses a unified underlying vision pre-training method based on a multi-scale diffusion model, and the method comprises the steps: constructing and training a multi-scale degradation robust variational auto-encoder (VAE) which is used for extracting the multi-scale hidden space representation of degradation robustness; training a degradation invariant image feature encoder for extracting visual semantic features irrelevant to degradation types; based on a pre-trained diffusion model backbone network, in combination with the robust hidden space representation and the visual semantic features, constructing and training a condition-controllable bottom layer visual diffusion model, and modeling a diffusion process by adopting an improved stochastic differential equation (SDE); and integrating the multi-scale degradation robust VAE with a bottom-layer visual diffusion model, and constructing a large multi-scale bottom-layer visual pre-training model. According to the method, unified pre-training of various underlying degeneration is realized, and the fidelity and robustness of image restoration are improved.

The invention relates to the technical field of medical image enhancement, and discloses a time-driven Transform diffusion model-based photoacoustic image enhancement method, which comprises the following steps of: constructing a diffusion model based on a stochastic differential equation; based on the forward process of a diffusion model, noise is added into the photoacoustic image reconstructed under the non-ideal condition, and the photoacoustic image reconstructed under the non-ideal condition is converted into a known prior distribution image; the prior information is continuously updated through the reverse process of the diffusion model, and the noise is predicted by using a time-related network, so that the training of the diffusion model is realized; sampling the noise state of the photoacoustic image reconstructed under the non-ideal condition through the trained diffusion model, and iteratively solving a reverse stochastic differential equation by using a numerical scheme so as to enhance the photoacoustic image reconstructed under the non-ideal condition to obtain an enhanced photoacoustic image; distortion and artifacts in the photoacoustic image can be suppressed, and detail information of the photoacoustic image can be recovered to the greatest extent.

Disclosed is an approach for learning probability distributions as differentiable quantum circuits (DQC) that enable efficient quantum generative modelling (QGM) and synthetic data generation. A method includes training of a differentiable quantum circuits (DCQ) based model, where data is encoded in a latent space with a phase feature map, followed by a variational quantum circuit. The trained model is then mapped to the bit basis using a fixed unitary transformation, coinciding with a quantum Fourier transform circuit in the simplest case. This allows fast sampling from parametrized distributions using a single-shot readout. Simplified latent space training provides models that are automatically differentiable. Samples from propagated stochastic differential equations (SDEs) can be accessed by solving a stationary Fokker-Planck equation and time-dependent Kolmogorov backward equation on a quantum computer. A route to multidimensional generative modelling is opened with qubit registers explicitly correlated via a (fixed) entangling layer.

The invention relates to the field of ship route optimization, discloses a route optimization method and system based on a particle swarm optimization algorithm and geographic information, and is used for providing comprehensive decision support covering route planning, energy scheduling, risk control and carbon emission management for shipping enterprises. Comprising the steps that a commercial decision basic data set integrating geographic information, ship multi-energy system configuration and marine environment forecast data is constructed, and a commercial fractional order parameter set is generated; environment uncertainty is modeled by using a stochastic differential equation, a multi-energy system operation cost function is integrated, and a commercial cost prediction model capable of outputting operation cost probability distribution is formed; and carrying out risk-aware business strategy optimization based on a particle swarm optimization algorithm, and finally generating an executable business decision proposal containing a detailed route trajectory, a speed plan and a multi-energy equipment scheduling strategy. According to the invention, collaborative optimization of route planning and energy management is realized, and economy, safety and environmental protection of ship operation are significantly improved.

Methods for training and using a machine learning diffusion model to generate graph data based on samples from a data distribution as input. The diffusion model includes one or more diffusion layers, and the graph data include node attributes and edge attributes. The training method includes a diffusion process, including a forward- and a reverse-time pass, to learn parameters of the diffusion layers, and a joint diffusion process, including solving a forward- and a reverse-time stochastic differential equation. Both equations are based on both the node and edge attributes, the reverse-time equation being additionally based on the learnt parameters of the diffusion layers. Both equations are solved for both for the node and the edge attributes simultaneously. The trained diffusion model is provided for use, in which use method the trained diffusion model repeatedly performs the reverse-time pass to obtain graph data based on input samples.

In some aspects, the present disclosure provides a method for generating a geometrical structure of a binding complex formed between a protein and a ligand. In some embodiments, the method comprises sampling an initial geometrical structure of the binding complex from a geometry prior. In some embodiments, the method comprises denoising, using a machine-learned stochastic differential equation (SDE), the initial geometrical structure to generate the geometrical structure of the binding complex.

Systems and methods described relate to the synthesis of content using generative models. In at least one embodiment, a score-based generative model can use a stochastic differential equation with critically-damped Langevin diffusion to learn to synthesize content. During a forward diffusion process, noise can be introduced into a set of auxiliary (e.g., “velocity”) values for an input image to learn a score function. This score function can be used with the stochastic differential equation during a reverse diffusion denoising process to remove noise from the image to generate a reconstructed version of the input image. A score matching objective for the critically-damped Langevin diffusion process can require only the conditional distribution learned from the velocity data. A stochastic differential equation based integrator can then allow for efficient sampling from these critically-damped Langevin diffusion models.

The invention discloses a multi-scene risk perception method based on a knowledge graph and deep learning, and the method comprises the following steps: collecting multi-source heterogeneous data, carrying out the unified standardization of the multi-source heterogeneous data into a multi-modal data vector, converting the data vector into polar coordinate representation with direction and amplitude characteristics, and carrying out the inter-modal fusion through a polar coordinate transformation network; the fusion result is matched with a preset risk knowledge graph, and a dynamic graph structure containing risk nodes and edge weights is constructed; by introducing a reverse stochastic differential equation, the state of risk evolution along with time is simulated; and analyzing the risk density distribution in different directions, generating a dynamic vector for controlling risk reasoning, dynamically activating a risk channel in an asynchronous mode, and comprehensively outputting a unified multi-scene risk perception map. According to the invention, continuous and interpretable identification and prediction of the risk state in a complex environment are realized.

The invention relates to the technical field of hydrogen energy scheduling, in particular to a hydrogen energy full-link scheduling system and method based on big data. The system comprises a hydrogen energy link data acquisition module, a hydrogen energy supply and demand prediction module, a hydrogen production storage and transportation matching module, a hydrogen energy link scheduling optimization module and a safety feedback module. Multi-dimensional operation data covering hydrogen production to terminal utilization is collected through a distributed sensor network, a supply and demand trend is predicted by adopting a stochastic differential equation and a neural random process, and a reinforcement learning optimization model is constructed in combination with link constraints to generate an initial scheduling scheme; and constructing a graph structure path planning model by using an artificial potential field algorithm to optimize full-link scheduling, and finally realizing safety evaluation and closed-loop feedback regulation and control through fuzzy logic control. According to the invention, the digital, intelligent and efficient development of the operation of each link of the hydrogen energy industry chain is promoted, and the full-link operation efficiency and intelligent level of hydrogen energy are comprehensively improved.

The invention provides a hockey team tactical position early warning and correction method and system fusing spatio-temporal features, and relates to the technical field of hockey tactical analysis, and the method comprises the steps: generating tactical feature vectors through employing a self-adaptive diffusion probability graph algorithm, and comparing the tactical feature vectors with a standard tactical library to generate an early warning; calculating position deviation according to tactical rules; a stochastic differential equation is constructed, a plurality of groups of candidate trajectories are generated, and the trajectory with the minimum energy value is selected as the correction path, so that real-time early warning and scientific correction of the hockey tactics are realized, and the tactical execution accuracy is improved.

The invention discloses a refrigerator life prediction method based on an adaptive physical information recurrent neural network, and the method comprises the steps: extracting multi-source statistical health features through a steady-state control window, capturing time sequence dependence through a recurrent neural network, and generating a degeneration state estimator; embedding the estimator into a stochastic differential equation driven by a Wiener stochastic process to enable a drift term to reflect a deterministic degradation mechanism and a diffusion term to quantify the uncertainty of a working condition; based on data deviation, equation residual and monotonicity violation, multi-constraint joint loss is constructed, Bayesian uncertainty estimation is adopted to dynamically optimize each loss weight, and adaptive balance of physical consistency and observation fitting is realized. According to the method, residual life prediction with probability distribution is output, and the precision and robustness of refrigerator health management under complex working conditions are effectively improved.

The invention provides a Schrodinger bridge-based spatial audio synthesis method and system, and the method comprises the steps: obtaining a monaural sound source signal, and constructing a prior signal and a noisy representation based on the monaural sound source signal; inputting the prior signal and the noisy representation into a pre-trained spatial audio synthesis model, and generating a final dual-track spatial audio through the spatial audio synthesis model based on a stochastic differential equation iterative sampling path; wherein the spatial audio synthesis model is obtained by training a preset neural network model through a predefined dual-channel Schrodinger bridge parameterization target and boundary auxiliary supervision. According to the invention, the problems of slow synthesis speed and poor quality of the existing spatial audio are solved.

The invention relates to the technical field of information, and discloses a data migration scheduling method based on equipment anomaly detection, which comprises the following steps of: S1, modeling through the health state of equipment, monitoring and estimating the health state of the equipment in real time, and judging whether the equipment is in an abnormal state or not based on an estimated value; s2, when the equipment is in an abnormal state, triggering data migration scheduling; s3, designing an optimization scheduling target, minimizing data migration delay, equipment load fluctuation and bandwidth resource consumption, and calculating a data migration path and a scheduling strategy based on an equipment health state and bandwidth constraints; and S4, solving an optimal scheduling problem with constraints by adopting a Lagrange multiplier method. Modeling and real-time estimation are carried out on the equipment health state through the stochastic differential equation and the Kalman filtering algorithm, accurate abnormal state detection is achieved, and the problems that in the prior art, response to equipment state changes is not timely, and accuracy is insufficient are effectively solved.

The invention relates to the field of new energy batteries, and discloses a new energy battery temperature monitoring control method comprising the following steps: deploying a multi-mode sensor array comprising a nanowire thermopile, a terahertz time-domain spectroscopy system and a flexible film temperature sensor; constructing a non-local heat conduction model, and introducing a heat flow relaxation time term and a non-local diffusion term to correct the traditional Fourier law; establishing a jump diffusion type stochastic differential equation of a microscopic heat source, and quantifying the randomness of lithium ion flux fluctuation and thermal runaway triggering; the optimal layout of the sensor is determined through a topological optimization algorithm, and the temperature field reconstruction error and the hardware cost are minimized. Through the cross-scale sensing fusion of the nanowire thermopile and the terahertz spectrum, the limitation of traditional single-point temperature measurement is broken through, and millisecond-level capture of the thermal behavior of an electrode-electrolyte interface microcell is realized. Compared with a method depending on surface temperature extrapolation in the prior art, the problem that microcosmic heat source recognition precision is insufficient is effectively solved, and the early-stage thermal runaway early warning capacity is remarkably improved.

The present invention discloses an electric vehicle energy storage charging and discharging optimization scheduling method based on a V2G feasible domain, which specifically relates to the technical field of electric vehicle energy management. The method is used to solve the problem of scheduling failure caused by the instability of energy flow interaction when existing electric vehicles are performing regenerative braking energy recovery and V2G discharge simultaneously; by obtaining regenerative braking signals and remaining power data, it is determined based on a multi-source prediction algorithm whether the initial conditions of the feasible domain are met; when the conditions are met, the wheel axle braking distribution characteristics are analyzed through dynamic modeling and stochastic differential equations, and the real-time characteristic curve of vehicle discharge and grid adaptation is identified based on grid load data; the energy flow identification result of regenerative braking and V2G discharge is determined according to the characteristic curve and dynamic change characteristics; when the identification result is in a preset coupling interval, the feasible domain constraint conditions are calculated, the output power of the drive system and the bidirectional inverter are scheduled, and the regenerative braking recovery power and the grid discharge power are dynamically allocated.

A multi-modal uncertainty perception mechanical life prediction method based on random model updating comprises the steps that firstly, a drift network is constructed, a representation extraction block RDB comprises four layers including a data structure DS, short-term local feature refinement, long-term time trend development and down-sampling, and residual connection is used for capturing cross-layer information and promoting gradient flow; a diffusion network is constructed, Lipschitz continuity is followed, and ReLU is selected as an activation function; then each subnet is trained, the first two parts of a target function of the subnet are regression terms of a distributed drift network and an adjustment diffusion network, and the latter component belongs to an OOD sample; then, the uncertainty of each subnet is quantified, and finally, multi-modal fusion is carried out based on the respective uncertainty; according to the method, modeling is carried out on the uncertainty of each mode through the fractional order stochastic differential equation sub-network, dynamic fusion is carried out according to the uncertainty of modal characteristics, and a final prediction result and the corresponding uncertainty are obtained through regression reasoning.

The invention discloses a photovoltaic power generation capacity credibility evaluation method based on joint probability modeling and a random process, and the method comprises the steps: collecting historical irradiance and temperature data of a target region, extracting an hour-level mean value and a standard deviation of the historical irradiance and temperature data, and dividing a seasonal data set; a Copula function is adopted to construct a joint probability distribution model of irradiance and temperature, a Gumbel Copula function is selected in summer, and a Clayton Copula function is selected in winter; generating a random fluctuation sequence of irradiance and temperature based on a stochastic differential equation, and calculating photovoltaic real-time output by combining a photovoltaic cell characteristic equation; a double-state Markov model is adopted to simulate traditional unit faults and load fluctuation, and system loss of load expectation (LOLE) is quantified through Monte Carlo simulation; carrying out dynamic iteration by utilizing an equivalent conventional capacity method (ECP), and calculating the capacity credibility of photovoltaic power generation; according to the invention, a reliability evaluation basis is provided for high-proportion photovoltaic grid-connected planning.

The invention relates to the technical field of carbon emission prediction and early warning, and provides a multi-scale neural distribution prediction and hierarchical migration early warning method, which comprises the following steps: acquiring historical carbon emission data, respectively inputting a historical sequence and a to-be-predicted sequence into an energy consumption stochastic differential equation model and a carbon factor stochastic differential equation model, generating a multi-scale carbon emission path sample set through an independent random disturbance term; calculating a path-level suitability score of each path sample based on the standard-exceeding risk integral, the first standard-exceeding moment and the path fluctuation variance; layering the calibration data set into a plurality of working condition layers according to working condition labels, sharing distribution shape parameters among the working condition layers through a hierarchical Bayesian method, and regularizing quantiles of small sample working condition layers to obtain an early warning threshold value of each working condition layer; and selecting a corresponding early warning threshold value according to the current working condition label to compare and trigger early warning. According to the method, the accuracy of carbon emission distribution prediction and the robustness of an early warning system are improved, and the problem that the early warning threshold value is unstable under the small sample working condition is relieved.

The invention discloses a method for evaluating lag time of influence of blast furnace ironmaking operation parameters on state parameters, which comprises the following steps of: constructing a mapping relation between time delay sequence data of operation variables and blast furnace state parameters, and processing blast furnace state parameter data by adopting modes such as wavelet filtering and difference so as to construct a drift term, a diffusion term and a random process model. And then, evaluating the operation time delay through the established stochastic differential equation simulation result. Constraint conditions are added for optimization of the model according to the blast furnace smelting characteristics and the hierarchical clustering result, and the optimization efficiency and precision are further improved. Finally, the importance of the parameters is evaluated based on the gain or SHAP algorithm, the regulation and control priorities of the operation variables are ranked, and a basis is provided for regulation and control of the blast furnace ironmaking process. The method has the advantages that the delay time of the influence of the operation variables on the state parameters of the blast furnace can be accurately evaluated by fusing the production rule of blast furnace smelting, the production cost is reduced while the quality and the efficiency of the ironmaking process are improved, and the method has a wide application prospect.

The embodiment of the disclosure provides a controllable palmprint sample generation method based on a stochastic differential equation; it relates to the field of computer vision. The method comprises constructing an unconditional diffusion model network architecture and defining a forward process and a reverse process; training an unconditional diffusion model according to the forward process and a palmprint image to obtain a noise image; according to the reverse process, the trained unconditional diffusion model is used for sampling the noise image to generate an unconditional palmprint sample; a conditional encoder is introduced into the trained unconditional diffusion model; the weights of the unconditional diffusion model are frozen, and the conditional encoder is trained according to a conditional image; when the conditional encoder loss function converges, the frozen weights are unlocked, and the unconditional diffusion model introduced with the conditional encoder is trained according to the forward process, the palmprint image and the conditional image; according to the reverse process, the trained unconditional diffusion model with the conditional encoder is used for sampling to generate a conditional palmprint sample. Thus, the palmprint sample authenticity, diversity and controllability are improved.

The invention discloses a random stability analysis method for an electromechanical composite transmission system, and belongs to the technical field of electromechanical transmission. The method comprises the following steps: acquiring operation data of the electromechanical composite transmission system, and estimating drift term parameters and diffusion term parameters of an Irat stochastic differential equation representing a random road load by adopting a maximum likelihood method; establishing a random state space model according to the current balance relationship, and judging the random stability of the model by adopting a moment stability analysis method to obtain a system stability analysis result; and performing numerical simulation according to the random state space model, verifying the accuracy of the system stability analysis result, and outputting a random stability verification conclusion. According to the method, random excitation can be processed by estimating the Italian process parameters based on the maximum likelihood method and adopting the moment stability analysis method, and the method has the advantages of being good in adaptability, high in calculation efficiency and high in stability.

The present application relates to the field of intelligent medical treatment, in particular to a tumor treatment dosing regimen Pareto optimization method based on forward-backward stochastic differential equation. The present application unifies the terminal tumor load, dosing cost and treatment risk into the same stochastic control framework, can establish the Pareto optimality condition corresponding to the nonlinear model, and further converts into the implementable closed-loop state feedback control law in the LQ case. Through the combination with the numerical example and the Pareto frontier analysis, an implementable, interpretable and scalable technical path is provided for the multi-objective individualized treatment regimen design, which can solve the problem that the existing technology is difficult to uniformly process the terminal therapeutic effect, dosing consumption and cumulative risk, and difficult to obtain the implementable closed-loop feedback strategy under the forward-backward stochastic control framework.