7 results about "Explicit model" patented technology

The invention provides an explicit model prediction control system and method, and the method comprises the steps: obtaining a current yaw angle and a yaw angular velocity of a three-degree-of-freedom helicopter, and generating an error vector of the three-degree-of-freedom helicopter; determining the current yaw motion state of the three-degree-of-freedom helicopter, and determining a control state vector for explicit model prediction control according to the error vector and the yaw motion state; the state space region of the three-degree-of-freedom helicopter is calculated off line based on explicit model predictive control in advance, and a target state space partition corresponding to a state vector is identified and controlled based on the state space region; and retrieving a control law associated with the identified state space partition, and controlling the yaw motion of the three-degree-of-freedom helicopter by using a yaw control signal generated by the retrieved control law. According to the technical scheme provided by the invention, the control signal can be updated in time when the three-degree-of-freedom helicopter is subjected to yaw control.

The invention discloses a fan trigger type fault-tolerant control method fusing zotope and a fatigue load prediction model. The method comprises the following steps: firstly, constructing a quasi-linear parameter time-varying model of state-driven scheduling, setting a key dynamic parameter as a state-dependent function, and realizing smooth parameter scheduling and global closed-loop stability through a convex combination structure; secondly, designing a sliding mode interval observer based on zonotope set membership estimation, defining external disturbance and obtaining upper and lower bounds of a residual error, enhancing set estimation convergence in combination with a super-twisted sliding mode arrival law, and realizing efficient fault detection by checking whether a residual error interval contains zero or not; thirdly, designing a multi-cell expressed semi-explicit model predictive controller under the Q-LPV framework, constructing control input into a convex combination of vertex control input, and balancing real-time control performance and calculation processability by pre-calculating a generalized multi-parameter quadratic programming critical region and a control law; and meanwhile, a fatigue load prediction model is fused to realize multi-objective optimization of power tracking and load suppression.

The application provides a kind of test time self-adaptive method and device based on sparse supervision explicit modeling field variable, comprising: first, for the setting that test stage only contains untagged stream data, the exclusive field representation of each sample is explicitly constructed;Then, the dense and continuous multimodal features are extracted using a visual-language pre-training model, and the field parameter of each sample is parameterized as a learnable distribution variable;Then, a momentum-updated sparse field library is designed, which provides structured prior for the field variable through decoupling supervision mechanism;On this basis, the learned explicit field clues are injected into the downstream model, and efficient adaptation can be achieved by using only the basic entropy minimization criterion without relying on complex online optimization strategy;Finally, through the system experiment verification on multiple standard robustness benchmarks, by introducing explicit and structured field representation mechanism, the application reveals that the key of robust adaptation does not come from complex adaptation algorithm.

The invention relates to a reinforcement learning-based multi-microgrid energy system optimization regulation and control method. The method comprises the following steps of 1, performing system modeling on an EMS (Energy Management System); step 2, constructing a micro-grid hybrid series-parallel connection MPSC framework; 3, defining that power borrowing can be performed between adjacent MGs, constructing a friendly mutual-aid MG cluster, and performing power transmission borrowing between the MGs in the face of energy shortage; 4, constructing a double-layer energy management strategy; and 5, designing deep reinforcement learning of the EMS. According to the method, a global optimization task is decomposed into upper-layer strategy generation and lower-layer distributed autonomous optimization through a hierarchical reinforcement learning framework, source load uncertainty is responded in real time by utilizing the perception capability of a deep neural network, and multi-objective optimization is split into independent sub-problems by adopting a hierarchical constraint reinforcement learning framework. Hard constraints such as power balance and energy storage SOC limitation are embedded into a reward function, and explicit modeling nonlinear constraints are avoided.

The invention relates to a power system short-term load prediction method based on a topology constraint space-time diffusion model, which comprises the steps of 1, preprocessing virtual power plant data, and constructing a standardized data set, 2, constructing a space-time diagram structure model, and defining a node feature matrix X and an adjacent matrix A, the method comprises the steps of (1) obtaining a historical load sequence, (2) gradually adding Gaussian noise epsilon-N (0, I) to the historical load sequence according to a cosine scheduling strategy beta n, (4) carrying out denoising and feature extraction through UNet, (5) outputting a final load prediction result, and (6) updating network parameters through a loss function. According to the method, on one hand, the strong fitting capability of a diffusion model to a non-stationary sequence is inherited, and the sensitivity to load mutation is enhanced through a cosine noise scheduling strategy; on the other hand, the space-time coupling relation between the graph structure explicit modeling nodes is utilized, the collaborative prediction precision of regional load fluctuation is remarkably improved, and a reliable risk quantification basis is provided for the virtual power plant to participate in power market bidding and dynamic scheduling.

This invention provides an explicit-implicit fusion modeling method and system, belonging to the field of 3D modeling. The method includes acquiring modeling data, performing feature analysis on the modeling data to obtain a discrimination vector; performing data splitting according to the discrimination vector and preset splitting rules, routing the modeling data to explicit processing paths and implicit processing paths; constructing an implicit scalar field based on the data in the implicit processing path, and constructing an implicit model based on the implicit scalar field; constructing an explicit model based on the data in the explicit processing path; mapping the isosurface mesh of the implicit model to the geometric space of the explicit parametric model and performing fusion processing with the explicit model to obtain an explicit-implicit fusion 3D model. This invention completes discrimination and splitting based on multiple features of the modeling data, divides corresponding processing paths, constructs explicit and implicit models respectively, and then completes model fusion. It can adapt to modeling data with different density distributions, achieving a deep integration of two types of modeling techniques.

The invention discloses a hypergraph link prediction method based on structure perception and edge feature fusion, and the method comprises the steps: introducing edge information explicit modeling under a hypergraph frame, combining a self-adaptive sub-graph sampling mechanism, taking the difference of a local structure into consideration while modeling a high-order relation, and introducing a hypergraph modeling method with edge information as the center. According to the method, nodes in an original graph are converted into edges in a hypergraph, the edges of the original graph are converted into nodes in the hypergraph, by means of the natural high-order relation modeling capacity of a hypergraph convolutional network, the over-smoothing problem is relieved while the expression capacity of a model for a complex structure is effectively improved, and by dynamically adjusting the hop count, the over-smoothing performance of the model is improved. And the most relevant neighbor node is selected to construct a local context so as to improve the structural information capture capability of link prediction.