6 results about "Single agent" patented technology

The application discloses a virtual power plant cooperative control method based on multi-agent layered reinforcement learning, and a layered multi-agent system (MAS) processes complex tasks through hierarchical cooperation and distributed decision among agents, and specifically comprises the following steps: S1, an attribute information of a single agent is acquired by a bottom execution layer, specific equipment is directly controlled, and an inverter output power is adjusted or a load transfer instruction is executed according to a middle management layer; S2, a middle management layer receives an instruction of a top coordination layer, a sub-group is mixedly divided according to a region or a resource type, the instruction of the top coordination layer is executed, and real-time data is fed back; S3, the top coordination layer serves as a global decision center, is responsible for power market strategy formulation and resource aggregation optimization, and finally outputs a structured decision instruction and dynamic feedback; after the virtual power plant (VPP) receives the structured decision instruction and dynamic feedback output by the top coordination layer, layered processing and dynamic optimization are carried out. The application can obviously reduce delay and reduce load.

The invention discloses a multi-agent-based product research and development decision optimization method and system, and belongs to the technical field of artificial intelligence. The invention aims to solve the technical problems that an existing product aided design system adopts a single agent architecture and lacks an effective cooperation mechanism, so that an information island phenomenon is serious, and a unified product research and development scheme is difficult to form. According to the method, an assistant agent analyzes an issue and automatically calls cross-domain professional agents, each agent makes a proposal for the issue according to the own professional field, and the assistant agent organizes to perform review, screens out the proposal decided by the review, and arranges the proposal into a comprehensive decision scheme. The method is mainly used for optimizing the product research and development decision scheme.

The invention relates to the technical field of power distribution network dispatching and virtual power plant cooperative control, in particular to a power distribution network load recovery method based on a multi-agent coordination mechanism under a virtual power plant, which comprises the following steps of: modeling adjustable resources of the virtual power plant to quantify power regulation potential, abstracting a load recovery process into a Markov decision process, and extracting the load recovery process into the Markov decision process; then configuring three types of agents, namely a switch control agent, a virtual power plant coordination agent and a global coordination agent, and finally realizing multi-agent training and execution; according to the method, the power supporting potential of the virtual power plant can be fully released, the problems that a single agent is slow in convergence and prone to sinking into local optimum, and multi-agent mixed action space adaptation defects exist are solved, the operation safety of the power grid is guaranteed, and the flexibility, efficiency and safety of load recovery after the power distribution network breaks down are improved.

The invention relates to the technical field of intelligent decision, and discloses a PPO-based multi-agent cooperation mechanism generation method, device, equipment and medium, and the method comprises the steps: generating a single-agent output action through PPO, an independent agent and a current state vector, and calculating a single-agent advantage value according to GAE; and generating an agent cooperation mechanism according to the negative extrinsic punishment mechanism, the priority rule, the single agent output action and the single agent advantage value. Through the above mode, the heterogeneous data is converted into the unified state vector, the single-agent action is generated through the PPO algorithm, and the single-agent advantage value is calculated through the GAE algorithm, so that the problems of training oscillation and convergence difficulty of multiple agents are inhibited, and through a penalty mechanism and a priority rule, the multi-agent training efficiency is improved. The problem of deviation between local optimum and global target of multiple agents is solved. The method can be applied to the business fields such as financial science and technology and medical health care, and the reliability of parallel processing of multiple businesses by the intelligent customer service system is improved.

The application relates to the technical field of intelligent computing centers, in particular to an agent context optimization method for an intelligent computing center, which comprises the following steps: step 1), system initialization; step 2), task receiving and disassembly; 3), intelligent routing and selective injection; step 4), subtask execution; step 5), result aggregation and feedback. Through MCP capability boundary binding and a selective injection mechanism, the MCP tool is changed from "full injection" to "precise injection", the Token occupation of a single agent is greatly reduced, the effective working space of the model is released, and the problem that Token resources are excessively squeezed is solved; the context of each expert agent only contains MCP tool definitions in the own field, MCP tool definitions under long context are avoided to accumulate, the problem of intermediate information loss caused by the U-shaped bias of the attention mechanism is solved, and the identification and execution accuracy of key business information is improved.

The invention discloses a multi-agent collaborative reasoning method and system, a terminal and a storage medium. According to the method, structural reasoning is firstly carried out on the user demand through the first-level intelligent agent, the first reasoning conclusion with the structural mark is output, the general user demand is divided into a plurality of clear structural intention bodies, each structural intention body has a respective reasoning result, confusion when a single intelligent agent processes complex information is avoided, and the user experience is improved. And understanding deviation is reduced. Secondly, secondary structured reasoning is conducted on the specific structured intention body through the secondary agent, and the primary agent is optimized through output of the secondary agent; in the optimization process, superposition correction is not carried out in original disordered dialogues, interference of historical deviation information is avoided, it is ensured that correction focuses on specific problems every time, deviation expansion cannot be caused, and therefore a reasoning conclusion meeting user requirements is updated step by step.