CN-121981426-A - Wind-solar-hydrogen storage intelligent regulation and control method and system based on pareto multi-model collaborative evaluation

CN121981426ACN 121981426 ACN121981426 ACN 121981426ACN-121981426-A

Abstract

The invention relates to a wind-solar-hydrogen storage intelligent regulation and control method and system based on pareto multi-model collaborative evaluation, wherein the method comprises the following steps of step 1, multi-source heterogeneous data acquisition and storage; step 2, cognition diagnosis and anomaly detection based on a large language model, step 3, digital twin parallel simulation and multi-algorithm test, step 4, multi-agent cooperative score and dynamic pareto optimization, and step 5, pareto feedback preference and self-adaptive strategy switching. The system executes the wind, light and hydrogen storage intelligent regulation and control method, and comprises a cloud side large model, a simulation platform, an edge side energy management platform and an end side energy device. The multi-mode understanding and knowledge reasoning capability of the large model is applied to operation monitoring and algorithm decision of the wind, light and hydrogen storage integrated system, and multi-target collaborative optimization scheduling of hydrogen production, energy storage, surfing and the like is realized by constructing a digital twin simulation test environment and a layering intelligent evaluation system.

Inventors

Qin Manting
Xuan Shunde
LIN YUKAI
ZHANG MENG
HU YUFEI
NAN XIONG
MA KANG
LIU BOWEI
LIU WEI
CHENG YANTING
SHI WENGANG

Assignees

中国大唐集团科技创新有限公司

Dates

Publication Date: 20260505
Application Date: 20251210

Claims (9)

1. The intelligent regulation and control method for wind, light and hydrogen storage based on pareto multi-model collaborative evaluation is characterized by comprising the following steps: Step 1, multi-source heterogeneous data acquisition and storage, comprising: the method comprises the steps of collecting operation data, economic index data and environmental parameter data of all subsystems in the wind, light and hydrogen storage integrated system in real time through a multi-protocol data collection engine, and storing the operation data, the economic index data and the environmental parameter data in a time sequence database; Step 2, cognitive diagnosis and anomaly detection based on a large language model, comprising: Deploying a monitoring agent based on a large language model, and carrying out knowledge enhancement time sequence mode identification on real-time and stored data in the step 1; carrying out online reasoning by adopting a sliding window mechanism to realize real-time abnormality judgment, and automatically extracting an abnormal state data packet corresponding to a time window when an abnormal state is identified, wherein the abnormal state data packet comprises complete operation data of 5 minutes before and after abnormality; step 3, digital twin parallel simulation and multi-algorithm test, comprising: injecting the abnormal state data packet into a simulation platform, and running a plurality of preset regulation and control algorithms in parallel, wherein a regulation and control algorithm library at least comprises a hydrogen production priority algorithm, a surfing priority algorithm, an energy storage priority algorithm and a waste wind and waste light rate minimization algorithm; step 4, multi-agent cooperative score and dynamic pareto optimization, comprising: Constructing a multi-agent evaluation system, carrying out cooperative grading on each scheduling instruction set through three special evaluation agents, and introducing a pareto optimization method to realize optimal balance among three indexes of economy, low carbon and operation energy efficiency, wherein the three special evaluation agents comprise an economy evaluation agent, an operation energy efficiency evaluation agent and an operation energy efficiency evaluation agent; step 5, pareto feedback preferential and self-adaptive strategy switching, comprising: And acquiring feedback of the pareto optimal scoring result, selecting an algorithm with the highest current comprehensive scoring, comparing the algorithm with the current energy management platform algorithm, switching the energy management platform algorithm if the algorithms are different, and maintaining the original energy management platform algorithm if the algorithms are the same.
2. The intelligent regulation and control method for wind, light and hydrogen storage based on pareto multi-model collaborative evaluation according to claim 1, wherein the acquisition protocol in step 1 supports Modbus/TCP, IEC 61850 and MQTT standards, and the acquisition frequency is dynamically configured according to data types; The time sequence database adopts a cluster architecture, supports millions of measuring point storage, and the data retention strategy is configured to be 1 year of original data and 5 years of aggregated data, so that traceability of historical data is ensured.
3. The intelligent regulation and control method for wind, solar and hydrogen storage based on pareto multi-model collaborative evaluation according to claim 1 is characterized in that in step 2, a mechanism knowledge graph and a historical fault case library of intelligent body built-in equipment are monitored, and an electrolytic tank voltage characteristic formula and expert knowledge of an energy storage system decay mechanism are injected into a model through prompt engineering.
4. The wind-solar-hydrogen storage intelligent regulation and control method based on pareto multi-model collaborative evaluation according to claim 1 is characterized in that in the step 3, a simulation platform is arranged in a container mode, resources are isolated and can be expanded transversely, an electrolytic tank model in the simulation platform comprises electrochemical polarization, thermal dynamics and two-phase flow dynamic characteristics, simulation precision errors are smaller than 3%, an energy storage battery model is an aging model coupled by a second-order RC equivalent circuit, SOC estimation errors are smaller than 5%, and simulation acceleration ratio is not lower than 10:1, so that evaluation timeliness is ensured.
5. The intelligent regulation and control method for wind, solar and hydrogen storage based on pareto multi-model collaborative evaluation according to claim 1 is characterized in that in step 4, the economic evaluation agent calculates the standardized hydrogen production cost, net present value and internal yield; aiming at different evaluation indexes, designing a corresponding evaluation agent, wherein the agent takes a large language model as a core, and comprehensively scores a scheduling instruction by combining expert experience and a customized evaluation mechanism; And constructing an automatic evaluation tool by using the large language model, calling the large language model to rapidly evaluate the scheduling instruction, scoring the scheduling instruction according to a set evaluation standard, and providing a quantitative and qualitative evaluation result.
6. The intelligent regulation and control method for wind, solar and hydrogen storage based on pareto multi-model collaborative evaluation according to claim 5, wherein in step 4, the implementation of optimal balance among three indexes of economy, low carbon and operation energy efficiency by introducing the pareto optimization method comprises the following steps: setting an optimization target as follows: ; Wherein, the For the purpose of an economy score, For the purpose of low-carbon scoring, Scoring operational energy efficiency; the scoring values are derived from the collaborative scores of the multiple models and the scores are formed into a solution set Each solution Corresponds to a scoring vector: ; in the solution on the pareto front, the composite score is further calculated The following method is introduced: calculating weights from the distribution of each score: ; Wherein the method comprises the steps of ; The comprehensive score is as follows: ; By adopting an entropy weight method and (3) autonomous adjustment of a working condition factor: ; Kj is a working condition factor, alpha j is a sensitivity coefficient, and the value range is 0.2-0.5, and the sensitivity coefficient is obtained through offline historical data training; According to comprehensive scoring Selecting an optimal solution from the pareto front solution set : ; And feeding back the pareto optimal scoring result, and guiding algorithm selection to ensure that an optimal algorithm in the current running state is selected, thereby being used for energy management platform algorithm switching.
7. The intelligent regulation and control method for wind, light and hydrogen storage based on pareto multi-model collaborative evaluation according to claim 6, wherein the evaluation index is as follows: The economic performance is used for evaluating the comprehensive profitability of the system operation and reflecting the economic performance and cost control level of the system in the electric power market environment, the low-carbon performance is used for quantifying the carbon emission reduction benefit of the system operation and reflecting the contribution degree of the system to the low-carbon transformation of the energy structure in the full life cycle, and the operation energy efficiency is used for evaluating the comprehensive efficiency of the energy conversion and utilization of the system and reflecting the effective energy output and operation optimization level of the system under given input.
8. The intelligent regulation and control method for wind, solar and hydrogen storage based on pareto multi-model collaborative evaluation according to claim 6 is characterized in that in step 5, shadow mode verification is adopted in the switching process, the new algorithm performs parallel calculation output in an independent memory space, deviation verification is performed on the new algorithm and the active algorithm, manual confirmation is triggered when the deviation exceeds 15%, hot switching is completed in integer multiples of the power frequency period of a power grid, a performance monitoring window is established for 5 minutes after switching, and key indexes are reduced by more than 10% and automatically roll back.
9. A wind-solar-hydrogen storage intelligent regulation system based on pareto multi-model collaborative assessment, characterized in that the method of any one of claims 1 to 8 is performed, comprising: The cloud side comprises a large model and a simulation platform and is used for global optimization, strategy generation, anomaly detection and effect evaluation; the system comprises an energy management platform, an edge side, a control platform and a control platform, wherein the energy management platform is used as a core scheduling and data integration hub and is used for issuing scheduling instructions and receiving operation data of various energy systems; and arranging energy equipment at the end side, wherein the energy equipment comprises wind power, photovoltaic, energy storage, hydrogen production and a grid connection point, and each energy equipment is provided with a sensor and a controller and is used for collecting states and executing control instructions so as to realize real-time monitoring and control of an energy system.

Description

Wind-solar-hydrogen storage intelligent regulation and control method and system based on pareto multi-model collaborative evaluation Technical Field The invention belongs to the technical field of intelligent management and control of wind, light and hydrogen storage systems, and particularly relates to a wind, light and hydrogen storage intelligent regulation and control method and system based on pareto multi-model collaborative evaluation. Background Under the background of novel power system construction, green hydrogen coupling coal chemical industry, photoelectric hydrogen storage and charging integration and other various energy coupling systems become key research directions. The system covers the conversion and flow processes of various energy sources and substances such as electric power, chemical raw materials and the like, the dynamic characteristics of internal units are different, and the running states are complex and changeable. In addition, the multi-energy collaborative energy network is easily influenced by external factors such as market environment, natural conditions and the like, and the traditional scheduling method based on fixed rules is difficult to realize real-time optimal operation of the system. The prior art has the following obvious limitations: Firstly, the boundary of the system running condition is complex and changeable, the fluctuation of the new energy output, the randomness of the load demand and the dynamic nature of the market environment form a high uncertainty together, and the traditional algorithm is difficult to adapt to the dynamic working condition conversion demand of the wind-light-hydrogen energy storage source system in multiple time scales. Typically, the complexity of online calculation of the traditional Model Predictive Control (MPC) method is O (n 3), the response time reaches 5-10 seconds, and the real-time regulation and control requirement of second-level fluctuation cannot be met. Secondly, the traditional mathematical modeling method relies on linearization assumption, so that the strong nonlinear coupling relation between the multi-energy flow and the multi-material flow is difficult to accurately represent, and the problem of model misalignment exists in the operation state analysis. Specifically, the nonlinear error of the voltage characteristic of the electrolytic cell and the SOC estimation error of the energy storage system are high, and the scheduling decision accuracy is seriously affected. Thirdly, the existing method lacks the capability of deep backtracking and experience migration of historical operation data, and abnormal events in a real scene have sparsity and unpredictability, and have the disadvantages of response lag, high misjudgment rate and obvious adaptability when facing to extreme or complex scenes. And the industrial scene has strict requirements on real-time performance and safety, the traditional large model deployment scheme has the problems of high response delay (average 2-3 seconds), unexplained decision, lack of safety redundancy mechanism and the like, cannot meet the IEC 61508 functional safety standard, and restricts the industrial application of the large model in the field of energy control. Disclosure of Invention The invention aims to provide a wind-solar-hydrogen storage intelligent regulation and control method and system based on pareto multi-model collaborative evaluation, which utilize parallel computing capacity of a large-scale simulation platform and multi-model understanding capacity of a large language model, combine multi-source operation data of a low-carbon energy system, including equipment state, environmental parameters, market information and the like, and construct a closed-loop decision mechanism of acquisition, test, evaluation and switching. The comprehensive evaluation of the performance of multiple algorithms is realized through data acquisition and simulation test, an algorithm performance database is established, an optimal model based on the pareto front is established through performance evaluation, the accurate decision of the algorithm is realized, the optimal matching of the algorithm strategy and the operation working condition is ensured through dynamic switching, a real-time feedback mechanism continuous optimization decision model is established, a hierarchical evaluation system, a multi-objective weighing strategy and a systematic algorithm performance verification process are adopted, the accuracy, adaptability and economy of system regulation and control are improved, an efficient and reliable intelligent algorithm scheduling framework of the low-carbon energy system is constructed, and the first low-carbon energy management platform supporting autonomous decision of a large model in China is realized. The invention provides a wind-solar-hydrogen storage intelligent regulation and control method based on pareto multi-model collaborative evaluation, which comprises the f