CN-121978954-A - Building energy-saving self-adaptive regulation and control method based on scene analysis

CN121978954ACN 121978954 ACN121978954 ACN 121978954ACN-121978954-A

Abstract

The invention provides a building energy-saving self-adaptive regulation and control method based on scene analysis, which comprises the following steps of forming a state space representing a building scene through time characteristic information, environment data and equipment running state data, discretizing or parameterizing executable regulation and control operation of building energy equipment to be controlled to form an action space, taking the state space as input, taking the action space as output, taking a multi-objective rewarding function as an optimization target, training a DQN model, inputting the current state space into the pre-trained DQN model, outputting equipment control actions by the DQN model, generating equipment regulation strategies according to the equipment control actions, executing the equipment regulation strategies, obtaining new state and actual energy consumption data of building equipment after the equipment regulation strategies are executed, regulating the DQN model through periodic sampling, and updating the action space of the DQN model. The method and the system can dynamically construct the building operation scene, generate the cross-equipment cooperative optimization control strategy and reduce the building energy consumption.

Inventors

LI LIHONG
LIAO JIANMIN

Assignees

桂林理工大学南宁分校

Dates

Publication Date: 20260505
Application Date: 20260209

Claims (10)

1. The building energy-saving self-adaptive regulation and control method based on scene analysis is characterized by comprising the following steps of: Acquiring environment data and equipment operation state data through a multi-mode sensor network arranged in a building, and setting time characteristic information; Discretizing or parameterizing executable regulation and control operation of building energy equipment to be controlled to form an action space; setting a multi-objective rewarding function according to the energy efficiency control target, the comfort control target and the air quality control target; Setting a to-be-trained DQN model, taking a state space as input, taking an action space as output, taking a multi-objective rewarding function as an optimization target, and training the DQN model to obtain a pre-trained DQN model; Acquiring environment data and equipment running state data of building equipment to be adjusted, constructing time characteristic information at the current moment, and generating a current state space; Inputting the current state space into a pre-trained DQN model, and outputting a control action of equipment by the DQN model; generating a device adjustment strategy according to the device control action, and executing the device adjustment strategy; The method comprises the steps of obtaining new state and actual energy consumption data of building equipment after equipment regulation strategy execution, calculating instant rewards according to a multi-objective rewarding function, storing the instant rewards, the new state and the actual energy consumption data into an experience playback pool of the DQN model, adjusting the DQN model through periodic sampling, and updating an action space of the DQN model.
2. The scene analysis-based building energy-saving adaptive regulation and control method according to claim 1, wherein the environmental data comprises indoor temperature, indoor humidity, carbon dioxide concentration, illumination intensity and personnel presence information, and outdoor meteorological data acquired through an internet of things gateway; the equipment operation state data comprise real-time power and energy consumption data of main energy consumption equipment and setting parameters of the energy consumption equipment; The time characteristic information comprises a current moment, a date type, season information and a sunlight period mark.
3. The method for adaptively controlling energy conservation in a building based on scene analysis according to claim 1, wherein the discretizing or parameterizing the executable control operation of the building energy device to be controlled, comprises: For an air conditioning system, discretizing the adjustment action of the temperature set point into one-step rising, one-step falling or keeping unchanged; For the ventilation fan, discretizing the rotation speed control into three gears of high, medium and low; an illumination system parameterizing brightness adjustment to continuously adjust over a range of zero to one hundred percent; And a binary control instruction for controlling the start and stop of the fresh air system.
4. The method for adaptively adjusting and controlling building energy conservation based on scene analysis according to claim 1, wherein the setting of the multi-objective rewarding function according to the energy efficiency control objective, the comfort control objective and the air quality control objective comprises: Setting a composite rewarding function formed by weighted summation, wherein the composite rewarding function mainly comprises an energy efficiency rewarding item, a comfort punishment item and an air quality punishment item; the value of the energy efficiency rewarding item is inversely proportional to the real-time total energy consumption of the system; the value of the comfort penalty term is increased according to the degree that the indoor actual temperature deviates from the human body comfort temperature interval; And the air quality penalty term gives negative rewards when the carbon dioxide concentration exceeds a preset health threshold.
5. The building energy-saving self-adaptive regulation and control method based on scene analysis according to claim 1, wherein the training of the DQN model to obtain the pre-trained DQN model takes a state space as input, an action space as output, and a multi-objective rewarding function as an optimization target comprises the following steps: randomly initializing an evaluation neural network and a target neural network with the same structure; in the simulation environment or historical data, enabling an intelligent agent to randomly explore or select a control action according to a strategy of an optimal decision of a current model according to a current environment state and a certain probability; After the action is executed, the environment generates a new state and calculates an instant rewarding value as an interactive experience; in training iteration, a batch of past experiences are randomly extracted from an experience pool at regular intervals and used for calculating the prediction error of the neural network; and synchronizing the parameters of the evaluation network to the target network every fixed step number, and repeating until convergence.
6. The method for building energy-saving adaptive regulation and control based on scene analysis according to claim 1, wherein the device regulation strategy comprises: Generating a device control command sequence capable of being issued according to a specific action instruction output by the intelligent decision; Automatically reducing the illumination brightness according to the readings of the illumination sensor so as to fully utilize natural light; According to the indoor and outdoor temperature difference and the carbon dioxide concentration, different ventilation modes are adopted in an intelligent decision.
7. The scene analysis-based building energy-saving adaptive regulation and control method according to claim 1, wherein the storing of instant rewards, new states and actual energy consumption data into an experience playback pool of the DQN model, the adjusting of the DQN model by periodic sampling, the updating of the action space of the DQN model, comprises the steps of: Storing new experience data generated by each decision interaction, including the current state, the execution action, the acquisition of rewards and the next state, into an experience playback pool; A small batch of sample data is randomly extracted from the experience pool, the model predicted loss is calculated, and the network parameters of the DQN model are updated through a back propagation algorithm.
8. The building energy-saving self-adaptive regulation and control system based on scene analysis is characterized by comprising a first processing module, a second processing module, a third processing module and a fourth processing module, wherein the first processing module is used for acquiring environment data and equipment operation state data through a multi-mode sensor network arranged in a building and setting time characteristic information; The second processing module is used for discretizing or parameterizing the executable regulation and control operation of the building energy-saving equipment to be controlled to form an action space; the third processing module is used for setting the DQN model to be trained, taking a state space as input, taking an action space as output, taking a multi-objective rewarding function as an optimization target, and training the DQN model to obtain the pre-trained DQN model; The fourth processing module is used for acquiring environment data and equipment running state data of building equipment to be adjusted, constructing time characteristic information at the current moment and generating a current state space; A fifth processing module, configured to input the current state space into a pre-trained DQN model, where the DQN model output device controls actions; generating a device adjustment strategy according to the device control action, and executing the device adjustment strategy; the system comprises a device regulation strategy, a sixth processing module, a prompt rewarding function, an experience playback pool, a DQN model and a motion space updating module, wherein the device regulation strategy is used for executing a device regulation strategy, the sixth processing module is used for acquiring new state and actual energy consumption data of building devices after the device regulation strategy is executed, calculating the prompt rewarding according to the multi-target rewarding function, storing the prompt rewarding, the new state and the actual energy consumption data into the experience playback pool of the DQN model, and adjusting the DQN model through periodic sampling to update the motion space of the DQN model.
9. A computer device, comprising: memory, transceiver, processor, and bus system; Wherein the memory is used for storing programs; The processor is used for executing the program in the memory, and comprises executing a building energy-saving self-adaptive regulation and control method based on scene analysis as claimed in any one of claims 1 to 7; The bus system is used for connecting the memory and the processor so as to enable the memory and the processor to communicate.
10. A readable storage medium storing computer readable instructions, wherein the computer readable instructions when executed by a processor implement a building energy saving adaptive regulation method step based on scene analysis according to any one of claims 1 to 7.

Description

Building energy-saving self-adaptive regulation and control method based on scene analysis Technical Field The invention relates to the technical field of building management, in particular to a building energy-saving self-adaptive regulation and control method based on scene analysis. Background The building energy-saving regulation and control technology is realized mainly by means of a building automation system. The traditional building control system generally adopts a programmed control based on a fixed time table or a feedback control strategy based on a single parameter threshold value to independently start, stop or regulate equipment such as an air conditioner, illumination and the like. In recent years, with the development of control theory, advanced methods such as proportional-integral-derivative control and fuzzy logic control have been introduced into the field to improve the adaptability of a single control loop. Meanwhile, the popularization of the internet of things technology enables real-time acquisition of building environment and equipment data through multiple sensor networks, and a foundation is laid for data-driven monitoring and management. However, the prior art still faces significant limitations and industry pain points in practical applications. The traditional control method is static or semi-static in nature, is difficult to dynamically respond to complex scenes coupled by multiple factors such as personnel density fluctuation, weather change, equipment performance attenuation and the like, and is easy to cause energy waste or comfort level reduction. More importantly, subsystems such as illumination, air conditioning, ventilation and the like often run and control independently, a cooperative optimization mechanism is lacked, and global optimization of the overall energy efficiency of the building is difficult to achieve. Although intelligent methods such as fuzzy control and the like improve the adaptability to a certain extent, the intelligent methods are highly dependent on expert experience to construct a knowledge base, and have the inherent problems of high design cost and difficult coverage of unknown scenes. In addition, most of the existing systems still belong to open-loop or semi-closed-loop control, the energy consumption analysis is generally lagged, on-line strategy optimization cannot be performed based on real-time operation feedback, and the self-adaptive capability of continuous self-learning and evolution is essentially lacking. Disclosure of Invention The invention provides a building energy-saving self-adaptive regulation and control method based on scene analysis, which can dynamically construct a building operation scene through multi-mode perception data, generate a cross-equipment cooperative optimization control strategy based on a deep reinforcement learning model, obviously reduce the overall energy consumption of a building on the premise of ensuring indoor environment comfort and air quality, and have online self-learning capability to continuously adapt to the changes of environment and use modes. The invention provides a building energy-saving self-adaptive regulation and control method based on scene analysis, which comprises the following steps: Acquiring environment data and equipment operation state data through a multi-mode sensor network arranged in a building, and setting time characteristic information; Discretizing or parameterizing executable regulation and control operation of building energy equipment to be controlled to form an action space; setting a multi-objective rewarding function according to the energy efficiency control target, the comfort control target and the air quality control target; Setting a to-be-trained DQN model, taking a state space as input, taking an action space as output, taking a multi-objective rewarding function as an optimization target, and training the DQN model to obtain a pre-trained DQN model; Acquiring environment data and equipment running state data of building equipment to be adjusted, constructing time characteristic information at the current moment, and generating a current state space; Inputting the current state space into a pre-trained DQN model, and outputting equipment control actions by the DQN model; The method comprises the steps of obtaining new state and actual energy consumption data of building equipment after equipment regulation strategy execution, calculating instant rewards according to a multi-objective rewarding function, storing the instant rewards, the new state and the actual energy consumption data into an experience playback pool of the DQN model, adjusting the DQN model through periodic sampling, and updating an action space of the DQN model. Further, the environmental data comprises indoor temperature, indoor humidity, carbon dioxide concentration, illumination intensity and personnel presence information, and outdoor meteorological data obtained through an Internet