Large-scale public building wind and water joint debugging method containing digital twinning and deep reinforcement learning

Abstract

The invention discloses a large-scale public building wind and water joint debugging method containing digital twinning and deep reinforcement learning. The method comprises the following steps: constructing a digital twinning model of a multi-level wind and water joint debugging system of a large-scale public building; constructing an hourly load demand prediction model of each user building, and calculating to obtain an hourly load demand prediction result of each building of the large public building; the hourly air volume demand result, the water consumption demand result and the water temperature demand result of each user building and the hourly water consumption demand result of each building are analyzed, and the operation modes of the multi-level wind-water joint debugging system of the large public building are divided in combination with external condition changes; taking ventilation energy consumption minimization, water consumption minimization and temperature deviation minimization of each user building area as targets, and setting a multi-level control variable combination target value of the wind-water joint debugging system adapted to each operation mode; and performing multi-level equipment control instruction calculation of the wind-water joint debugging system by adopting a deep reinforcement learning algorithm, and issuing to related equipment of each level for execution.

Description

Technical Field

[0001] The present invention belongs to the technical field of feng shui joint adjustment of large public buildings, and specifically relates to a feng shui joint adjustment method for large public buildings containing digital twins and deep reinforcement learning. Background Art

[0002] With the rapid development of urbanization in my country, the construction industry has been driven forward. The volume of my country's construction industry has continued to increase. In addition to residential buildings, the proportion of large public buildings is also relatively high, such as newly built large airports, subway stations, commercial buildings, office buildings, etc. The increase in the volume of these large public buildings has increased energy consumption. After they are built and put into use, they need to provide corresponding energy according to the building function type, environmental conditions, and crowd density to meet the energy needs of the building itself during operation.

[0003] At present, large public buildings will be centrally installed with central air conditioning systems to provide heating and cooling to meet the demand for hot and cold loads in the building. The load at the end of the central air conditioning changes dynamically. The load change corresponds to the change in the hot water or chilled water flow of the air conditioning water system and the change in the required air volume of the air conditioning air system. In addition, the air conditioning air system and the water system are a pair of coupled systems. Compared with the traditional air conditioning system, the Feng Shui joint adjustment system has higher energy efficiency. The Feng Shui joint adjustment system usually regulates the chilled water flow in the fan coil unit based on the constant temperature difference of the chilled water, and also regulates the fan speed to adjust the air supply volume to meet the indoor load demand and reduce the energy consumption of the system operation. However, there are some problems with the Feng Shui joint adjustment system. For example, in the face of changes in various external conditions, such as different seasons, different outdoor environments, changes in indoor area traffic, extreme weather, disasters, etc., the Feng Shui joint adjustment system cannot operate adaptively. Therefore, the Feng Shui joint adjustment system of large public buildings needs to be able to adapt to the changes in various conditions to switch the system operation mode and optimize the equipment. In addition, the accuracy of the Feng Shui joint adjustment system equipment optimization is low.

[0004] Based on the above technical problems, it is necessary to design a new feng shui coordination method for large public buildings containing digital twins and deep reinforcement learning. Summary of the invention

[0005] The purpose of the present invention is to provide a Feng Shui joint regulation method for large public buildings containing digital twins and deep reinforcement learning. The method can achieve Feng Shui joint regulation of large public buildings through multi-level joint regulation by adjusting the energy center on the source side, adjusting the water flow of buildings in front of the buildings, adjusting the air volume of air conditioners in front of households, and adjusting the water flow and water temperature of the air conditioning water system. The Feng Shui joint regulation system can be divided into fine operating modes, and the Feng Shui joint regulation systems with different operating modes can be optimized and controlled, which can adapt to the changes in conditions under various operating modes.

[0006] In order to solve the above technical problems, the present invention provides a large-scale public building feng shui joint adjustment method containing digital twins and deep reinforcement learning, and the large-scale public building feng shui joint adjustment method comprises:

[0007] S1. Install electric regulating valves for regulating water flow in front of each building at the end of a large public building, install air conditioning and ventilation devices for variable frequency adjustment of air volume in front of each user in each building, and install air conditioning water systems for regulating user water flow and water temperature, and combine the energy center on the source side and the heat exchange station on the station side to form a multi-level wind and water joint adjustment system for large public buildings;

[0008] S2. Construct a digital twin model of the multi-level Feng Shui coordination system of large public buildings, including the physical layer, interaction layer, digital twin layer and application layer, and verify the consistency between the digital twin model and the actual system physical entity;

[0009] S3. Based on the verified digital twin model of the multi-level Feng Shui coordination system of large public buildings, the hourly load demand prediction model of each user building is constructed considering the function, location, population density, external environmental conditions and environmental parameters of each user building area in each building, and the hourly load demand prediction results of each building of large public buildings are calculated;

[0010] S4. Based on the hourly load demand forecast results of each user building and the hourly load demand forecast results of each building, analyze the hourly wind volume demand results, water demand results and water temperature demand results of each user building, and the hourly water demand results of each building, and divide the operation mode of the multi-level Feng Shui joint adjustment system of large public buildings in combination with changes in external conditions;

[0011] S5. According to the operation modes of the multi-level Feng Shui joint adjustment system of large public buildings, with the goal of minimizing ventilation energy consumption, water consumption and temperature deviation of each user building area, set the target value of the multi-level control variable combination of the Feng Shui joint adjustment system that adapts to each operation mode;

[0012] S6. According to the target value of the multi-level control variable combination of the Feng Shui coordination system in each operating mode, a deep reinforcement learning algorithm is used to calculate the multi-level equipment control instructions of the Feng Shui coordination system, and the instructions are sent to the relevant equipment at each level for execution.

[0013] Furthermore, in step S1, the source-side energy center includes at least a plurality of chillers for cooling and a plurality of boilers for heating; the number of the station-side heat exchange stations is multiple, and each heat exchange station is connected to the air conditioning and ventilation device and the air conditioning water system of each user;

[0014] The air conditioning and ventilation device comprises: a fan and a fan coil;

[0015] The air conditioning water system includes: air conditioning equipment, water distributor, water collector, valve and water pump;

[0016] The regulation strategy of the multi-level Feng Shui joint regulation system of large public buildings includes: first, the hot water supplied by the boiler of the energy center on the source side or the chilled water supplied by the chiller is distributed to each building through the regulating electric regulating valve, and then the water pump in the air-conditioning water system sends the hot water or chilled water to the water distributor, and then the water distributor sends it to the air-conditioning equipment of each user through each water supply branch pipe, and after the water temperature treatment of the air-conditioning equipment and the water flow adjustment of the valve, it flows back to the water collector through each return branch pipe and returns to the boiler or chiller through the water collector, and the cycle is repeated to change the water flow and water temperature, and finally flows through the fan coil units of each user's building area, and then the fan frequency conversion adjusts the air volume to adjust the temperature distribution and air flow of each user's building area.

[0017] Further, the S2 specifically includes:

[0018] Construct the physical layer: Use multi-source sensors to sense the equipment operating conditions and environmental information of the multi-level Feng Shui joint adjustment system of large public buildings and transmit them to the interactive layer, and realize interactive mapping with the digital twin layer through interactive feedback of multi-source data, and control the equipment of the Feng Shui joint adjustment system according to the application layer; the equipment operating conditions include the operating status parameters of each device in the Feng Shui joint adjustment system; the environmental information includes outdoor weather temperature, weather humidity, building area pedestrian flow and building area temperature;

[0019] Build the interaction layer: Transmit the equipment operating conditions and environmental information collected by multi-source sensors to the application layer, and establish state synchronization and interactive collaborative control between the digital twin layer and the physical entity through data preprocessing, data fusion, and data mining;

[0020] Constructing the digital twin layer: constructing a digital twin model that integrates the physical model, mechanism model, data model, and simulation model; the physical model indicates the structural characteristics of the physical entity of the Feng Shui joint adjustment system of large public buildings; the mechanism model indicates the logical description and mechanism explanation of the physical entity when performing Feng Shui joint adjustment; the data model includes the perceived equipment operating conditions, environmental information, and simulation operation data of the physical entity of the Feng Shui joint adjustment system, and uses the simulation operation data to monitor the digital twin model in real time, and optimizes the parameters and structure of the digital twin model based on the real-time monitoring data; the simulation model indicates the virtual-real interaction between the physical entity of the Feng Shui joint adjustment system and the digital twin model;

[0021] Build the application layer: Establish an intelligent decision-making module for the Feng Shui joint adjustment system to monitor the system operation status and make Feng Shui joint adjustment control decisions;

[0022] The digital twin model of the multi-level Feng Shui coordination system of large public buildings is constructed by interactively integrating the physical layer, interaction layer, digital twin layer and application layer;

[0023] Perform consistency check on the physical entity and digital twin model of the Feng Shui coordination system: Evaluate whether the physical structure, operating status and consistency of the digital twin model with the physical entity are within the preset range. If so, the consistency check passes; otherwise, perform parameter optimization and structural optimization on the digital twin model.

[0024] Furthermore, the evaluation of the consistency between the physical structure and operating status of the digital twin model and the object entity includes:

[0025] Establish a multi-dimensional first-level evaluation index system for consistency evaluation, including: model consistency index, simulation consistency index, data consistency index and interaction consistency index;

[0026] Establish a secondary evaluation indicator system corresponding to each indicator in the multi-dimensional primary evaluation indicator system;

[0027] After calculating the subjective and objective weights of the evaluation index system, the subjective and objective weights are integrated to obtain a comprehensive evaluation result of the consistency between the digital twin model and the physical entity.

[0028] The sovereign weight calculation adopts a fuzzy comprehensive evaluation algorithm that combines the hierarchical analysis method and the fuzzy matrix operation method; the objective weight calculation adopts the entropy weight method; the fusion of the subjective and objective weights adopts the game theory method.

[0029] Further, in S3, the functions of each user building area in each building include: business premises, office premises; the functions of each user building area in each building will cause different degrees of influence of heat load or cold load, and different functional areas require different load durations; the heat load or cold load influence is defined as:

[0030]

[0031] ε i is the influence degree of heating load or cooling load at the i-th moment. The greater the influence degree, the smaller the demand for heating load or cooling load. ij is the heating load data or cooling load data at the i-th moment of the j-th day; n is the total number of days;

[0032] The pedestrian flow density in the area includes the pedestrian flow density on working days and the pedestrian flow density on non-working days, and the pedestrian flow density is predicted by using a first machine learning algorithm;

[0033] The construction of the hourly load demand prediction model for each user building includes: using the second machine learning algorithm and the third machine learning algorithm to perform predictions respectively, and then performing weighted combination of the prediction results through a reinforcement learning algorithm to construct the hourly load demand prediction model for each user building;

[0034] The hourly load demand forecast results of each building of the large public building are obtained by cumulatively calculating the hourly load demand forecast results of each user building under the building.

[0035] Furthermore, the first machine learning algorithm is an ARIMA algorithm; the second machine learning algorithm is an NBEATSx time series data prediction algorithm; the third machine learning algorithm is a DELM deep extreme learning machine algorithm, and a snow melting optimization algorithm is used to optimize the parameters of the DELM algorithm; and the reinforcement learning algorithm is a Q learning algorithm.

[0036] Further, the S4 specifically includes:

[0037] The hourly load demand forecast results of each user building and the hourly load demand forecast results of each building are input into the digital twin model of the multi-level Feng Shui joint adjustment system of large public buildings, and Feng Shui joint adjustment simulation analysis is carried out to obtain the hourly air volume demand results, water demand results and water temperature demand results of each user building that meet the hourly load demand, and the hourly water demand results of each building;

[0038] Obtain changes in external conditions, including seasonal changes, time changes, and abnormal changes; seasonal changes include focusing on heat load and ventilation requirements in winter, focusing on cooling load and ventilation requirements in summer, and focusing on ventilation requirements in transition seasons; time changes include whether the operating time is daytime or nighttime; abnormal changes include extreme weather changes, equipment failure changes, and disaster changes;

[0039] According to the hourly wind volume demand results, water consumption demand results and water temperature demand results of each user building, the hourly water consumption demand results of each building, combined with seasonal changes, time changes and abnormal changes, a cluster analysis algorithm is used to establish a Feng Shui coordination system operation mode recognition model, and the operation mode is identified and divided; the operation mode includes a daily operation mode that meets the air volume demand, water consumption demand, and water temperature demand, an energy-saving operation mode during non-peak demand periods or transition seasons while ensuring comfort, and an emergency operation mode in extreme weather, equipment failure changes, and disasters.

[0040] Furthermore, in S5, the objectives of minimizing ventilation energy consumption, minimizing water consumption, and minimizing temperature deviations in each user's building area are respectively expressed as:

[0041]

[0042] f1 is the minimization of ventilation energy consumption; W fan,k,i is the fan energy consumption of the i-th user building area in the k-th operation mode; M is the number of user buildings; f2 is the minimization of water consumption; W unit,k,j W is the energy consumption of the jth unit under the kth operating mode; pump,k,j is the energy consumption of the jth water pump under the kth operation mode; N is the number of water-using devices; f3 is the minimum temperature deviation of each user's building area; T set,k,i is the set temperature of the i-th user building area under the k-th operating mode, and the temperature is set according to different operating modes; T room,k,i The actual temperature of the i-th user building area under the k-th operating mode;

[0043] The setting of the multi-level control variable combination target value of the Feng Shui coordination system that adapts to each operating mode includes: solving each target through an intelligent optimization algorithm to obtain the control variable combination target value for adjusting water flow, water temperature and ventilation volume in the multi-level Feng Shui coordination system that adapts to each operating mode.

[0044] Further, the S6 specifically includes:

[0045] Constructing a mathematical model of the Markov decision process for equipment control of the Feng Shui joint control system, including setting a state space, an action space, a reward function, a state transition probability and a discount factor; the state space represents the equipment operating state parameters and environmental information of the Feng Shui joint control system; the action space represents the multi-level equipment control variable actions of the Feng Shui joint control system, including the chiller operating power, the boiler operating power, the electric regulating valve opening, the air conditioning equipment operating power, the water pump operating frequency, the valve opening and the fan operating frequency; the reward function represents the minimum deviation between the target value and the actual value of the multi-level control variable combination of the Feng Shui joint control system in each operating mode; the state transition probability is the probability that the state is transferred from the current state to another state after executing the equipment control variable action in the action space; the discount factor represents the reward attenuation factor;

[0046] The Actor network in the deep reinforcement learning TD3 algorithm interacts with the environment to obtain the action at the time t, and obtains the state and reward at the next moment. Then, the Critic network and the Actor network are updated by taking out the state, action, reward at the time t and the state at the next moment from the experience replay pool, and the target Actor network and the target Critic network are updated regularly. Finally, after continuous iterative updates, the network strategy is trained in the direction of higher reward values, and the multi-level equipment control action instructions of the Feng Shui joint control system are obtained, and sent to the relevant equipment at each level for execution.

[0047] Furthermore, the deep reinforcement learning TD3 algorithm structure includes an online network and a target network; the online network includes an online Actor network and two online Critic networks; the target network includes a target Actor network and two target Critic networks.

[0048] The beneficial effects of the present invention are:

[0049] (1) The present invention forms a multi-level Feng Shui joint adjustment system for large public buildings by installing electric regulating valves for adjusting the water flow of buildings in front of each building at the end of a large public building, installing air conditioning and ventilation devices for frequency conversion adjustment of air volume in front of each user in each building, and installing an air conditioning water system for adjusting the water flow and water temperature of the user, and combining the source-side energy center and the station-side heat exchange station; the Feng Shui joint adjustment of large public buildings can be achieved through the adjustment of the source-side energy center, the building water flow adjustment in front of the building, the air volume adjustment of the air conditioning in front of the household, and the water flow and water temperature adjustment of the air conditioning water system through multi-level joint adjustment. The cold and hot loads of the subordinate areas can be changed through the water flow in front of the building, and the cold and hot fine-tuning of each local area can be performed by adjusting the water temperature and water flow in front of the household and adjusting the air volume through frequency conversion to meet the load requirements of each area;

[0050] (2) The present invention constructs a digital twin model of a multi-level Feng Shui coordination system for large public buildings, including a physical layer, an interaction layer, a digital twin layer, and an application layer, and verifies the consistency of the digital twin model with the actual system physical entity; the Feng Shui coordination system for large public buildings can be constructed in a virtual space to form a system digital twin model, complete the mapping of the Feng Shui coordination physical system, facilitate subsequent prediction, decision-making, and control, and improve the accuracy of the digital twin model by verifying the consistency of the digital twin model with the physical entity;

[0051] (3) The present invention constructs an hourly load demand prediction model for each user building based on the verified digital twin model of the multi-level Feng Shui joint adjustment system of large public buildings, taking into account the function, location, population density, external environmental conditions and environmental parameters of each user building area in each building, and calculates and obtains the hourly load demand prediction results of each building of the large public building; it can integrate various static and dynamic data parameters to establish a load demand prediction model to achieve subsequent on-demand precise regulation;

[0052] (4) Based on the hourly load demand prediction results of each user building and the hourly load demand prediction results of each building, the present invention analyzes the hourly air volume demand results, water consumption demand results and water temperature demand results of each user building, and the hourly water consumption demand results of each building, and divides the operation mode of the multi-level Feng Shui joint regulation system of large public buildings in combination with changes in external conditions; it can further obtain the air volume, water consumption and water temperature requirements of each area and each building based on the load prediction results, and perform a refined division of the operation mode of the Feng Shui joint regulation system, which is convenient for the subsequent optimization and control of the Feng Shui joint regulation system in different operation modes, and can adapt to the changes in conditions under various operation modes;

[0053] (5) The present invention sets the target value of the multi-level control variable combination of the Feng Shui joint adjustment system adapted to each operation mode according to the divided operation mode of the multi-level Feng Shui joint adjustment system of large public buildings, with the goal of minimizing ventilation energy consumption, water consumption and temperature deviation of each user building area; it can ensure that the energy-saving and comfort goals are met as much as possible in each operation mode, and the target value of the system control variable combination changes according to the different operation modes, so as to realize the control strategy of the Feng Shui joint adjustment operation mode self-adaptation;

[0054] (6) The present invention uses a deep reinforcement learning algorithm to calculate the multi-level equipment control instructions of the Feng Shui coordination system according to the target value of the multi-level control variable combination of the Feng Shui coordination system in each operating mode, and sends them to the relevant equipment at each level for execution; the use of the deep reinforcement learning algorithm can provide the equipment control effect of the Feng Shui coordination system.

[0055] Other features and advantages will be described in the following description, and partly become apparent from the description, or understood by practicing the invention. The purpose and other advantages of the invention are realized and obtained by the structures particularly pointed out in the description and the drawings.

[0056] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0058] Figure 1 This is a flow chart of a large-scale public building feng shui joint adjustment method including digital twins and deep reinforcement learning in the present invention;

[0059] Figure 2 It is a schematic diagram of the structure of the multi-level Feng Shui coordination system for large public buildings of the present invention. DETAILED DESCRIPTION

[0060] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0061] Example 1

[0062] like Figure 1 As shown, this embodiment 1 provides a large-scale public building feng shui joint adjustment method including digital twins and deep reinforcement learning, and the large-scale public building feng shui joint adjustment method includes:

[0063] S1. Install electric regulating valves for regulating water flow in front of each building at the end of a large public building, install air conditioning and ventilation devices for variable frequency adjustment of air volume in front of each user in each building, and install air conditioning water systems for regulating user water flow and water temperature, and combine the energy center on the source side and the heat exchange station on the station side to form a multi-level wind and water joint adjustment system for large public buildings;

[0064] S2. Construct a digital twin model of the multi-level Feng Shui coordination system of large public buildings, including the physical layer, interaction layer, digital twin layer and application layer, and verify the consistency between the digital twin model and the actual system physical entity;

[0065] S3. Based on the verified digital twin model of the multi-level Feng Shui coordination system of large public buildings, the hourly load demand prediction model of each user building is constructed considering the function, location, population density, external environmental conditions and environmental parameters of each user building area in each building, and the hourly load demand prediction results of each building of large public buildings are calculated;

[0066] S4. Based on the hourly load demand forecast results of each user building and the hourly load demand forecast results of each building, analyze the hourly wind volume demand results, water demand results and water temperature demand results of each user building, and the hourly water demand results of each building, and divide the operation mode of the multi-level Feng Shui joint adjustment system of large public buildings in combination with changes in external conditions;

[0067] S5. According to the operation modes of the multi-level Feng Shui joint adjustment system of large public buildings, with the goal of minimizing ventilation energy consumption, water consumption and temperature deviation of each user building area, set the target value of the multi-level control variable combination of the Feng Shui joint adjustment system that adapts to each operation mode;

[0068] S6. According to the target value of the multi-level control variable combination of the Feng Shui coordination system in each operating mode, a deep reinforcement learning algorithm is used to calculate the multi-level equipment control instructions of the Feng Shui coordination system, and the instructions are sent to the relevant equipment at each level for execution.

[0069] In this embodiment, in step S1, the source-side energy center includes at least a plurality of chillers for cooling and a plurality of boilers for heating; the number of the station-side heat exchange stations is multiple, and each heat exchange station is connected to the air conditioning and ventilation device and the air conditioning water system of each user;

[0070] The air conditioning and ventilation device comprises: a fan and a fan coil;

[0071] The air conditioning water system includes: air conditioning equipment, water distributor, water collector, valve and water pump;

[0072] The regulation strategy of the multi-level Feng Shui joint regulation system of large public buildings includes: first, the hot water supplied by the boiler of the energy center on the source side or the chilled water supplied by the chiller is distributed to each building through the regulating electric regulating valve, and then the water pump in the air-conditioning water system sends the hot water or chilled water to the water distributor, and then the water distributor sends it to the air-conditioning equipment of each user through each water supply branch pipe, and after the water temperature treatment of the air-conditioning equipment and the water flow adjustment of the valve, it flows back to the water collector through each return branch pipe and returns to the boiler or chiller through the water collector, and the cycle is repeated to change the water flow and water temperature, and finally flows through the fan coil units of each user's building area, and then the fan frequency conversion adjusts the air volume to adjust the temperature distribution and air flow of each user's building area.

[0073] In actual applications, taking large airport buildings as an example, the optimization control of the airport central air conditioning Feng Shui joint adjustment system is proposed. The system automatically calculates the prediction results of the airport's air conditioning load demand based on parameters such as indoor and outdoor temperature, humidity, inbound and outbound passenger flow, and decides the number of units and boilers to be started in the Feng Shui joint adjustment system, as well as the operating parameters of units, boilers, fans, pumps, valves and other equipment. Among them, the loads of various building areas in the airport are also different. For example, the flow of people in the hall changes with the time of the flight, and the flow of people in the business area of ​​the airport also changes during the day and at night. Therefore, it is necessary to accurately predict and control the load of each building area to meet on-demand control and energy saving.

[0074] like Figure 2As shown in the structural schematic diagram of the multi-level Feng Shui coordination system of a large public building, the energy center on the source side may include a chiller, a gas boiler, an electric boiler and a flue gas recovery device. There can be multiple units of each device, which are equipped according to the load requirements of the end users, and the output is hot water or chilled water; the building water flow regulation is to adjust the flow of hot water or chilled water in each building through the provision of multiple electric regulating valves, and then after heat exchange treatment through multiple heat exchange stations on the station side, it is transported to the corresponding air-conditioning water system, and after treatment by water distributors, water pumps, air-conditioning equipment, water collectors, valves and other equipment, the water temperature and water flow are regulated and controlled, and then it is transported to the corresponding air-conditioning air system, and after adjustment and control by fan coil units and fans, the temperature distribution and air flow of each user area at the end are realized to meet the load requirements of each user area, and finally the multi-level Feng Shui coordination control from top to bottom of large public buildings is realized.

[0075] In this embodiment, S2 specifically includes:

[0076] Construct the physical layer: Use multi-source sensors to sense the equipment operating conditions and environmental information of the multi-level Feng Shui joint adjustment system of large public buildings and transmit them to the interactive layer, and realize interactive mapping with the digital twin layer through interactive feedback of multi-source data, and control the equipment of the Feng Shui joint adjustment system according to the application layer; the equipment operating conditions include the operating status parameters of each device in the Feng Shui joint adjustment system; the environmental information includes outdoor weather temperature, weather humidity, building area pedestrian flow and building area temperature;

[0077] Build the interaction layer: Transmit the equipment operating conditions and environmental information collected by multi-source sensors to the application layer, and establish state synchronization and interactive collaborative control between the digital twin layer and the physical entity through data preprocessing, data fusion, and data mining;

[0078] Constructing the digital twin layer: constructing a digital twin model that integrates the physical model, mechanism model, data model, and simulation model; the physical model indicates the structural characteristics of the physical entity of the Feng Shui joint adjustment system of large public buildings; the mechanism model indicates the logical description and mechanism explanation of the physical entity when performing Feng Shui joint adjustment; the data model includes the perceived equipment operating conditions, environmental information, and simulation operation data of the physical entity of the Feng Shui joint adjustment system, and uses the simulation operation data to monitor the digital twin model in real time, and optimizes the parameters and structure of the digital twin model based on the real-time monitoring data; the simulation model indicates the virtual-real interaction between the physical entity of the Feng Shui joint adjustment system and the digital twin model;

[0079] Build the application layer: Establish an intelligent decision-making module for the Feng Shui joint adjustment system to monitor the system operation status and make Feng Shui joint adjustment control decisions;

[0080] The digital twin model of the multi-level Feng Shui coordination system of large public buildings is constructed by interactively integrating the physical layer, interaction layer, digital twin layer and application layer;

[0081] Perform consistency check on the physical entity and digital twin model of the Feng Shui coordination system: Evaluate whether the physical structure, operating status and consistency of the digital twin model with the physical entity are within the preset range. If so, the consistency check passes; otherwise, perform parameter optimization and structural optimization on the digital twin model.

[0082] In this embodiment, the evaluation of the consistency between the physical structure and operating state of the digital twin model and the object entity includes:

[0083] Establish a multi-dimensional first-level evaluation index system for consistency evaluation, including: model consistency index, simulation consistency index, data consistency index and interaction consistency index;

[0084] Establish a secondary evaluation indicator system corresponding to each indicator in the multi-dimensional primary evaluation indicator system;

[0085] After calculating the subjective and objective weights of the evaluation index system, the subjective and objective weights are integrated to obtain a comprehensive evaluation result of the consistency between the digital twin model and the physical entity.

[0086] The sovereign weight calculation adopts a fuzzy comprehensive evaluation algorithm that combines the hierarchical analysis method and the fuzzy matrix operation method; the objective weight calculation adopts the entropy weight method; the fusion of the subjective and objective weights adopts the game theory method.

[0087] In actual applications, the secondary evaluation index system corresponding to the model consistency index includes: conceptual model consistency, mathematical mechanism logic model consistency, and simulation model consistency; the secondary evaluation index system corresponding to the simulation consistency index includes: simulation algorithm consistency, simulation environment consistency, and simulation program consistency; the secondary evaluation index system corresponding to the data consistency index includes: operating data consistency, operating condition consistency, and environmental data consistency; the interaction consistency index includes data transmission speed, data transmission delay, data transmission packet loss, communication stability, and communication security.

[0088] The subjective weights are calculated by using a fuzzy comprehensive evaluation algorithm that combines the analytic hierarchy process with the fuzzy matrix operation method, including:

[0089] Using the 0.1-0.9 scaling method, the first fuzzy judgment matrix is ​​established by comparing the two indicators in the primary evaluation index system; the second fuzzy judgment matrix is ​​established by comparing the two indicators in the secondary evaluation index system;

[0090] Calculate the initial weights of the two-level indicators, expressed as:

[0091]

[0092] n is the number of rows or columns of the first fuzzy judgment matrix and the second fuzzy judgment matrix; α is the discrimination coefficient; a ik is the i-th row and k-th column element of the first fuzzy judgment matrix and the second fuzzy judgment matrix;

[0093] The subjective weight is calculated using the fuzzy analytic hierarchy process and is expressed as:

[0094]

[0095] z ij The subjective weight matrix Z = {Z1, Z2, ..., Z i}Z i The i-th row and j-th column element; r ij are elements of the first fuzzy judgment matrix and the second fuzzy judgment matrix.

[0096] In this embodiment, in S3, the functions of each user building area in each building include: business premises, office premises; the functions of each user building area in each building may cause different degrees of influence of heat load or cold load, and different functional areas require different load durations; the heat load or cold load influence is defined as:

[0097]

[0098] ε i is the influence degree of heating load or cooling load at the i-th moment. The greater the influence degree, the smaller the demand for heating load or cooling load. ij is the heating load data or cooling load data at the i-th moment of the j-th day; n is the total number of days;

[0099] The pedestrian flow density in the area includes the pedestrian flow density on working days and the pedestrian flow density on non-working days, and the pedestrian flow density is predicted by using a first machine learning algorithm;

[0100] The construction of the hourly load demand prediction model for each user building includes: using the second machine learning algorithm and the third machine learning algorithm to perform predictions respectively, and then performing weighted combination of the prediction results through a reinforcement learning algorithm to construct the hourly load demand prediction model for each user building;

[0101] The hourly load demand forecast results of each building of the large public building are obtained by cumulatively calculating the hourly load demand forecast results of each user building under the building.

[0102] It should be noted that the external environmental conditions include outdoor temperature, outdoor humidity, and outdoor wind speed; the regional environmental parameters include indoor temperature, indoor humidity, indoor wind speed, and indoor carbon dioxide concentration.

[0103] In this embodiment, the first machine learning algorithm is the ARIMA algorithm; the second machine learning algorithm is the NBEATSx time series data prediction algorithm; the third machine learning algorithm is the DELM deep extreme learning machine algorithm, and the snow melting optimization algorithm is used to optimize the parameters of the DELM algorithm; the reinforcement learning algorithm is the Q learning algorithm.

[0104] It should be noted that the NBEATSx time series data prediction algorithm is a deep neural network structure based on backward residual links, forward residual links and fully connected layer stacks. A complete NBEATSx time series data prediction algorithm structure consists of multiple stacks, and a stack consists of multiple blocks. The basic component of each block is a multi-layer fully connected network based on nonlinear activation functions; the snow ablation optimization algorithm simulates the sublimation and melting process of snow, which can be used to balance the trade-off between development and exploration of the solution space, and can avoid premature convergence.

[0105] The Q learning algorithm is a reinforcement learning algorithm that acquires experience and knowledge through continuous trial and error by the intelligent agent, establishes a Q table of state-action relationship, and continuously iterates and optimizes to obtain the best strategy to maximize the expected return. The Q table can be expressed as:

[0106] Q k+1 (s k ,a k )=Q k (s k ,a k )+α[R(s k ,s k+1 ,a k )+γmaxQ k 9s k+1 ,a′)-Q k (s k ,a k )];

[0107] Q k (s k ,a k ) is in s k Execute action a in state k Q value; R(s k ,s k+1 ,a k ) is state s k To state s k+1 The reward value of a′ is s k+1Any action value in the action space; α is the learning factor; γ is the discount factor; k is the time.

[0108] The weighted combination of prediction results by the reinforcement learning algorithm includes:

[0109] Set the prediction result weight of the NBEATSx time series data prediction algorithm to c1 and the prediction result weight of the DELM deep extreme learning machine algorithm to c2; c1+c2=1;

[0110] Establish an environment state set and an action space set, take c1 as the environment state, divide [0,1] into subsets including multiple values ​​with a preset value as the interval, as the environment state set; for example, take 0.0001 as the preset value, then there are 10001 states in the environment state set; establish a Q table with rows representing states and columns representing actions with the action space set Z = [-0.0001, 0.0001];

[0111] Establish the loss function L and reward function R, which are expressed as:

[0112]

[0113]

[0114] H(t) is the actual value of heating load or cooling load; is the predicted value of heating load or cooling load; N is the number of samples when predicting heating load or cooling load; L m The loss function value calculated for the mth selected action;

[0115] The action is selected using the action selection mechanism, and the Q table is updated to determine whether the Q table converges. If it converges and converges in subsequent iterative learning, the weight is output; otherwise, the action is selected again and the Q table is updated.

[0116] In this embodiment, S4 specifically includes:

[0117] The hourly load demand forecast results of each user building and the hourly load demand forecast results of each building are input into the digital twin model of the multi-level Feng Shui joint adjustment system of large public buildings, and Feng Shui joint adjustment simulation analysis is carried out to obtain the hourly air volume demand results, water demand results and water temperature demand results of each user building that meet the hourly load demand, and the hourly water demand results of each building;

[0118] Obtain changes in external conditions, including seasonal changes, time changes, and abnormal changes; seasonal changes include focusing on heat load and ventilation requirements in winter, focusing on cooling load and ventilation requirements in summer, and focusing on ventilation requirements in transition seasons; time changes include whether the operating time is daytime or nighttime; abnormal changes include extreme weather changes, equipment failure changes, and disaster changes;

[0119] According to the hourly wind volume demand results, water consumption demand results and water temperature demand results of each user building, the hourly water consumption demand results of each building, combined with seasonal changes, time changes and abnormal changes, a cluster analysis algorithm is used to establish a Feng Shui coordination system operation mode recognition model, and the operation mode is identified and divided; the operation mode includes a daily operation mode that meets the air volume demand, water consumption demand, and water temperature demand, an energy-saving operation mode during non-peak demand periods or transition seasons while ensuring comfort, and an emergency operation mode in extreme weather, equipment failure changes, and disasters.

[0120] It should be noted that under different operating modes, the heat load or cold load will change with the operating mode. When switching from one operating mode to another, when the load increases, the ventilation volume is adjusted first to change the regional ambient temperature. When the ambient temperature cannot be met, the water flow rate is increased to change the ambient temperature target value. When the load decreases, the water flow rate is reduced first, and then the ventilation volume is adjusted.

[0121] In this embodiment, in S5, the objectives of minimizing ventilation energy consumption, minimizing water consumption, and minimizing temperature deviations in each user building area are respectively expressed as:

[0122]

[0123]

[0124] f1 is the minimization of ventilation energy consumption; W fan,k,i is the fan energy consumption of the i-th user building area in the k-th operation mode; M is the number of user buildings; f2 is the minimization of water consumption; W unit,k,j W is the energy consumption of the jth unit under the kth operating mode; pump,k,j is the energy consumption of the jth water pump under the kth operation mode; N is the number of water-using devices; f3 is the minimum temperature deviation of each user's building area; T set,k,i is the set temperature of the i-th user building area under the k-th operating mode, and the temperature is set according to different operating modes; T room,k,i The actual temperature of the i-th user building area under the k-th operating mode;

[0125] The setting of the multi-level control variable combination target value of the Feng Shui coordination system that adapts to each operating mode includes: solving each target through an intelligent optimization algorithm to obtain the control variable combination target value for adjusting water flow, water temperature and ventilation volume in the multi-level Feng Shui coordination system that adapts to each operating mode.

[0126] In this embodiment, S6 specifically includes:

[0127] Constructing a mathematical model of the Markov decision process for equipment control of the Feng Shui joint control system, including setting a state space, an action space, a reward function, a state transition probability and a discount factor; the state space represents the equipment operating state parameters and environmental information of the Feng Shui joint control system; the action space represents the multi-level equipment control variable actions of the Feng Shui joint control system, including the chiller operating power, the boiler operating power, the electric regulating valve opening, the air conditioning equipment operating power, the water pump operating frequency, the valve opening and the fan operating frequency; the reward function represents the minimum deviation between the target value and the actual value of the multi-level control variable combination of the Feng Shui joint control system in each operating mode; the state transition probability is the probability that the state is transferred from the current state to another state after executing the equipment control variable action in the action space; the discount factor represents the reward attenuation factor;

[0128] The Actor network in the deep reinforcement learning TD3 algorithm interacts with the environment to obtain the action at the time t, and obtains the state and reward at the next moment. Then, the Critic network and the Actor network are updated by taking out the state, action, reward at the time t and the state at the next moment from the experience replay pool, and the target Actor network and the target Critic network are updated regularly. Finally, after continuous iterative updates, the network strategy is trained in the direction of higher reward values, and the multi-level equipment control action instructions of the Feng Shui joint control system are obtained, and sent to the relevant equipment at each level for execution.

[0129] In this embodiment, the deep reinforcement learning TD3 algorithm structure includes an online network and a target network; the online network includes an online Actor network and two online Critic networks; the target network includes a target Actor network and two target Critic networks.

[0130] It should be noted that the online critic network is used to calculate the current Q value and update its own network parameters. The target critic network is used to calculate the target Q value, and its network parameters are updated regularly. The Q value refers to the value of performing an action in the current state. In each iterative learning process, the Q value of the action with the maximum current value is estimated, which is greater than the Q value of the action that actually interacts with the environment. Therefore, it will lead to the problem of overestimation of the Q value. The TD3 algorithm uses two critic networks to estimate the Q value separately and selects the smaller Q value as the target value.

[0131] The input of the online Actor network is the state of the environment, and the output is the action. The input of the target Actor network is the state of the environment at the next moment, and the output is the action under that state. The two online Critic networks output corresponding Q values ​​according to the state and action respectively. The TD3 algorithm adopts the soft update method of the target network, that is, the target network is updated after the current network is updated several times, and the delayed update is used to update the Actor network. The action is executed according to the output of the Actor network first, and the Actor network is updated after the Critic network is updated a set number of times.

[0132] In several embodiments provided in the present application, it should be understood that the disclosed system and method can also be implemented in other ways. The system embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the systems, methods and computer program products according to multiple embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of the code, and a part of the module, program segment or code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0133] In addition, the functional modules in each embodiment of the present invention can be integrated together to form an independent part, or each module can exist separately, or two or more modules can be integrated to form an independent part. If the function is implemented in the form of a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, RandomAccess Memory), disk or optical disk and other media that can store program codes.

[0134] Based on the above ideal embodiments of the present invention, the relevant staff can make various changes and modifications without departing from the technical concept of the present invention through the above description. The technical scope of the present invention is not limited to the contents of the specification, and its technical scope must be determined according to the scope of the claims.

Claims

1. A feng shui joint adjustment method for large public buildings using digital twins and deep reinforcement learning, characterized in that: The large-scale public building feng shui joint adjustment method comprises: S1. Install electric regulating valves for regulating water flow in front of each building at the end of a large public building, install air conditioning and ventilation devices for variable frequency adjustment of air volume in front of each user in each building, and install air conditioning water systems for regulating user water flow and water temperature, and combine the energy center on the source side and the heat exchange station on the station side to form a multi-level wind and water joint adjustment system for large public buildings; S2. Construct a digital twin model of the multi-level Feng Shui coordination system of large public buildings, including the physical layer, interaction layer, digital twin layer and application layer, and verify the consistency between the digital twin model and the actual system physical entity; S3. Based on the verified digital twin model of the multi-level Feng Shui coordination system of large public buildings, the hourly load demand prediction model of each user building is constructed considering the function, location, population density, external environmental conditions and environmental parameters of each user building area in each building, and the hourly load demand prediction results of each building of large public buildings are calculated; S4. Based on the hourly load demand forecast results of each user building and the hourly load demand forecast results of each building, analyze the hourly wind volume demand results, water demand results and water temperature demand results of each user building, and the hourly water demand results of each building, and divide the operation mode of the multi-level Feng Shui joint adjustment system of large public buildings in combination with changes in external conditions; S5. According to the operation modes of the multi-level Feng Shui joint adjustment system of large public buildings, with the goal of minimizing ventilation energy consumption, water consumption and temperature deviation of each user building area, set the target value of the multi-level control variable combination of the Feng Shui joint adjustment system that adapts to each operation mode; S6. According to the target value of the multi-level control variable combination of the Feng Shui coordination system in each operating mode, a deep reinforcement learning algorithm is used to calculate the multi-level equipment control instructions of the Feng Shui coordination system, and the instructions are sent to the relevant equipment at each level for execution.

2. The feng shui joint adjustment method for large public buildings according to claim 1 is characterized in that: In step S1, the source-side energy center includes at least a plurality of chillers for cooling and a plurality of boilers for heating; the number of the station-side heat exchange stations is multiple, and each heat exchange station is connected to the air conditioning and ventilation device and air conditioning water system of each user; The air conditioning and ventilation device comprises: a fan and a fan coil; The air conditioning water system includes: air conditioning equipment, water distributor, water collector, valve and water pump; The regulation strategy of the multi-level Feng Shui joint regulation system of large public buildings includes: first, the hot water supplied by the boiler of the energy center on the source side or the chilled water supplied by the chiller is distributed to each building through the regulating electric regulating valve, and then the water pump in the air-conditioning water system sends the hot water or chilled water to the water distributor, and then the water distributor sends it to the air-conditioning equipment of each user through each water supply branch pipe, and after the water temperature treatment of the air-conditioning equipment and the water flow adjustment of the valve, it flows back to the water collector through each return branch pipe and returns to the boiler or chiller through the water collector, and the cycle is repeated to change the water flow and water temperature, and finally flows through the fan coil units of each user's building area, and then the fan frequency conversion adjusts the air volume to adjust the temperature distribution and air flow of each user's building area.

3. The feng shui joint adjustment method for large public buildings according to claim 1 is characterized in that: The S2 specifically includes: Construct the physical layer: Use multi-source sensors to sense the equipment operating conditions and environmental information of the multi-level Feng Shui joint adjustment system of large public buildings and transmit them to the interactive layer, and realize interactive mapping with the digital twin layer through interactive feedback of multi-source data, and control the equipment of the Feng Shui joint adjustment system according to the application layer; the equipment operating conditions include the operating status parameters of each device in the Feng Shui joint adjustment system; the environmental information includes outdoor weather temperature, weather humidity, building area pedestrian flow and building area temperature; Build the interaction layer: Transmit the equipment operating conditions and environmental information collected by multi-source sensors to the application layer, and establish state synchronization and interactive collaborative control between the digital twin layer and the physical entity through data preprocessing, data fusion, and data mining; Constructing the digital twin layer: constructing a digital twin model that integrates the physical model, mechanism model, data model, and simulation model; the physical model indicates the structural characteristics of the physical entity of the Feng Shui joint adjustment system of large public buildings; the mechanism model indicates the logical description and mechanism explanation of the physical entity when performing Feng Shui joint adjustment; the data model includes the perceived equipment operating conditions, environmental information, and simulation operation data of the physical entity of the Feng Shui joint adjustment system, and uses the simulation operation data to monitor the digital twin model in real time, and optimizes the parameters and structure of the digital twin model based on the real-time monitoring data; the simulation model indicates the virtual-real interaction between the physical entity of the Feng Shui joint adjustment system and the digital twin model; Build the application layer: Establish an intelligent decision-making module for the Feng Shui joint adjustment system to monitor the system operation status and make Feng Shui joint adjustment control decisions; The digital twin model of the multi-level Feng Shui coordination system of large public buildings is constructed by interactively integrating the physical layer, interaction layer, digital twin layer and application layer; Perform consistency check on the physical entity and digital twin model of the Feng Shui coordination system: Evaluate whether the physical structure, operating status and consistency of the digital twin model with the physical entity are within the preset range. If so, the consistency check passes; otherwise, perform parameter optimization and structural optimization on the digital twin model.

4. The method for joint adjustment of Feng Shui of large public buildings according to claim 3 is characterized in that: The evaluation of the physical structure and operating status of the digital twin model and the consistency of the object entity includes: Establish a multi-dimensional first-level evaluation index system for consistency evaluation, including: model consistency index, simulation consistency index, data consistency index and interaction consistency index; Establish a secondary evaluation indicator system corresponding to each indicator in the multi-dimensional primary evaluation indicator system; After calculating the subjective and objective weights of the evaluation index system, the subjective and objective weights are integrated to obtain a comprehensive evaluation result of the consistency between the digital twin model and the physical entity. The sovereign weight calculation adopts a fuzzy comprehensive evaluation algorithm that combines the hierarchical analysis method and the fuzzy matrix operation method; the objective weight calculation adopts the entropy weight method; the fusion of the subjective and objective weights adopts the game theory method.

5. The method for joint adjustment of Feng Shui for large public buildings according to claim 1 is characterized in that: In S3, the functions of each user building area in each building include: business premises, office premises; the functions of each user building area in each building may cause different degrees of influence of heat load or cold load, and different functional areas may require different load durations; the heat load or cold load influence is defined as: ε i is the influence degree of heating load or cooling load at the i-th moment. The greater the influence degree, the smaller the demand for heating load or cooling load. ij is the heating load data or cooling load data at the i-th moment of the j-th day; n is the total number of days; The pedestrian flow density in the area includes the pedestrian flow density on working days and the pedestrian flow density on non-working days, and the pedestrian flow density is predicted by using a first machine learning algorithm; The construction of the hourly load demand prediction model for each user building includes: using the second machine learning algorithm and the third machine learning algorithm to perform predictions respectively, and then performing weighted combination of the prediction results through a reinforcement learning algorithm to construct the hourly load demand prediction model for each user building; The hourly load demand forecast results of each building of the large public building are obtained by cumulatively calculating the hourly load demand forecast results of each user building under the building.

6. The method for joint adjustment of Feng Shui for large public buildings according to claim 5 is characterized in that: The first machine learning algorithm is the ARIMA algorithm; the second machine learning algorithm is the NBEATSx time series data prediction algorithm; the third machine learning algorithm is the DELM deep extreme learning machine algorithm, and the snow melting optimization algorithm is used to optimize the parameters of the DELM algorithm; the reinforcement learning algorithm is the Q learning algorithm.

7. The method for joint adjustment of Feng Shui for large public buildings according to claim 1 is characterized in that: The S4 specifically includes: The hourly load demand forecast results of each user building and the hourly load demand forecast results of each building are input into the digital twin model of the multi-level Feng Shui joint adjustment system of large public buildings, and Feng Shui joint adjustment simulation analysis is carried out to obtain the hourly air volume demand results, water demand results and water temperature demand results of each user building that meet the hourly load demand, and the hourly water demand results of each building; Obtain changes in external conditions, including seasonal changes, time changes, and abnormal changes; seasonal changes include focusing on heat load and ventilation requirements in winter, focusing on cooling load and ventilation requirements in summer, and focusing on ventilation requirements in transition seasons; time changes include whether the operating time is daytime or nighttime; abnormal changes include extreme weather changes, equipment failure changes, and disaster changes; According to the hourly wind volume demand results, water consumption demand results and water temperature demand results of each user building, the hourly water consumption demand results of each building, combined with seasonal changes, time changes and abnormal changes, a cluster analysis algorithm is used to establish a Feng Shui coordination system operation mode recognition model, and the operation mode is identified and divided; the operation mode includes a daily operation mode that meets the air volume demand, water consumption demand, and water temperature demand, an energy-saving operation mode during non-peak demand periods or transition seasons while ensuring comfort, and an emergency operation mode in extreme weather, equipment failure changes, and disasters.

8. The method for joint adjustment of Feng Shui for large public buildings according to claim 1, characterized in that: In S5, the objectives of minimizing ventilation energy consumption, minimizing water consumption, and minimizing temperature deviations in each user's building area are respectively expressed as: f1 is the minimization of ventilation energy consumption; W fan,k,i is the fan energy consumption of the i-th user building area in the k-th operation mode; M is the number of user buildings; f2 is the minimization of water consumption; W unit,k,j W is the energy consumption of the jth unit under the kth operating mode; pump,k,j is the energy consumption of the jth water pump under the kth operation mode; N is the number of water-using devices; f3 is the minimum temperature deviation of each user's building area; T set,k,i is the set temperature of the i-th user building area under the k-th operating mode, and the temperature is set according to different operating modes; T room,k,i The actual temperature of the i-th user building area under the k-th operating mode; The setting of the multi-level control variable combination target value of the Feng Shui coordination system that adapts to each operating mode includes: solving each target through an intelligent optimization algorithm to obtain the control variable combination target value for adjusting water flow, water temperature and ventilation volume in the multi-level Feng Shui coordination system that adapts to each operating mode.

9. The method for joint adjustment of Feng Shui for large public buildings according to claim 1, characterized in that: The S6 specifically includes: Constructing a mathematical model of the Markov decision process for equipment control of the Feng Shui joint control system, including setting a state space, an action space, a reward function, a state transition probability and a discount factor; the state space represents the equipment operating state parameters and environmental information of the Feng Shui joint control system; the action space represents the multi-level equipment control variable actions of the Feng Shui joint control system, including the chiller operating power, the boiler operating power, the electric regulating valve opening, the air conditioning equipment operating power, the water pump operating frequency, the valve opening and the fan operating frequency; the reward function represents the minimum deviation between the target value and the actual value of the multi-level control variable combination of the Feng Shui joint control system in each operating mode; the state transition probability is the probability that the state is transferred from the current state to another state after executing the equipment control variable action in the action space; the discount factor represents the reward attenuation factor; The Actor network in the deep reinforcement learning TD3 algorithm interacts with the environment to obtain the action at the time t, and obtains the state and reward at the next moment. Then, the Critic network and the Actor network are updated by taking out the state, action, reward at the time t and the state at the next moment from the experience replay pool, and the target Actor network and the target Critic network are updated regularly. Finally, after continuous iterative updates, the network strategy is trained in the direction of higher reward values, and the multi-level equipment control action instructions of the Feng Shui joint control system are obtained, and sent to the relevant equipment at each level for execution.

10. The method for joint adjustment of Feng Shui of large public buildings according to claim 9, characterized in that: The deep reinforcement learning TD3 algorithm structure includes an online network and a target network; the online network includes an online Actor network and two online Critic networks; the target network includes a target Actor network and two target Critic networks.

Images