Indoor environment regulation and control system based on multiple intelligent systems

Abstract

The invention discloses an indoor environment regulation and control system based on a multi-intelligent system, and belongs to the technical field of indoor environment regulation and control, and the system comprises a data obtaining subsystem which is used for obtaining multi-dimensional environment data of an indoor region based on an inductor intelligent agent; the data processing subsystem is used for preprocessing the multi-dimensional environment data of the indoor area to generate a data set; the model construction subsystem is used for constructing a bidirectional long-short-term memory network model and training the bidirectional long-short-term memory network model based on the historical data set to obtain an environment regulation and control model; the calculation subsystem is used for inputting the data set into an environment regulation and control model to generate a regulation and control strategy; and the execution subsystem is used for feeding back the regulation and control result to the actuator intelligent agent for execution. Through optimized operation of the intelligent regulation and control system, the comfort level of the indoor environment is improved, and unnecessary energy waste can be reduced and energy efficiency maximization can be achieved by accurately controlling work of environment equipment.

Description

Technical Field

[0001] The present invention belongs to the technical field of indoor environment control, and in particular relates to an indoor environment control system based on a multi-intelligent system. Background Art

[0002] A multi-agent system consists of multiple interacting agents, each of which can operate independently and respond to changes in the environment it perceives. This structure enables the system to flexibly adapt to different environmental conditions and user needs in different indoor areas, such as temperature, light, and humidity, and can automatically adjust its behavior to best suit the current environmental state. Multi-agent-based systems are easy to expand. More agents can simply be added as needed to cover larger areas or handle more environmental factors. Each agent can operate independently as a module without affecting other parts of the system, which improves the overall robustness and maintainability of the system.

[0003] However, due to the dynamic changes and uncertainty of the indoor environment, directional control has brought difficulties. How to achieve real-time perception and rapid response of the intelligent agent to environmental changes, and continuously optimize the control strategy to adapt to new environmental conditions is a technical problem that needs to be solved urgently. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes an indoor environment control system based on a multi-intelligent system to solve the problems existing in the above prior art.

[0005] To achieve the above object, the present invention provides an indoor environment control system based on a multi-intelligent system, comprising:

[0006] A data acquisition subsystem, used to acquire multi-dimensional environmental data of indoor areas based on sensor agents;

[0007] A data processing subsystem, connected to the data acquisition subsystem, for preprocessing the multi-dimensional environmental data of the indoor area to generate a data set of environmental parameter distribution characteristics of each area;

[0008] A model building subsystem, connected to the data processing subsystem, for building a bidirectional long short-term memory network model, training the bidirectional long short-term memory network model based on a historical data set, and obtaining an environmental control model;

[0009] A computing subsystem connected to the model building subsystem, used to input the environmental parameter distribution characteristic data set into the environmental control model to generate a control strategy;

[0010] The execution subsystem is connected to the computing subsystem and is used to feed back the control result to the actuator agent for execution.

[0011] Preferably, the data acquisition subsystem includes:

[0012] A deployment module, used to partition the indoor environment and deploy the sensor agent based on the partition result;

[0013] The acquisition module is used to acquire data of the indoor environment based on the sensor agent and generate multi-dimensional environmental data of the indoor area.

[0014] Preferably, the acquisition module is further used to: construct a wireless sensor network, and transmit the multi-dimensional environmental data of the indoor area to the data processing subsystem through the Zigbee protocol based on the wireless sensor network.

[0015] Preferably, the data processing subsystem includes:

[0016] An exception processing module, used to remove abnormal values ​​and noise from the multi-dimensional environmental data of the indoor area, and obtain the distribution characteristics of the environmental parameters of each area;

[0017] A normalization module, used to process the distribution characteristics of the environmental parameters of each region based on a maximum and minimum value normalization method to generate normalized data;

[0018] A vector acquisition module, used to acquire key features of regional environmental parameter distribution characteristics in the normalized data, and construct a feature vector based on the key features;

[0019] The clustering module is used to cluster and group the feature vectors using Euclidean distance as a similarity metric to generate a distribution feature data set of environmental parameters for each region.

[0020] Preferably, the model building subsystem includes:

[0021] Model generation module, used to build a bidirectional long short-term memory network model;

[0022] The training set generation module is used to obtain the position coordinates, capability parameters and status information of each intelligent agent, construct the intelligent agent attribute vector, group the intelligent agents based on the clustering algorithm, dynamically adjust the fitness weight of the intelligent agent through the optimization algorithm, generate the intelligent agent data set, and obtain the environment data set corresponding to the intelligent agent;

[0023] A model training module is used to train the bidirectional long short-term memory network model based on the agent data set and the environment data set to obtain the environment control model.

[0024] Preferably, the computing subsystem comprises:

[0025] A result generation module is used to input the environmental parameter distribution characteristic data set into the environmental control model to estimate the environmental trend in the future;

[0026] A control strategy generation module is used to generate the control strategy based on the environmental trend in the future time.

[0027] Preferably, the execution subsystem includes:

[0028] A control module, used to feed back the control strategy to the actuator agent for dynamic optimization of the environment;

[0029] A feedback module is used to feed back the adjusted environmental information to the environmental control model during the dynamic optimization process to optimize the parameters of the environmental control model.

[0030] Preferably, the feedback module is also used to: analyze the input environmental parameter change data through the environmental control model, estimate the environmental trend in the future, and determine whether the environmental trend prediction result obtained by the prediction model matches the currently implemented control strategy. If the prediction result does not match the current control strategy, the dynamic optimization process of the control strategy is triggered.

[0031] Compared with the prior art, the present invention has the following advantages and technical effects:

[0032] The present invention uses a data acquisition subsystem and a sensor agent to acquire multi-dimensional data from the indoor environment, so that the system can obtain comprehensive environmental perception. The data processing subsystem preprocesses the collected data and generates a distribution feature data set of environmental parameters in each area. The model building subsystem uses a bidirectional long short-term memory network model, which can effectively capture the front-to-back dependencies in time series data, thereby improving the accuracy of prediction. The system trains the model based on historical data sets and can predict and adjust environmental parameters to achieve the optimal state. The computing subsystem inputs the environmental parameter feature data set into the trained environmental control model to generate specific control instructions. The execution subsystem then transmits these control instructions to each actuator agent to achieve precise control of the indoor environment. This process supports real-time response and adjustment to ensure that the indoor environment is always in the optimal state set by the user. Through the optimized operation of the intelligent control system, not only the comfort of the indoor environment is improved, but also unnecessary energy waste can be reduced by accurately controlling the operation of environmental equipment, thereby maximizing energy efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0034] Figure 1 Schematic diagram of an indoor environment control system based on a multi-intelligent system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0035] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0036] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0037] Embodiment 1

[0038] like Figure 1 As shown, in this embodiment, a multi-intelligent system-based indoor environment control system is provided, including:

[0039] A data acquisition subsystem, used to acquire multi-dimensional environmental data of indoor areas based on sensor agents;

[0040] Further optimizing the scheme, the data acquisition subsystem includes:

[0041] A deployment module, used to partition the indoor environment and deploy the sensor agent based on the partition result;

[0042] The acquisition module is used to acquire data of the indoor environment based on the sensor agent and generate multi-dimensional environmental data of the indoor area.

[0043] According to a further optimization scheme, the acquisition module is further used to: construct a wireless sensor network, and transmit the multi-dimensional environmental data of the indoor area to the data processing subsystem through the Zigbee protocol based on the wireless sensor network.

[0044] An environmental sensor network is deployed in each area to collect multi-dimensional environmental data such as temperature, humidity, and light in the area, and transmit the collected environmental data to the cloud data center; the environmental data of each area received by the cloud data center is summarized and analyzed to obtain the distribution characteristics of the environmental parameters of each area.

[0045] According to the regional division plan, an environmental sensor network is deployed in each area. The sensors include temperature sensors, humidity sensors, and light sensors, etc., which are used to collect multi-dimensional environmental parameter data of the area;

[0046] The environmental data collected by each sensor is transmitted to the cloud data center in real time through the wireless communication network, and the cloud data center receives the environmental data reported by the sensors in each area;

[0047] The cloud data center pre-processes the received environmental data of each region, including data cleaning, denoising and normalization, to improve data quality;

[0048] According to the preset data aggregation rules, the pre-processed environmental data are aggregated to obtain data sets of different environmental parameters in each region;

[0049] Clustering algorithms are used to perform cluster analysis on the environmental parameter data sets of each region, and the distribution range of the environmental parameters of each region is divided according to the clustering results;

[0050] Specifically, when deploying environmental sensor networks in each area, the Zigbee wireless communication protocol can be used, with a sensor node density of 10 per 100 square meters and a sensor sampling frequency of once per minute. The environmental data collected by each sensor is transmitted to the regional aggregation node in real time through the Zigbee network, and then transmitted to the cloud data center through the 4G network.

[0051] A data processing subsystem, connected to the data acquisition subsystem, for preprocessing the multi-dimensional environmental data of the indoor area to generate a data set of environmental parameter distribution characteristics of each area;

[0052] Further optimizing the scheme, the data processing subsystem includes:

[0053] An exception processing module, used to remove abnormal values ​​and noise from the multi-dimensional environmental data of the indoor area, and obtain the distribution characteristics of the environmental parameters of each area;

[0054] A normalization module, used to process the distribution characteristics of the environmental parameters of each region based on a maximum and minimum value normalization method to generate normalized data;

[0055] A vector acquisition module, used to acquire key features of regional environmental parameter distribution characteristics in the normalized data, and construct a feature vector based on the key features;

[0056] The clustering module is used to cluster and group the feature vectors using Euclidean distance as a similarity metric to generate a distribution feature data set of environmental parameters for each region.

[0057] The acquired environmental parameter data is preprocessed to remove outliers and noise, and the distribution characteristics of environmental parameters in each area are obtained. According to the preset environmental comfort threshold, it is judged whether the environmental parameters of each area are within the comfortable range. If they exceed the threshold range, they are marked as uncomfortable areas. The K-means clustering algorithm is used to group the areas, and the areas with similar environmental parameter distribution characteristics are divided into the same group to obtain several regional groups. For different regional groups, differentiated control strategies are formulated according to the characteristics of their environmental parameters. For example, for the regional group with high temperature, cooling measures such as enhanced ventilation and shading can be adopted. On the basis of the control strategy, the control targets of each regional group are further determined, such as setting the target temperature of the high temperature area to 26 degrees Celsius and the target humidity of the low humidity area to 50%. The formulated control strategies and targets are sent to the control equipment of each area, and the dynamic adjustment of the regional environmental parameters is realized through the automatic control system to meet the comfort requirements.

[0058] Specifically, wireless sensor networks are deployed in each area, and sensor nodes collect real-time data of environmental parameters such as temperature, humidity, and light, and transmit the data to the central control system through the Zigbee protocol. The control system uses the Kalman filter algorithm to preprocess the sensor data, remove outliers and high-frequency noise that exceed the normal range by 20%, and obtain the statistical distribution characteristics of environmental parameters in each area, such as temperature concentrated at 22-28℃ and humidity concentrated at 40%-60%. The system presets the environmental comfort threshold of temperature 18-26℃, humidity 40%-70%, and illumination 500-1000lux. According to the threshold judgment, 5 temperature discomfort areas and 3 humidity discomfort areas are marked. The K-means clustering algorithm is used to cluster and group the environmental parameter characteristics of each area with Euclidean distance as the similarity metric. The optimal clustering number K=4 is obtained through the Calinski-Harabasz index evaluation, and each area is divided into high temperature and low humidity group, low temperature and high humidity group, high temperature and high humidity group, and low temperature and low humidity group.

[0059] Specifically, after obtaining the environmental parameter data of each region, the data is first preprocessed. For missing values, the mean filling or median filling method can be used for processing. For example, for temperature data, if the temperature value at a certain time point is missing, it can be filled with the mean temperature of other time points in the region on the same day. For outliers, the box plot method can be used for identification and processing, and data that exceeds the upper and lower quartiles by more than 5 times is regarded as an outlier and replaced with the median. Then, the data is normalized to unify the numerical range of each environmental parameter into the [0,1] interval to eliminate the dimension effect. Normalization can use the maximum and minimum value normalization method, that is, subtract the minimum value of the parameter from each data, and then divide it by the difference between the maximum and minimum values. After obtaining the standardized data set, the key features that can reflect the distribution characteristics of regional environmental parameters, such as the mean and variance of temperature, the mean and variance of humidity, etc., are extracted to construct feature vectors. Then, the K-means clustering algorithm is used to perform cluster analysis on the feature vectors. The K-means algorithm divides the closest regions into the same group by calculating the Euclidean distance between each region's feature vector and the cluster center. In order to determine the optimal number of groups, different K values ​​can be set, such as 2, 3, 4, 5, etc., run multiple clusterings, and calculate evaluation indicators such as the silhouette coefficient and Davies-Bouldin index. The closer the silhouette coefficient is to 1, the better the clustering effect; the smaller the Davies-Bouldin index is, the better the clustering effect is. For example, when K = 3, the silhouette coefficient is 8 and the Davies-Bouldin index is 2; when K = 4, the silhouette coefficient is 85 and the Davies-Bouldin index is 1, then K = 4 is selected as the optimal number of groups. According to the clustering results corresponding to the optimal number of groups, each region is divided into several groups. The clustering results can be displayed in a visual way such as a scatter plot and a radar chart to intuitively present the grouping of each region and the similarity within the group.

[0060] A model building subsystem, connected to the data processing subsystem, for building a bidirectional long short-term memory network model, training the bidirectional long short-term memory network model based on a historical data set, and obtaining an environmental control model;

[0061] Further optimizing the scheme, the model building subsystem includes:

[0062] Model generation module, used to build a bidirectional long short-term memory network model;

[0063] The training set generation module is used to obtain the location coordinates, capability parameters and status information of each intelligent agent, construct the intelligent agent attribute vector, group the intelligent agents based on the clustering algorithm, and dynamically adjust the fitness weight of the intelligent agent through the optimization algorithm to generate the intelligent agent data set, and at the same time obtain the environmental data set corresponding to the intelligent agent; obtain the location information of each intelligent agent through sensors and positioning devices, such as using GPS positioning technology to determine the precise coordinates of each intelligent agent, and collect its capability parameters such as load capacity, speed, etc., and the current working status, such as the remaining percentage of power, through the wireless communication module. These data are sent to the central processing system, and the decision tree algorithm is used to score the fitness of the intelligent agent. The algorithm will take into account the location distance, capability matching and current status

[0064] A model training module is used to train the bidirectional long short-term memory network model based on the agent data set and the environment data set to obtain the environment control model.

[0065] The calculation formula for extracting features of the bidirectional long short-term memory network model is:

[0066]

[0067] in, is the output vector of the forward LSTM hidden layer at time t, (t=1, 2, ..., N), which is the input vector x at the current time t and the forward LSTM output vector at the previous moment jointly determined;

[0068] in, is the output vector of the reverse LSTM hidden layer at time t, (t=1, 2, ..., N), which is the input vector x at the current time t and the reverse LSTM output vector of the previous moment jointly determined;

[0069] h t is the output of the BiLSTM model, ω t is the weight matrix of the forward LSTM output; v t is the weight matrix of the reverse LSTM output; bt is the bias of the weight matrix.

[0070] A computing subsystem connected to the model building subsystem, used to input the environmental parameter distribution characteristic data set into the environmental control model to generate a control strategy;

[0071] Further optimizing the scheme, the computing subsystem includes:

[0072] A result generation module is used to input the environmental parameter distribution characteristic data set into the environmental control model to estimate the environmental trend in the future;

[0073] A control strategy generation module is used to generate the control strategy based on the environmental trend in the future time.

[0074] Use the trained machine learning model to predict changes in environmental parameters in the future. Based on the prediction results, determine the changing trend of future environmental parameters. If the predicted value exceeds the preset threshold, warn of environmental deterioration. Automatically trigger predetermined response measures for the warned environmental deterioration trend, such as increasing monitoring frequency and initiating emergency plans. Continuously track the actual changes in environmental parameters, and feed the actual data back to the machine learning model to achieve online updating and optimization of the model.

[0075] The execution subsystem is connected to the computing subsystem and is used to feed back the control result to the actuator agent for execution.

[0076] Further optimizing the solution, the execution subsystem includes:

[0077] A control module, used to feed back the control strategy to the actuator agent for dynamic optimization of the environment;

[0078] A feedback module is used to feed back the adjusted environmental information to the environmental control model during the dynamic optimization process to optimize the parameters of the environmental control model.

[0079] To further optimize the solution, the feedback module is also used to: analyze the input environmental parameter change data through the environmental control model, estimate the environmental trend in the future, and determine whether the environmental trend prediction results obtained by the prediction model match the currently implemented control strategy. If the prediction results do not match the current control strategy, the dynamic optimization process of the control strategy is triggered.

[0080] Specifically, in practical applications, high-precision environmental sensors deployed at urban traffic intersections are first used to obtain environmental parameters such as air quality, temperature, and humidity in real time. These data are updated every minute and transmitted to the central processing system via a wireless network. The central processing system uses a time series-based prediction model, such as a long short-term memory network (LSTM), to analyze the received environmental parameter data and predict environmental change trends within the next hour. For example, the model may predict that the PM5 concentration will increase by 10% within the next hour. The system then compares this prediction result with the current air quality improvement strategy, such as increasing the frequency of watering urban green spaces to reduce dust. If the prediction result shows that the PM5 concentration is rising, this strategy is obviously not enough to cope with the predicted changes. At this point, the system automatically triggers the dynamic optimization process of the control strategy, and uses reinforcement learning algorithms, such as deep Q networks (DQN), to adaptively adjust the strategy, such as increasing the working intensity and frequency of air purification equipment. Through multiple rounds of iterations, the system learns and optimizes the control strategy, and finally forms a new strategy, such as increasing the operating time of air purification equipment from 8 hours a day to 12 hours, and monitors its effect in real time. The optimized control strategy is then deployed to the smart city management system and dynamically adjusted based on real-time updated environmental data to ensure that the strategy can effectively adapt to future environmental changes, thereby effectively improving urban air quality.

[0081] This embodiment may also include: establishing an energy consumption model according to the energy consumption data of the intelligent body to obtain an energy consumption feature vector; using a reinforcement learning algorithm, taking the energy consumption feature vector as the state input and the energy management strategy of the intelligent body as the action output, and obtaining the optimal energy management strategy through trial and error learning; controlling the energy utilization of the intelligent body according to the optimal energy management strategy, including adjusting the sensor sampling frequency, optimizing the data transmission method, reducing the processor main frequency, etc., to reduce the overall energy consumption; in the energy management process, continuously monitoring the energy consumption of the intelligent body, if the energy consumption exceeds the preset threshold, triggering the reinforcement learning algorithm to learn and optimize again, and updating the energy management strategy; through the continuous optimization of the energy efficiency management strategy, the energy consumption of the intelligent body is minimized, while ensuring that its performance meets the task requirements, thereby extending the service life of the intelligent body; for different types of intelligent bodies, such as mobile robots, drones, etc., establishing personalized energy consumption models, formulating differentiated energy management strategies, and further improving energy utilization efficiency; associating the energy consumption of the intelligent body with its task execution, statistically analyzing the correlation between energy consumption and performance, providing data support and decision-making basis for subsequent energy efficiency management, and achieving global optimal energy utilization.

[0082] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. An indoor environment control system based on a multi-intelligent system, characterized in that: include: A data acquisition subsystem, used to acquire multi-dimensional environmental data of indoor areas based on sensor agents; A data processing subsystem, connected to the data acquisition subsystem, for preprocessing the multi-dimensional environmental data of the indoor area to generate a data set of environmental parameter distribution characteristics of each area; A model building subsystem, connected to the data processing subsystem, for building a bidirectional long short-term memory network model, training the bidirectional long short-term memory network model based on a historical data set, and obtaining an environmental control model; A computing subsystem connected to the model building subsystem, used to input the environmental parameter distribution characteristic data set into the environmental control model to generate a control strategy; The execution subsystem is connected to the computing subsystem and is used to feed back the control result to the actuator agent for execution.

2. The indoor environment control system based on multi-intelligent system according to claim 1 is characterized in that: The data acquisition subsystem includes: A deployment module, used to partition the indoor environment and deploy the sensor agent based on the partition result; The acquisition module is used to acquire data of the indoor environment based on the sensor agent and generate multi-dimensional environmental data of the indoor area.

3. The indoor environment control system based on multi-intelligent system according to claim 2 is characterized in that: The acquisition module is also used to: construct a wireless sensor network, and based on the wireless sensor network, transmit the multi-dimensional environmental data of the indoor area to the data processing subsystem through the Zigbee protocol.

4. The indoor environment control system based on multi-intelligent system according to claim 1 is characterized in that: The data processing subsystem comprises: An exception processing module, used to remove abnormal values ​​and noise from the multi-dimensional environmental data of the indoor area, and obtain the distribution characteristics of the environmental parameters of each area; A normalization module, used to process the distribution characteristics of the environmental parameters of each region based on a maximum and minimum value normalization method to generate normalized data; A vector acquisition module, used to acquire key features of regional environmental parameter distribution characteristics in the normalized data, and construct a feature vector based on the key features; The clustering module is used to cluster and group the feature vectors using Euclidean distance as a similarity metric to generate a distribution feature data set of environmental parameters for each region.

5. The indoor environment control system based on multi-intelligent system according to claim 1 is characterized in that: The model building subsystem includes: Model generation module, used to build a bidirectional long short-term memory network model; The training set generation module is used to obtain the position coordinates, capability parameters and status information of each intelligent agent, construct the intelligent agent attribute vector, group the intelligent agents based on the clustering algorithm, dynamically adjust the fitness weight of the intelligent agent through the optimization algorithm, generate the intelligent agent data set, and obtain the environment data set corresponding to the intelligent agent; A model training module is used to train the bidirectional long short-term memory network model based on the agent data set and the environment data set to obtain the environment control model.

6. The indoor environment control system based on multi-intelligent system according to claim 1 is characterized in that: The computing subsystem includes: A result generation module is used to input the environmental parameter distribution characteristic data set into the environmental control model to estimate the environmental trend in the future; A control strategy generation module is used to generate the control strategy based on the environmental trend in the future time.

7. The indoor environment control system based on multi-intelligent system according to claim 1 is characterized in that: The execution subsystem includes: A control module, used to feed back the control strategy to the actuator agent for dynamic optimization of the environment; A feedback module is used to feed back the adjusted environmental information to the environmental control model during the dynamic optimization process to optimize the parameters of the environmental control model.

8. The indoor environment control system based on multi-intelligent system according to claim 7 is characterized in that: The feedback module is also used to: analyze the input environmental parameter change data through the environmental control model, estimate the environmental trend in the future, and determine whether the environmental trend prediction result obtained by the prediction model matches the currently implemented control strategy. If the prediction result does not match the current control strategy, the dynamic optimization process of the control strategy is triggered.

Images