Building energy consumption monitoring system and method based on Internet of Things data acquisition

Abstract

The invention relates to the technical field of building energy consumption monitoring, in particular to a building energy consumption monitoring system and method based on data acquisition of the Internet of Things, and the system comprises a data collection module, a strategy generator module and a calculation and updating module, generating a state vector of the current time step; the strategy generator module is used for generating a specific management strategy according to the state vector of the current time step; the method comprises the following steps: setting a neural network as a strategy generator, initializing the weight of the neural network in the strategy generator, and generating a current optimal strategy by using an exploration strategy method; the calculation and updating module is used for applying the current optimal strategy to equipment in the building and calculating a reward result according to an environment interaction result after the management strategy is executed; and updating the neural network weight and the management strategy in the strategy generator according to the environment interaction result and the reward result.

Description

Technical Field

[0001] The present invention relates to the technical field of building energy consumption monitoring, and in particular to a building energy consumption monitoring system and method based on Internet of Things data collection. Background Art

[0002] An important problem with traditional building energy management systems is the inability to dynamically adjust management strategies, which can lead to poor energy management and equipment operation efficiency. Traditional systems often rely on preset fixed rules and schedules to control equipment in buildings, such as air conditioning and lighting systems. These rules and schedules may have been carefully considered and optimized at the beginning of the design, but in actual operation, factors such as changes in indoor and outdoor environments, users' personalized needs, and the operating status of equipment cannot be reflected in a timely and accurate manner. For example, a fixed schedule may not be able to automatically adjust the set temperature of the air conditioner when the outdoor temperature changes, resulting in overheating in summer or overcooling in winter, which not only affects indoor comfort but also increases unnecessary energy consumption.

[0003] Similarly, the lack of light sensors makes it impossible for the lighting system to dynamically adjust according to changes in indoor and outdoor light, resulting in lighting equipment being turned on when there is sufficient light during the day, wasting electricity. The lack of equipment status sensors makes it impossible for traditional systems to detect and handle equipment failures in a timely manner, causing equipment to run for a long time in an inefficient or faulty state, further increasing energy consumption and maintenance costs. Therefore, the static management strategies of these traditional systems cannot adapt to dynamically changing environments and needs, resulting in deficiencies in energy consumption management and equipment operation efficiency. Summary of the invention

[0004] The purpose of the present invention is to provide a building energy consumption monitoring system and method based on Internet of Things data collection to solve the problems raised in the above background technology.

[0005] To achieve the above objectives, one of the objectives of the present invention is to provide a building energy consumption monitoring system based on Internet of Things data collection, including a data collection module, a strategy generator module and a calculation and update module, wherein:

[0006] The data collection module collects multi-source sensor data, merges the multi-source sensor data, and generates a state vector for the current time step;

[0007] The strategy generator module is used to generate a specific management strategy according to the state vector of the current time step; by setting a neural network as a strategy generator and initializing the weights of the neural network in the strategy generator, the current optimal strategy is generated using the exploration strategy method;

[0008] The calculation and update module is used to apply the current optimal strategy to the equipment in the building, and calculate the reward result according to the environmental interaction result after executing the management strategy; update the neural network weights and management strategy in the strategy generator according to the environmental interaction result and the reward result.

[0009] As a further improvement of the technical solution, the multi-source sensor data collected by the data collection module using the multi-source sensors specifically includes:

[0010] Temperature sensor: collects indoor and outdoor temperature data Tin and Tout;

[0011] Humidity sensor: collects indoor and outdoor humidity data Hin and Hout;

[0012] Light sensor: collects indoor and outdoor light data Lin and Lout;

[0013] Energy consumption monitoring sensor: collects energy consumption data E of energy-consuming equipment;

[0014] Device status sensor: collects the current operating status information Sdevice of energy-consuming devices.

[0015] As a further improvement of the technical solution, the data collection module merges the multi-source sensor data to generate the state vector St of the current time step, where:

[0016] St=[t,Tin,Tout,Hin,Hout,Lin,Lout,E,Sdevice].

[0017] As a further improvement of the technical solution, the process of generating the current optimal strategy by the strategy generator module specifically includes:

[0018] A neural network is selected as a strategy generator, including an input layer, a hidden layer, and an output layer. The input layer receives the state vector St of the current time step, the hidden layer extracts features through multiple fully connected layers or convolutional layers, and the output layer generates a specific management strategy At and a probability distribution. The action space of the management strategy At includes air conditioning temperature adjustment, lighting equipment switch, dehumidifier switch, and heating equipment switch.

[0019] Initialize the weights of the neural network using a random initialization method;

[0020] The exploration strategy adopts the ε method, which randomly selects the action space of the management strategy At with a certain probability ε, and selects the action space with the largest probability distribution in the management strategy At with a certain probability 1-ε;

[0021] As training progresses, the frequency of randomly selected actions is reduced by gradually reducing the value of ε. ε is multiplied by a decay factor every 100 training rounds. When the value of ε is reduced to 0.1, the policy generator selects the action space in the current management policy At as the current best policy.

[0022] As a further improvement of the technical solution, the calculation and update module includes a reward result calculation unit, and the reward result calculation unit generates a calculation reward result and the process of the reward result specifically includes:

[0023] The generated management strategy is applied to the equipment in the building through device control. After the management strategy is executed, a new state vector St+1 is obtained from the data collection module as the generated environmental interaction result. According to the multi-source sensor data in St+1, the reward result Rt+1 is calculated. The calculation formula is as follows:

[0024] Rt+1=-E(t+1)+α·C(t+1)+β·Sdevice(t+1), where E(t+1) represents the energy consumption data after the execution of the management strategy; C(t+1) represents the indoor comfort based on the combination of indoor and outdoor temperature data, indoor and outdoor humidity data, and indoor and outdoor light data in St+1; Sdevice(t+1) represents the device operation status information after the execution of the management strategy, indicating the device efficiency; α is used to adjust the contribution of indoor comfort to the reward result, and β is used to adjust the contribution of device efficiency to the reward result.

[0025] As a further improvement of the technical solution, the calculation and update module includes an update unit, and the update process of the update unit specifically includes:

[0026] Calculate the advantage function. The advantage function is the difference between the reward result of the state vector St at the current time step and the reward result of the new state vector St+1. The advantage function reflects the actual effect of the current action.

[0027] The update direction and amplitude of the neural network weights are determined according to the advantage function and the action probability ratio, where the action probability ratio represents the ratio of the probability that the current policy generator selects the current best policy to the probability that the policy generator after the last update selects the current best policy. If the probability that the current policy generator selects the current best policy is higher than the last time, the selection probability of the action is increased; otherwise, the selection probability is reduced. The weights of the neural network are gradually adjusted through the gradient ascent method.

[0028] The second object of the present invention is to provide a method for a building energy consumption monitoring system based on Internet of Things data collection, comprising the following method steps:

[0029] S1, collect multi-source sensor data, merge the multi-source sensor data, and generate the state vector of the current time step;

[0030] S2. Generate a specific management strategy based on the state vector of the current time step; generate the current optimal strategy by setting the neural network as the strategy generator and initializing the neural network weights in the strategy generator, and using the exploration strategy method;

[0031] S3. Apply the current best strategy to the equipment in the building, and calculate the reward result based on the environmental interaction result after executing the management strategy; update the neural network weights and management strategy in the strategy generator based on the environmental interaction result and the reward result.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] The building energy consumption monitoring system based on IoT data collection generates the state vector of the current time step by collecting multi-source sensor data in real time; these state vectors contain information such as indoor and outdoor temperature, humidity, light, energy consumption and equipment status; and use the strategy generator to dynamically generate management strategies through neural networks, which can respond quickly to environmental changes, adjust air conditioning temperature, switch lighting equipment, etc., to ensure the best balance between indoor comfort and energy efficiency.

[0034] In addition, rewards are calculated based on the execution results of the generated management strategy, and the weights of the neural network are updated through the advantage function and the action probability ratio. This feedback-based update mechanism enables the strategy generator to continuously learn and optimize, gradually reducing energy consumption and improving equipment operating efficiency. As training progresses, the system can identify which management strategies are most effective in specific environments, thereby achieving smarter and more dynamic energy management, ensuring that the building can operate efficiently under various conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a schematic diagram of the overall method flow of the present invention;

[0036] Figure 2 It is a schematic diagram of the overall module unit of the present invention.

[0037] In the figure: 100, data collection module; 200, strategy generator module; 300, calculation and update module; 301, reward result calculation unit; 302, update unit. DETAILED DESCRIPTION

[0038] The following will be combined with the accompanying drawings in the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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.

[0039] Next, see Figure 1-Figure 2 One of the purposes of this embodiment is to provide a building energy consumption monitoring system based on Internet of Things data collection, including a data collection module 100, a strategy generator module 200 and a calculation and update module 300.

[0040] The data collection module 100 uses a multi-source sensor to collect multi-source sensor data, merges the multi-source sensor data, and generates a state vector of the current time step, wherein the multi-source sensor data collected by the multi-source sensor specifically includes:

[0041] Temperature sensor: collects indoor and outdoor temperature data Tin and Tout. The temperature sensor collects indoor and outdoor temperature data, which helps the strategy generator determine when to adjust the air conditioning temperature to maintain indoor comfort and reduce unnecessary energy consumption.

[0042] Humidity sensor: collects indoor and outdoor humidity data Hin and Hout; the humidity sensor collects indoor and outdoor humidity data, which enables the strategy generator to decide whether to turn on the dehumidifier to avoid excessive humidity affecting comfort and optimize energy consumption;

[0043] Light sensor: collects indoor and outdoor light data Lin and Lout. The light sensor collects indoor and outdoor light data, which helps the strategy generator decide when to turn on or off lighting equipment to ensure that the indoor light is appropriate and save electricity to the greatest extent.

[0044] Energy consumption monitoring sensor: collects energy consumption data E of energy-consuming equipment (such as air conditioning, lighting, heating systems, etc.); Energy consumption monitoring sensors collect energy consumption data of energy-consuming equipment, which enables the strategy generator to evaluate the power consumption of the equipment, thereby optimizing the operation of the equipment to reduce energy consumption and control costs;

[0045] Device status sensor: collects the current operating status information Sdevice of energy-consuming devices; the device status sensor collects the current operating status information of energy-consuming devices. This data helps the strategy generator understand the working status of the device, ensure that the device operates in the optimal state, and detect and handle faults in a timely manner.

[0046] The data collection module 100 merges the multi-source sensor data to generate a state vector St of the current time step, where St = [t, Tin, Tout, Hin, Hout, Lin, Lout, E, Sdevice]; the state vector provides the policy generator with comprehensive environmental information, helping it to more accurately evaluate the current situation and generate the optimal management strategy; by integrating data such as temperature, humidity, light, energy consumption and equipment status, the policy generator can make comprehensive judgments in multiple dimensions to ensure the optimal balance between comfort and energy efficiency of the building environment.

[0047] The strategy generator module 200 is used to generate a specific management strategy according to the state vector of the current time step; by setting a neural network as a strategy generator and initializing the weights of the neural network in the strategy generator, the current optimal strategy is generated using the exploration strategy method, which specifically includes:

[0048] A neural network (such as a multi-layer perceptron, a convolutional neural network or a recurrent neural network) is selected as a strategy generator, including an input layer, a hidden layer and an output layer, wherein the input layer receives the state vector St of the current time step, the hidden layer performs feature extraction through multiple fully connected layers or convolutional layers, and the output layer generates a specific management strategy At and a probability distribution. The action space of the management strategy At includes: air conditioning temperature adjustment (such as setting a temperature value), lighting equipment switch (such as turning on or off), dehumidifier switch (such as turning on or off) and heating equipment switch (such as turning on or off).

[0049] Initialize the weights of the neural network using a random initialization method;

[0050] In the early stage of training, the strategy generator needs to explore to understand the effects of different management strategies. The exploration strategy uses the ε method, which randomly selects the action space of the management strategy At with a certain probability ε, and selects the action space with the largest probability distribution in the management strategy At with a certain probability 1-ε;

[0051] As training progresses, the policy generator gradually accumulates a large amount of data and experience, and is able to identify which management strategies are most effective under specific conditions. At this point, the policy generator begins to rely more on the known optimal strategy; by gradually reducing the value of ε to reduce the frequency of randomly selected actions, ε is multiplied by a decay factor every 100 training rounds. When the ε value is reduced to 0.1, the policy generator selects the action space in the current management strategy At as the current optimal strategy.

[0052] The reward result calculation unit 301 in the calculation and update module 300 is used to apply the generated current optimal strategy to the equipment in the building, generate the environment interaction result, and calculate the reward result according to the environment interaction result after executing the management strategy; specifically including:

[0053] The generated management strategy is applied to the equipment in the building through device control, such as adjusting the air conditioning temperature and switching the lighting equipment. After the management strategy is executed, a new state vector St+1 is obtained from the data collection module 100 as the generated environment interaction result. According to the multi-source sensor data in St+1, the reward result Rt+1 is calculated. The calculation formula is as follows:

[0054] Rt+1=-E(t+1)+α·C(t+1)+β·Sdevice(t+1), where E(t+1) represents the energy consumption data after the execution of the management strategy; C(t+1) represents the indoor comfort based on the combination of indoor and outdoor temperature data, indoor and outdoor humidity data, and indoor and outdoor light data in St+1; Sdevice(t+1) represents the device operation status information after the execution of the management strategy, indicating the device efficiency; α is used to adjust the contribution of indoor comfort to the reward result, and β is used to adjust the contribution of device efficiency to the reward result.

[0055] The updating unit 302 in the calculation and updating module 300 updates the neural network weights and management strategies in the strategy generator according to the environmental interaction results and the reward results, specifically including:

[0056] Calculate the advantage function. The advantage function is the difference between the reward result of the state vector St at the current time step and the reward result of the new state vector St+1. The advantage function reflects the actual effect of the current action.

[0057] The update direction and amplitude of the neural network weights are determined according to the advantage function and the action probability ratio, where the action probability ratio represents the ratio of the probability that the current policy generator selects the current best policy to the probability that the policy generator after the last update selects the current best policy; if the probability that the current policy generator selects the current best policy is higher than the last time, the selection probability of the action is increased; otherwise, the selection probability is reduced; and the weights of the neural network are gradually adjusted through the gradient ascent method to ensure that the policy generator can generate more effective management policies;

[0058] After the neural network weights of the strategy generator are updated, the management strategy is also updated. The new management strategy will pay more attention to reducing energy consumption, improving indoor comfort and optimizing equipment operation efficiency in future decision-making, thereby gradually improving overall performance.

[0059] The second purpose of this embodiment is to provide a method for a building energy consumption monitoring system based on Internet of Things data collection, including the following method steps:

[0060] S1, collect multi-source sensor data, merge the multi-source sensor data, and generate the state vector of the current time step;

[0061] S2. Generate a specific management strategy based on the state vector of the current time step; generate the current optimal strategy by setting the neural network as the strategy generator and initializing the neural network weights in the strategy generator, and using the exploration strategy method;

[0062] S3. Apply the current best strategy to the equipment in the building, and calculate the reward result based on the environmental interaction result after executing the management strategy; update the neural network weights and management strategy in the strategy generator based on the environmental interaction result and the reward result.

[0063] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. Building energy consumption monitoring system based on IoT data collection, characterized by: The system comprises a data collection module (100), a strategy generator module (200) and a calculation and update module (300), wherein: The data collection module (100) collects multi-source sensor data, merges the multi-source sensor data, and generates a state vector of the current time step; The strategy generator module (200) is used to generate a specific management strategy according to the state vector of the current time step; by setting a neural network as a strategy generator and initializing the weight of the neural network in the strategy generator, using an exploration strategy method to generate the current optimal strategy; The calculation and update module (300) is used to apply the current optimal strategy to the equipment in the building, and calculate the reward result according to the environmental interaction result after executing the management strategy; and update the neural network weight and management strategy in the strategy generator according to the environmental interaction result and the reward result.

2. The building energy consumption monitoring system based on Internet of Things data collection according to claim 1 is characterized in that: The multi-source sensor data collected by the data collection module (100) using the multi-source sensors specifically include: Temperature sensor: collects indoor and outdoor temperature data Tin and Tout; Humidity sensor: collects indoor and outdoor humidity data Hin and Hout; Light sensor: collects indoor and outdoor light data Lin and Lout; Energy consumption monitoring sensor: collects energy consumption data E of energy-consuming equipment; Device status sensor: collects the current operating status information Sdevice of energy-consuming devices.

3. The building energy consumption monitoring system based on Internet of Things data collection according to claim 2 is characterized in that: The data collection module (100) combines the multi-source sensor data to generate a state vector St of the current time step, where: St=[t,Tin,Tout,Hin,Hout,Lin,Lout,E,Sdevice].

4. The building energy consumption monitoring system based on Internet of Things data collection according to claim 3 is characterized in that: The process of generating the current optimal strategy by the strategy generator module (200) specifically includes: A neural network is selected as a strategy generator, including an input layer, a hidden layer, and an output layer. The input layer receives the state vector St of the current time step, the hidden layer extracts features through multiple fully connected layers or convolutional layers, and the output layer generates a specific management strategy At and a probability distribution. The action space of the management strategy At includes air conditioning temperature adjustment, lighting equipment switch, dehumidifier switch, and heating equipment switch. Initialize the weights of the neural network using a random initialization method; The exploration strategy adopts the ε method, which randomly selects the action space of the management strategy At with a certain probability ε, and selects the action space with the largest probability distribution in the management strategy At with a certain probability 1-ε; As training progresses, the frequency of randomly selected actions is reduced by gradually reducing the value of ε. ε is multiplied by a decay factor every 100 training rounds. When the value of ε is reduced to 0.1, the policy generator selects the action space in the current management policy At as the current best policy.

5. The building energy consumption monitoring system based on Internet of Things data collection according to claim 4 is characterized in that: The calculation and update module (300) includes a reward result calculation unit (301), and the reward result calculation unit (301) generates a calculation reward result and the reward result process specifically includes: The generated management strategy is applied to the equipment in the building through device control. After the management strategy is executed, a new state vector St+1 is obtained from the data collection module (100) as the generated environment interaction result. Based on the multi-source sensor data in St+1, the reward result Rt+1 is calculated. The calculation formula is as follows: Rt+1=-E(t+1)+α·C(t+1)+β·Sdevice(t+1), where E(t+1) represents the energy consumption data after the execution of the management strategy; C(t+1) represents the indoor comfort based on the combination of indoor and outdoor temperature data, indoor and outdoor humidity data, and indoor and outdoor light data in St+1; Sdevice(t+1) represents the device operation status information after the execution of the management strategy, indicating the device efficiency; α is used to adjust the contribution of indoor comfort to the reward result, and β is used to adjust the contribution of device efficiency to the reward result.

6. The building energy consumption monitoring system based on Internet of Things data collection according to claim 5 is characterized in that: The calculation and update module (300) includes an update unit (302), and the update process of the update unit (302) specifically includes: Calculate the advantage function. The advantage function is the difference between the reward result of the state vector St at the current time step and the reward result of the new state vector St+1. The advantage function reflects the actual effect of the current action. The update direction and amplitude of the neural network weights are determined according to the advantage function and the action probability ratio, where the action probability ratio represents the ratio of the probability that the current policy generator selects the current best policy to the probability that the policy generator after the last update selects the current best policy. If the probability that the current policy generator selects the current best policy is higher than the last time, the selection probability of the action is increased; otherwise, the selection probability is reduced. The weights of the neural network are gradually adjusted through the gradient ascent method.

7. A method for using the building energy consumption monitoring system based on Internet of Things data collection according to any one of claims 1 to 6, characterized in that: The method comprises the following steps: S1, collect multi-source sensor data, merge the multi-source sensor data, and generate the state vector of the current time step; S2. Generate a specific management strategy based on the state vector of the current time step; generate the current optimal strategy by setting the neural network as the strategy generator and initializing the neural network weights in the strategy generator, and using the exploration strategy method; S3, applying the current best strategy to the equipment in the building, and calculating the reward results based on the environmental interaction results after executing the management strategy; Update the neural network weights in the policy generator and manage the policy based on the environment interaction results and reward results.

Images