Flexible energy consumption excitation method for temperature prediction based on feed-forward model
Through the temperature prediction and intelligent sensor system based on the ‘feedforward’ model, users can independently adjust the air conditioner turn-on time, solving the problems of high electricity prices and unbalanced supply and demand during peak grids, and achieving comfortable rest and energy-saving effects.
Patent Information
- Application Number
- CN202411003288.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2024-07-04
- Filing Date
- 2024-07-25
- Publication Date
- 2024-11-12
Smart Images

Figure CN118940627A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of temperature prediction, and in particular to a flexible energy use incentive method for temperature prediction based on a "feedforward" model. Background Art
[0002] Flexible incentives are a human-centered and humane management method. Based on the study of human psychology and behavioral laws, it uses a non-compulsory approach to create a potential persuasiveness in the minds of users, making it easier for users to accept. It is a transformation of the traditional incentive mechanism. The "feedforward" model predicts data for a period of time in the future based on the input of existing data, and can quickly and accurately calculate the required output.
[0003] In life, people often need to go out to work, and hope to have a comfortable environment to rest when they return home. However, people often arrive home at the same time. Turning on the air conditioner at this time is during the peak period of the power grid. Not only is the electricity price expensive, but it is also easy to cause an imbalance in the supply and demand of the power grid, which puts pressure on the power grid regulation. During the startup phase, not only the load of the room but also the heat storage load of the enclosing structure must be handled. The energy consumption of the air conditioner is highest in the first hour of startup. Therefore, a flexible energy incentive method based on a "feedforward" model for temperature prediction is proposed. Summary of the invention
[0004] 1. Technical issues to be solved
[0005] In view of the deficiencies in the prior art, the present invention provides a flexible energy incentive method for temperature prediction based on a "feedforward" model, which has the advantages of enabling users to save electricity consumption while enjoying a comfortable resting environment. It solves the problem that in life, people often need to go out to work and hope to have a comfortable environment to rest when they return home, but people often arrive home at the same time. At this time, turning on the air conditioner is at the peak of the power grid, which not only makes the electricity price expensive but also easily leads to an imbalance in the supply and demand of the power grid, bringing pressure to the power grid regulation. Since the startup stage not only has to deal with the load of the room, but also the heat storage load of the enclosing structure, the air conditioning energy consumption is highest in the first hour of startup.
[0006] (II) Technical solution
[0007] To achieve the above object, the present invention provides the following technical solution: a flexible energy incentive method for temperature prediction based on a "feedforward" model, comprising the following steps:
[0008] Step 1: Obtain input parameters and output parameters of training samples;
[0009] Step 2: Use the long short-term memory neural network algorithm to obtain a mathematical model that represents the mapping relationship between the input parameters and output parameters of the training sample, and obtain a neural network control system, i.e., "feedforward"
[0010] Model calculation system;
[0011] Step 3: Install the smart sensor in the room and connect it to the user's mobile phone APP through signal connection;
[0012] Step 4: Substitute the input parameters into the system for prediction calculation to obtain the prediction results, including temperature change, thermal sensation change, and power consumption change, which are displayed on carriers such as web pages or mobile APPs through signal connections;
[0013] Step 5: After checking the temperature change curve and combining it with the real-time electricity price, the user can then independently choose whether to turn on the room air conditioner, adjust the set temperature, or pre-cool the room by turning on the air conditioner in advance;
[0014] Step 6: Determine the temperature for turning on the room air conditioner according to the user's selection, and input the set temperature into the "feedforward" model;
[0015] Step 7: Based on the room temperature curve calculated by the "feedforward" model and the real-time electricity fee data and whether the air conditioner will be at a peak electricity consumption period after it is turned on, choose whether to turn on the air conditioner in advance or delay turning it on;
[0016] Step 8: The system acquires new parameters through continuous monitoring for reinforcement learning and updates the APP version.
[0017] Preferably, the input parameters and output parameters of the training samples are obtained, including: obtaining the input parameters and output parameters through experiments or actual room monitoring, the input parameters including room information, air conditioning information, operating parameters, and environmental parameters; the output parameters including changes in indoor temperature, changes in subjective thermal sensation, and changes in power consumption.
[0018] Preferably, a mathematical model representing the mapping relationship between input parameters and output parameters of a training sample is obtained by using a long short-term memory neural network algorithm, and a neural network control system is obtained, including: collecting training samples containing control system input and expected output through historical data, simulation environment or actual operation; bringing the input data in the training sample into the neural network model with parameter initialization, obtaining the output through forward propagation calculation, comparing the output of the model with the expected output in the training sample, and calculating the error; based on the error, using a back propagation algorithm to update the parameters of the neural network to reduce the error. Repeat this process until the error reaches a preset threshold or reaches a predetermined number of training rounds; input the data to be controlled into the trained neural network model, and the model will output a corresponding control signal to drive the actuator of the control system, thereby achieving control of the system.
[0019] Preferably, the input parameters are substituted into the system for predictive calculation to obtain the predicted results, and are represented on a carrier such as a web page or a mobile phone APP through a signal connection: including: installing sensors to monitor the operation and environmental parameters of the indoor room air-conditioning equipment to obtain the benchmark operating conditions; entering the desired air-conditioning control conditions in the "input section" of the web page or mobile phone APP, such as power on / off, changing the set temperature, etc., to obtain the control information; bringing the benchmark operating conditions, control information, and other input information such as electricity prices into the trained algorithm for predictive calculation to obtain the predicted results; and displaying the predicted results: room temperature changes, comfortable temperature range, real-time electricity prices, electricity savings, electricity saving costs, and PMV on the "prediction results" section of the web page or mobile phone APP through a signal connection.
[0020] Preferably, after viewing the temperature change curve and combining it with the real-time electricity price, the user can independently choose whether to turn on the room air conditioner, adjust the set temperature, or pre-cool the room by turning on the air conditioner in advance. When the user returns from get off work or work outside, he can view the temperature curve change graph calculated by the "feedforward" model, and determine whether the room temperature reaches the user's expected temperature when returning to the room, and then choose whether to turn on the air conditioner to pre-cool the room.
[0021] Preferably, the temperature for turning on the room air conditioner is determined according to the user's selection, and the set temperature is input into the "feedforward" model. After obtaining the calculation result of the "feedforward" model, the user can set the temperature of the air conditioner in advance and input the set temperature into the "feedforward" model calculation system.
[0022] Preferably, the room temperature curve calculated according to the "feedforward" model and the real-time electricity bill data, as well as whether the air conditioner will be in a peak electricity consumption period after it is turned on, can be used to select whether to turn on the air conditioner in advance or delay turning on the air conditioner. The user can choose whether to turn on the air conditioner in advance based on the temperature change curve calculated by the "feedforward" model calculation system, the real-time electricity bill data generated, and whether the air conditioner is in a peak electricity consumption period. The set air-conditioning temperature is continuously changed to obtain different temperature change curves and real-time electricity bills. The user can choose the air-conditioning temperature that he thinks is most suitable and then turn on the air conditioner.
[0023] Preferably, after being put into use, the system can continuously monitor and obtain new data, and with the user's consent, the new input can be added to the training set for intensive training and the APP version can be updated.
[0024] Preferably, the smart sensor is installed in the room, and the indoor temperature, indoor humidity, wall temperature, outdoor temperature, and radiation of the room are monitored by installing smart sensors inside and outside the room. The smart sensor includes a temperature sensor, a humidity sensor, and a radiation intensity sensor. The smart sensor is connected to the user's mobile phone APP through a signal connection, and the temperature sensors, humidity sensors, and radiation intensity sensors installed inside and outside the room are connected to the user's mobile phone APP through a signal connection.
[0025] Preferably, it comprises an air conditioner and a server having the functions as described in any one of claims 1-8.
[0026] (III) Beneficial effects
[0027] Compared with the prior art, the present invention provides a flexible energy incentive method for temperature prediction based on a "feedforward" model, which has the following beneficial effects:
[0028] 1. The flexible energy incentive method for temperature prediction based on the "feedforward" model obtains input parameters and output parameters through experiments or actual room monitoring, wherein the input parameters include room information, air conditioning information, operating parameters, and environmental parameters; the output parameters include indoor temperature changes, subjective thermal sensation changes, and power consumption changes; the K-Means clustering algorithm is used to classify training samples according to parameters such as room information and air conditioning information, and then a long short-term memory neural network (LSTM) is used to model samples of different categories, and indoor temperature changes can be predicted based on input parameters such as operating parameters and environmental parameters; the model will output corresponding control signals, which can be used to drive the actuators of the control system, thereby obtaining a neural network control system.
[0029] 2. This flexible energy incentive method for temperature prediction based on the "feedforward" model can detect the temperature inside and outside the room by setting up intelligent sensors, so that users can check the room temperature by using their mobile phones. The temperature monitored by the sensor is transmitted to the user's mobile phone APP through signal connection. The "feedforward" model calculation system inside the mobile phone APP can predict the temperature of the room for a period of time in the future. It can also calculate the temperature change curve of the room after turning on the air conditioner and the real-time electricity bill generated by using the air conditioner, so that users can have a more intuitive understanding of the future temperature of the room and the electricity usage.
[0030] 3. The flexible energy incentive method based on the "feedforward" model for temperature prediction allows users to set the air-conditioning temperature in advance before turning on the air-conditioning, and then use the "feedforward" model calculation system to calculate the temperature change curve of the room in the future, so that users can choose the best time and temperature to turn on the air-conditioning, which greatly saves the power consumption of the air-conditioning and allows users to immediately enjoy a comfortable rest environment when they return to the room. By setting the air-conditioning to be delayed or turned off at a scheduled time, the temperature of the room can reach the user's expected temperature when the user returns to the room, or the air-conditioning can be automatically turned off when the user turns on the air-conditioning for a period of time to the preset temperature, which can greatly save the user's power consumption. By allowing users to make various autonomous changes, users can increase their sense of identity when using the method, so that they can quickly accept the method. By predicting in advance whether it is in the peak power consumption period, users can avoid the peak power consumption period in time when using the air-conditioning, thereby reducing the pressure on the power grid.
[0031] 4. After the flexible energy incentive method for temperature prediction based on the "feedforward" model is put into use, the system can continuously monitor and obtain new data. With the user's consent, the new input can be added to the training set for intensive training and the APP version can be updated. BRIEF DESCRIPTION OF THE DRAWINGS
[0032] Figure 1 It is a schematic flow chart of the control method of the present invention.
[0033] Figure 2 Flowchart of the flexible energy incentive method based on the "feedforward" model
[0034] Figure 3 Schematic diagram of the overall implementation model of this air conditioning system
[0035] Figure 4 K-Means clustering algorithm flow chart
[0036] Figure 5 Schematic diagram of long short-term memory neural network (LSTM)
[0037] Figure 6 Schematic diagram of the process of generating the LSTM mathematical model
[0038] Figure 7 Schematic diagram of the mobile APP display page DETAILED DESCRIPTION
[0039] The following will be combined with the embodiments of 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.
[0040] A flexible energy incentive method for temperature prediction based on a "feedforward" model includes the following steps:
[0041] Step 1: Obtain input parameters and output parameters of training samples;
[0042] Step 2: Use the long short-term memory neural network algorithm to obtain a mathematical model that represents the mapping relationship between the input parameters and output parameters of the training sample, and obtain a neural network control system, i.e., "feedforward"
[0043] Model calculation system;
[0044] Step 3: Install the smart sensor in the room and connect it to the user's mobile phone APP through signal connection;
[0045] Step 4: Substitute the input parameters into the system for prediction calculation to obtain the prediction results, including temperature change, thermal sensation change, and power consumption change, which are displayed on carriers such as web pages or mobile APPs through signal connections;
[0046] Step 5: After checking the temperature change curve and combining it with the real-time electricity price, the user can then independently choose whether to turn on the room air conditioner, adjust the set temperature, or pre-cool the room by turning on the air conditioner in advance;
[0047] Step 6: Determine the temperature for turning on the room air conditioner according to the user's selection, and input the set temperature into the "feedforward" model;
[0048] Step 7: Based on the room temperature curve calculated by the "feedforward" model and the real-time electricity fee data and whether the air conditioner will be at a peak electricity consumption period after it is turned on, choose whether to turn on the air conditioner in advance or delay turning it on;
[0049] Step 8: The system acquires new parameters through continuous monitoring for reinforcement learning and updates the APP version.
[0050] Embodiment 1:
[0051] Select a representative experimental room with air conditioning equipment in good condition to facilitate sensor installation and data collection. Indoor temperature and humidity sensors should be installed in representative locations in the room, such as the center or a certain distance from the air outlet of the air conditioner. Outdoor temperature and humidity sensors should avoid direct sunlight and shielding. Install a pyranometer to measure solar radiation illuminance. Record room information, such as the city where the room is located, the length, width and height of the room, window area, window direction, room function, etc.; record air conditioning information, such as air conditioning brand model, years of use, cooling capacity, heating capacity, power consumption, etc. By changing the air conditioning setting temperature, mode and wind speed to obtain different initial working conditions, let the subjects fill in the questionnaire every 10 minutes in the experimental room to obtain the subjects' subjective thermal perception.
[0052] Collect sample data sets containing room information and air conditioning information, extract features that have an important impact on classification from the original data, clean the data, remove missing values, outliers or insignificant features, determine the number of clusters to be generated, that is, the K value, use the K-Means algorithm to cluster the preprocessed data, and classify similar samples into the same category; use the long short-term memory neural network LSTM to model samples of different categories, and collect data sets containing operating parameters such as air conditioning on time, set temperature, mode, wind speed, power consumption, operating time, etc. and environmental parameters such as indoor and outdoor temperature, humidity, solar radiation, etc. At the same time, ensure that the data set also contains the target output parameters, that is, the changes in indoor temperature, subjective thermal perception and power consumption, clean and label the data, and divide the data set into training set, validation set and test set. The training set is used to train the LSTM model, the validation set is used to adjust the model parameters and prevent overfitting, and the test set is used to evaluate the performance of the model. Perform necessary preprocessing on the sequence data, such as filling missing values, standardization or normalization, to convert the sequence data into a format suitable for LSTM input, for example, using sliding window technology to create a fixed-length sequence. Build an LSTM model according to specific needs. Determine the number of layers of the network, the size of the hidden layer, the activation function, the loss function, and the optimizer. Use the sequence data of each clustering category to train the LSTM model separately. This usually includes steps such as forward propagation, calculating losses, backpropagation, and weight updating. Use the validation set or test set to evaluate the trained LSTM model to check its performance in predicting output parameters. Summarize or integrate the LSTM output results of samples of different categories for subsequent analysis or decision-making; when the neural network model is trained, it can be applied to the actual control system. In actual applications, the data to be controlled is input into the trained neural network model, and the model will output the corresponding control signals, which can be used to drive the actuators of the control system to achieve control of the system.
[0053] Install intelligent sensors in the room. Monitor the indoor temperature, indoor humidity, wall temperature, outdoor temperature, and radiation of the room by installing intelligent sensors inside and outside the room. The intelligent sensors include temperature sensors, humidity sensors, and radiation intensity sensors. Connect the intelligent sensors to the user's mobile phone APP through signal connection. Connect the temperature sensors, humidity sensors, and radiation intensity sensors set inside and outside the room to the user's mobile phone APP through signal connection, so that the user can view the temperature, humidity, and radiation intensity inside and outside the room in real time. There is a "feedforward" model calculation system inside the mobile phone APP. The "feedforward" model calculation system inputs the indoor temperature, indoor humidity, wall temperature, outdoor temperature, radiation, etc., and then calculates the load of the room according to these parameters. Then, combined with whether the air conditioner in the room is turned on and the cooling capacity of the air conditioner, calculate the heat change of the room, and then obtain the indoor temperature change of the room, and then obtain the indoor temperature change curve of the room in the next few hours, so as to achieve the "feedforward" effect. The mobile phone APP can display the room temperature change curve, comfortable temperature range, real-time electricity price, power saving amount, power saving cost, and PMV on the mobile phone display screen. The "feedforward" model calculation system obtains the initial temperature of the room, and then calculates the temperature in the next period of time according to the heat transfer of the room, and then calculates in turn to obtain the temperature change curve in the next period of time. The calculation frequency can be set according to the user, and it can be 10 minutes or 15 minutes. Then, the indoor temperature change curve of the room in the next few hours can be obtained, so as to achieve the "feedforward" effect. PMV is an evaluation index of human thermal comfort, which is calculated through temperature, humidity, wind speed, mean radiant temperature, clothing thermal resistance, and activity level. When -0.5 < PMV < 0.5, the surrounding environment is considered neither cold nor hot and is comfortable. In this way, when making decisions, users can not only view the future room temperature curve, electricity bills, etc., but also see whether the environmental changes are within the comfortable range, helping users make decisions, thus achieving flexible incentives.
[0054] The data acquired by the smart sensor is transmitted into the "feedforward" model. The smart sensor can send the temperature, humidity, radiation intensity and other information inside and outside the room to the "feedforward" model calculation system on the corresponding APP of the user's mobile phone through signal transmission at regular intervals. The room temperature curve and real-time electricity fee for a period of time in the future are calculated by the "feedforward" model. The "feedforward" model can calculate the temperature change curve of the room for a period of time in the future through the input smart sensor data. According to whether the air conditioner is turned on, the temperature change curve for a period of time in the future and the real-time electricity fee information can be calculated. The "feedforward" model calculates the change curve of the room temperature after the air conditioner is turned on. First, the initial temperature of the room is measured, such as 30°C, and then the air conditioner is turned on and the air conditioner temperature is set. For example, when the air conditioner temperature is set to 20°C, the "feedforward" model calculation system can calculate how long it takes to turn on the 20°C air conditioner to reduce the room temperature of 30°C to a temperature suitable for the user. The "feedforward" model calculation system can obtain the temperature change curve, and then the user sets the delayed opening time and closing time of the air conditioner according to the temperature change curve, and can also calculate the cost generated by the air conditioner during the opening time period.
[0055] After viewing the temperature change curve and combining it with the real-time electricity price, the user can then choose whether to turn on the room air conditioner, adjust the set temperature, or pre-cool the room by turning on the air conditioner in advance. The user can check the mobile phone APP to understand the temperature inside the room when returning from get off work or working outside, and then by checking the temperature curve change graph calculated by the "feedforward" model, determine whether the room temperature reaches the user's expected temperature when returning to the room, and then choose whether to turn on the air conditioner to pre-cool the room. If the air conditioner is turned on. The "feedforward" model can calculate whether it is the rush hour after work, and then remind the user whether to turn on the air conditioner in advance to avoid the peak electricity consumption period and reduce the pressure on the power grid. When the air conditioner is running, when the user turns on the air conditioner at a temperature of 22°C, and this is during the peak electricity consumption period of the power grid, that is, in the morning and evening, the power grid is insufficient and the electricity price may be higher at this time. At this time, a message will pop up through the mobile phone APP to remind the user, providing the user with the option of changing the air conditioner set temperature, and displaying the temperature change curve of the room after the set temperature changes from 22°C to 26°C, as well as the electricity bill after adjustment and the electricity bill if no adjustment is made. Then the user can judge whether to accept the adjustment based on the subsequent room temperature, and the saved electricity bill, PMV and other auxiliary information can motivate the user to make adjustments. If the user thinks the room temperature is appropriate, he can also turn off the air conditioner.
[0056] The temperature at which the room air conditioner is turned on is determined based on the user's selection, and the set temperature is input into the "feedforward" model. After obtaining the calculation results of the "feedforward" model, the user can set the temperature of the air conditioner according to the mobile phone APP to avoid peak electricity consumption, and then input the set temperature into the "feedforward" model calculation system, which can obtain the temperature change curve of the room in the future after the air conditioner is turned on, as well as the electricity bill information generated continuously after the air conditioner is turned on.
[0057] Based on the room temperature curve calculated by the "feedforward" model, the real-time electricity bill data, and whether the air conditioner will be at a peak electricity consumption period, the user can choose whether to turn on the air conditioner in advance or delay turning it on.
[0058] The "feedforward" model calculation system calculates the temperature change curve and the real-time electricity bill data, and whether it is in the peak electricity consumption period can choose whether to turn on the air conditioner in advance. The set air-conditioning temperature can also be continuously changed to obtain different temperature change curves and real-time electricity bills. Users can choose the most suitable air-conditioning temperature and then turn on the air conditioner. If the "feedforward" model calculates that it will be in the peak electricity consumption period in the future or it is in the peak electricity consumption period now, the air conditioner can be controlled to start in advance or delay, and the air conditioner can automatically turn on after a set period of time, or automatically turn off after the air conditioner is turned on for a period of time.
[0059] The system continuously monitors the room and collects various data including user behavior, performance data, feedback, etc. After obtaining the user's consent, the system will add the newly collected data to the reinforcement learning training set. The system uses the reinforcement learning algorithm to train the training set and continuously optimize and improve the model. Based on the results of reinforcement learning, the system will update the APP version. This may include fixing known problems, improving user experience, adding new features or optimizing algorithms. Through this process, the system can continuously adapt to and meet changes in user needs, and continuously improve the quality and competitiveness of the APP.
[0060] The system continuously monitors and acquires new parameters for reinforcement learning, and updates the APP version; wherein the input parameters include room information, air conditioning information, operating parameters, and environmental parameters; the output parameters include indoor temperature changes, subjective thermal sensation changes, and power consumption changes; the prediction results include temperature changes, thermal sensation changes, and power consumption changes;
[0061] The mathematical model representing the mapping relationship between the input parameters and the output parameters of the training samples is obtained, including:
[0062] (1) Use the K-Means clustering algorithm to classify training samples based on parameters such as room information and air conditioning information, including:
[0063] Collect a sample data set containing room information and air conditioner information. This information may include the room area, orientation, floor, usage frequency, air conditioner model, power, energy efficiency level, etc.
[0064] Extract features that have an important impact on classification from the original data. These features should be able to reflect the characteristics and status of the room and air conditioner.
[0065] Clean the data to remove missing values, outliers, or insignificant features. Feature scaling or normalization may be required to ensure that the scales of different features are consistent.
[0066] The number of clusters to be generated is determined as the K value. The K-Means algorithm is used to cluster the preprocessed data and classify similar samples into the same category.
[0067] (2) Using the long short-term memory neural network (LSTM) to model samples of different categories, the indoor temperature change can be predicted based on input parameters such as operating parameters and environmental parameters, including:
[0068] Collect data sets containing operating parameters, such as air conditioner on time, set temperature, mode, wind speed, power consumption, operating time and other environmental parameters, such as indoor and outdoor temperature, humidity, solar radiation, etc. At the same time, ensure that the data set also contains the target output parameters, that is, the changes in indoor temperature, subjective thermal sensation and power consumption, clean and label the data, and divide the data set into training set, validation set and test set. The training set is used to train the LSTM model, the validation set is used to adjust the model parameters and prevent overfitting, and the test set is used to evaluate the performance of the model;
[0069] Perform necessary preprocessing on the sequence data, such as filling missing values, standardization or normalization, to convert the sequence data into a format suitable for LSTM input. For example, use the sliding window technique to create a fixed-length sequence.
[0070] Build an LSTM model based on specific needs. Determine the number of network layers, the size of the hidden layer, the activation function, the loss function, and the optimizer;
[0071] Use the sequence data of each cluster category to train the LSTM model separately. This usually includes steps such as forward propagation, loss calculation, backpropagation, and weight update;
[0072] Evaluate the trained LSTM model using the validation set or test set to check its performance in predicting the output parameters;
[0073] The LSTM output results of samples of different categories are aggregated or integrated for subsequent analysis or decision-making.
[0074] (3) The mathematical model established by the steps: y(t) = Παx(t);
[0075] Among them, α represents a neural network, x(t)=[x1,x2,...xp] T p-dimensional input vector representing the input parameters, y(t) = [y1, y2, ...y q ] T q-dimensional output vector representing the output parameters, x1,x2,...x p Represents p input parameters, y1, y2, ...y q Represents q output parameters;
[0076] Specific:
[0077] Input parameter X = [x1, x2, x3, x4, x5] T =[R i ,A i ,M i ,E i ,S i ] T ,
[0078] Output parameter Y = [y1, y2, y3] T =[T i ,J i ,P i ] T ,
[0079] Further R i =[a i ,b i ,c i ,d i ,e i ], A i =[f i ,g i ,h i ,i i ],M i =
[0080] [j i ,k i ,l i ,m i ,n i ],E i =[o i ,p i ,q i ,r i ,s i ],S i =[t i ];
[0081] The mathematical model is:
[0082] [T i ,J i ,P i ] T =f{[R i ,A i ,M i ,E i ,S i ] T};
[0083] Among them: a i – The city where the room is located;
[0084] b i – Room length, width and height;
[0085] c i – Room window area;
[0086] d i --Room window orientation;
[0087] e i --Room function;
[0088] f i – Air conditioner model;
[0089] g i – Number of years the air conditioner has been used;
[0090] h i – Air conditioning cooling capacity;
[0091] i i --Heat capacity of air conditioner;
[0092] j i – How long the air conditioner is on;
[0093] k i – Air conditioning setting temperature;
[0094] l i --Air conditioning setting mode;
[0095] m i --Air conditioning wind speed;
[0096] n i – Power consumption of air conditioner;
[0097] o i – indoor temperature;
[0098] p i – Indoor humidity;
[0099] q i--Outdoor temperature;
[0100] r i -- Outdoor humidity;
[0101] s i – Solar radiation;
[0102] t i --PMV;
[0103] T i --Indoor temperature changes;
[0104] J i – Changes in subjective thermal perception;
[0105] P i – Changes in power consumption;
[0106] Get the input p-dimensional vector x(t) = [x1, x2, ... xp] T For a 20-dimensional input vector: x(t) = [a i ,b i ,c i ,d i ,e i , f i ,g i ,h i ,i i , j i ,k i ,l i ,m i ,n i , o i ,p i ,q i ,r i ,s i ,t i ] T , q-dimensional vector y(t)=[y1,y2,...y q ] T is a 3D output vector: y(t) = [T i ,J i ,P i ] T .
[0107] In summary, the flexible energy incentive method for temperature prediction based on the "feedforward" model first obtains the input parameters and output parameters of the training samples, uses the long short-term memory neural network algorithm to obtain the mathematical model representing the mapping relationship between the input parameters and the output parameters of the training samples, and obtains the neural network control system, that is, the "feedforward" model calculation system, and then installs smart sensors in the room, and connects the smart sensors to the user's mobile phone APP through signal connection. Then, by setting up smart sensors, the temperature inside and outside the room can be detected, and the monitoring data is transmitted to the user's mobile phone APP through signals. Then, the future room temperature change curve is calculated through the "feedforward" model calculation system. When the user needs to turn on the air conditioner, the user's preset temperature curve is input to calculate the room temperature change curve after turning on the air conditioner. Finally, the user decides whether to turn on the air conditioner or delay turning it on. If the user feels that the electricity bill is too high, the set temperature can be modified and the room temperature change curve can be recalculated.
[0108] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.
Claims
1. A flexible energy incentive method for temperature prediction based on a "feedforward" model, characterized by: Step 1: Obtain input parameters and output parameters of training samples; Step 2: Use the long short-term memory neural network algorithm to obtain a mathematical model representing the mapping relationship between the input parameters and output parameters of the training sample, and obtain a neural network control system, that is, a "feedforward" model calculation system; Step 3: Install the smart sensor in the room and connect it to the user's mobile phone APP through signal connection; Step 4: Substitute the input parameters into the system for prediction calculation to obtain the prediction results, including temperature change, thermal sensation change, and power consumption change, which are displayed on carriers such as web pages or mobile APPs through signal connections; Step 5: After checking the temperature change curve and combining it with the real-time electricity price, the user can then independently choose whether to turn on the room air conditioner, adjust the set temperature, or pre-cool the room by turning on the air conditioner in advance; Step 6: Determine the temperature for turning on the room air conditioner according to the user's selection, and input the set temperature into the "feedforward" model; Step 7: Based on the room temperature curve calculated by the "feedforward" model and the real-time electricity fee data and whether the air conditioner will be at a peak electricity consumption period after it is turned on, choose whether to turn on the air conditioner in advance or delay turning it on; Step 8: The system acquires new parameters through continuous monitoring for reinforcement learning and updates the APP version.
2. According to claim 1, a flexible energy incentive method for temperature prediction based on a "feedforward" model is characterized in that: Get the input and output parameters of the training sample, including: The input parameters and output parameters are obtained through experiments or actual room monitoring, wherein the input parameters include room information, air conditioning information, operating parameters, and environmental parameters; the output parameters include indoor temperature changes, subjective thermal sensation changes, and power consumption changes.
3. According to claim 1, a flexible energy incentive method for temperature prediction based on a "feedforward" model is characterized in that: The long short-term memory neural network algorithm is used to obtain the representation of the training samples.
4. A mathematical model of the mapping relationship between input parameters and output parameters, and obtains a neural network control system, including: Collect training samples containing control system inputs and expected outputs through historical data, simulation environments, or actual operations; The input data in the training sample is brought into the parameter-initialized neural network model, the output is obtained through forward propagation calculation, the output of the model is compared with the expected output in the training sample, and the error is calculated; Based on the error, the back propagation algorithm is used to update the parameters of the neural network to reduce the error. This process is repeated until the error reaches a preset threshold or a predetermined number of training rounds is reached; The data to be controlled is input into the trained neural network model, and the model will output the corresponding control signal to drive the actuator of the control system, thereby achieving control of the system.
5. According to claim 1, a flexible energy incentive method for temperature prediction based on a "feedforward" model is characterized in that: Substitute the input parameters into the system for prediction calculation, obtain the prediction results, and display them on carriers such as web pages or mobile APPs through signal connection: including; Install sensors to monitor the operation and environmental parameters of indoor room air conditioning equipment to obtain baseline operating conditions; Enter the desired air conditioning control conditions in the "input section" of the web page or mobile APP, such as turning on / off, changing the set temperature, etc., to obtain the control information; Bring the benchmark operating conditions, control information, and other input information such as electricity prices into the trained algorithm for prediction calculation to obtain the prediction results; The prediction results: room temperature changes, comfortable temperature range, real-time electricity prices, electricity savings, electricity saving costs and PMV are displayed on the web page or mobile APP "Prediction Results" section through signal connection.
6. According to claim 1, a flexible energy incentive method for temperature prediction based on a "feedforward" model is characterized in that: After viewing the temperature change curve and combining it with the real-time electricity price, the user can then independently choose whether to turn on the room air conditioner, adjust the set temperature, or pre-cool the room by turning on the air conditioner in advance. When the user returns from get off work or work outside, he can view the temperature curve change graph calculated by the "feedforward" model, and determine whether the room temperature reaches the user's expected temperature when returning to the room, and then choose whether to turn on the air conditioner to pre-cool the room.
7. According to claim 1, a flexible energy incentive method for temperature prediction based on a "feedforward" model is characterized in that: The temperature for turning on the room air conditioner is determined according to the user's selection, and the set temperature is input into the "feedforward" model. After obtaining the calculation result of the "feedforward" model, the user can set the temperature of the air conditioner in advance and input the set temperature into the "feedforward" model calculation system.
8. According to claim 1, a flexible energy incentive method for temperature prediction based on a "feedforward" model is characterized in that: The room temperature curve calculated according to the "feedforward" model and the real-time electricity bill data, as well as whether the air conditioner will be in a peak electricity consumption period after it is turned on, can be used to choose whether to turn on the air conditioner in advance or delay turning on the air conditioner. The user can choose whether to turn on the air conditioner in advance through the temperature change curve calculated by the "feedforward" model calculation system, the real-time electricity bill data generated, and whether the air conditioner is in a peak electricity consumption period. The set air-conditioning temperature is continuously changed to obtain different temperature change curves and real-time electricity bills. The user can choose the air-conditioning temperature that he thinks is most suitable and then turn on the air conditioner.
9. According to claim 1, a flexible energy incentive method for temperature prediction based on a "feedforward" model is characterized in that: After being put into use, the system can continuously monitor and obtain new data. With the user's consent, the new input can be added to the training set for reinforcement training and the APP version can be updated. The smart sensor is installed in the room, and the indoor temperature, indoor humidity, wall temperature, outdoor temperature and radiation of the room are monitored by installing smart sensors inside and outside the room. The smart sensor includes a temperature sensor, a humidity sensor and a radiation intensity sensor. The smart sensor is connected to the user's mobile phone APP through a signal connection, and the temperature sensors, humidity sensors and radiation intensity sensors installed inside and outside the room are connected to the user's mobile phone APP through a signal connection.
10. An air conditioning system, characterized in that: The invention comprises an air conditioner and a server having the functions as described in any one of claims 1 to 8.
