A household air conditioning load prediction method based on building load prediction and matching
By constructing a TRNSYS building load calculation model and backpropagation neural network, combining meteorological and building parameters, the problem of building load not being considered in the air conditioner load prediction is solved, and the prediction accuracy and stability are improved, and it is suitable for a variety of home environments.
Patent Information
- Application Number
- CN202410502290.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-04-25
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2044-04-25
Smart Images

Figure CN118532779B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of air conditioning load prediction, and in particular relates to a household air conditioning load prediction method based on building load prediction and matching. Background Art
[0002] With the advancement of science and technology and social development, my country's urbanization rate is also gradually increasing, which leads to a gradual increase in demand for electricity on the demand side, and the problem of energy shortage is becoming increasingly prominent. Therefore, mastering the information on residents' electricity consumption behavior is one of the important methods to master the law of electricity consumption in a city. For the demand side, it is an important basis for making good electricity decisions and planning; for the supply side, it is not only an important basis for formulating electricity management policies, but also an important reference for implementing electricity safety supervision. Mastering household electricity consumption can organically unify the demand side and the supply side, reduce energy waste, and improve the efficiency of electricity use.
[0003] To understand the household electricity consumption, it is necessary to understand the load of various electrical appliances in the home. Among them, household air conditioners, as the main household appliances for regulating indoor temperature, are almost essential appliances in the home. Their high frequency of use makes them a major component of household electricity load. Therefore, load forecasting for household air conditioners is an important intermediate link in understanding the household electricity consumption pattern. The existing methods mainly collect and analyze the electrical signals of the equipment, extract features from the electrical signals, and use multiple electrical signals combined with machine learning and deep learning algorithms to identify and predict the load. However, the existing methods have certain shortcomings, mainly including:
[0004] When predicting air conditioning load, the relationship between air conditioning load and building load is not considered. It is only predicted based on the historical air conditioning load law combined with time series related algorithms. This prediction method can usually only predict the air conditioning load in the short term in the future. It is not stable enough and the calculation time and sample size are large, so the actual application is not ideal. Therefore, it is necessary to improve the existing household air conditioning load prediction method to improve the accuracy and convenience of the prediction. Summary of the invention
[0005] In view of the problems existing in the prior art, the present invention provides a household air-conditioning load prediction method based on building load prediction and matching, which solves the problems that the existing method does not consider the relationship between air-conditioning load and building load when predicting air-conditioning load, but only predicts according to the historical air-conditioning load law combined with time series related algorithms, lacks stability, and requires a large amount of time and sample size for calculation, and the actual application is not ideal.
[0006] To achieve the above object, the present invention adopts the following technical solution:
[0007] A household air conditioning load prediction method based on building load prediction and matching includes the following steps:
[0008] 1) Obtain meteorological parameters, building envelope parameters and air conditioning load. Meteorological parameters include historical and future hourly dry-bulb temperature, relative humidity, solar azimuth, solar horizontal beam radiation, solar horizontal scattered radiation, and room temperature obtained from meteorological websites. Building envelope parameters include building length, width, and height, building orientation, window-to-wall ratio in each direction, wall thickness, window thickness, wall thermal conductivity, and window thermal conductivity. Air conditioning load includes historical hourly actual air conditioning load values corresponding to meteorological parameters.
[0009] 2) Use TRNSYS to build a building load calculation model for calculating building loads. The model consists of a building load calculation module, a meteorological parameter import module, a solar zenith angle calculation module, an effective sky temperature calculation module, a dew point temperature calculation module, a building load output module, and parameter transfer relationship lines between modules. The meteorological parameter import module imports the historical meteorological parameters obtained from the meteorological website, is connected to the zenith angle calculation module, outputs the solar azimuth angle, solar horizontal beam radiation, and solar horizontal scattered radiation parameters, is connected to the effective sky temperature calculation module, outputs the dry bulb temperature, solar horizontal beam radiation, and solar horizontal scattered radiation parameters, is connected to the dew point temperature calculation module, outputs the dry bulb temperature and relative humidity parameters, is connected to the building load calculation module, outputs the dry bulb temperature, solar azimuth angle, and room temperature parameters; the solar zenith angle calculation module is connected to the building load calculation module, outputs the solar zenith angle parameters; the effective sky temperature calculation module is connected to the building load calculation module, outputs the effective sky temperature parameters; the dew point temperature calculation module is connected to the effective sky temperature calculation module, outputs the dew point temperature parameters, is connected to the building load calculation module, and outputs the relative humidity parameters; the building load calculation module determines the building structure through the building envelope structure parameters, is connected to the building load output module, and outputs the building load.
[0010] 3) Construct an air conditioning load matching function for matching household air conditioning loads. The function is a single-valued piecewise function whose independent variable is the building load and the dependent variable is the household air conditioning load. The number of value intervals of the independent variable building load is equal to the number of air conditioning loads, and each building load value interval corresponds to one air conditioning load.
[0011] 4) Construct a neural network for building load prediction. This network is a back propagation neural network for building load prediction. The network consists of one input layer, three hidden layers, and one output layer, and each layer is fully connected. The input layer of the network is a meteorological parameter node, the hidden layer is a feature extraction node, and the output layer is a building load node. In the established neural network, each node in the input layer corresponds to a meteorological parameter value, and the output layer node corresponds to the building load value. The input layer nodes are dry bulb temperature nodes, relative humidity nodes, solar azimuth nodes, solar horizontal beam radiation nodes, solar horizontal scattered radiation nodes, and room temperature nodes. The output layer is a building load value node.
[0012] 5) When training the adjustment network, each node in the input layer corresponds to the parameter values of the historical meteorological parameters, and the corresponding building load of the output layer node is calculated by the TRNSYS building load calculation model. After the network training and adjustment, each node in the input layer corresponds to the parameter values of the future meteorological parameters, and the output layer node outputs the predicted building load value. The predicted building load value is used as the input of the air-conditioning load matching function to obtain the predicted air-conditioning load value.
[0013] In summary, the present invention provides a household air conditioning load prediction method based on building load prediction and matching. The building load calculation model is constructed through TRNSYS, and the building load is calculated in combination with historical meteorological parameters and building envelope structure parameters. In combination with the household air conditioning load, an air conditioning load matching function is established to match the building load with the household air conditioning load. The meteorological parameters are used as input and the building load is used as output. The building load back propagation neural network prediction model is established and adjusted, and the building load calculated by the back propagation neural network load prediction model is matched with the household air conditioning load through the air conditioning load matching function to realize the load prediction of household air conditioning. The present invention is conducive to the management and planning of residents' electricity consumption behavior, and has a good reference and auxiliary role in the electricity consumption strategy and management of scenes where air conditioners are frequently used, such as hotels, student dormitories, apartments, residents, etc., and has a good application prospect.
[0014] Compared with the current technology, the beneficial effects of the present invention are as follows:
[0015] (1) Using meteorological parameters as air conditioning load prediction variables increases the analysis dimension and improves the prediction accuracy;
[0016] (2) In the air conditioning load forecasting, the building load is calculated and the building load is used as an intermediate variable, making full use of the correlation between the building and the air conditioning, thereby improving the reliability of the forecast results;
[0017] (3) Using a neural network combined with a piecewise function, a relatively simple model is used to express the user's habits and rules of using air conditioners while ensuring high prediction accuracy;
[0018] (4) In addition, the load forecasting method provided by the present invention has certain versatility and generalization, and can be applied to the load forecasting of air-conditioning equipment in a variety of household environments. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] Figure 1 It is a flow chart of the implementation method of the present invention;
[0020] Figure 2 It is a specific embodiment diagram of the present invention;
[0021] Figure 3 A diagram of a building load calculation model according to an embodiment of the present invention;
[0022] Figure 4 This is a diagram of a neural network model for building load prediction according to an embodiment of the present invention. DETAILED DESCRIPTION
[0023] To make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be described in detail below. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other implementation methods obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention.
[0024] like Figure 1 The figure shows a flow chart of an implementation method of the present application.
[0025] According to the building envelope parameters, meteorological parameters and household air conditioning type, a household air conditioning load prediction model is built, such as Figure 2 shown.
[0026] The air conditioning load prediction model is divided into two parts. The first part is the building load prediction part, and the second part is the air conditioning load prediction part. The building load prediction part consists of the building load calculation model built by TRNSYS and the building load prediction neural network model. Figure 3 Figure 4 As shown in formula (3), the air conditioning load prediction part is composed of the air conditioning load matching piecewise function.
[0027] The building load calculation model in the building load prediction part consists of a building load calculation module, a meteorological parameter import module, a solar zenith angle calculation module, an effective sky temperature calculation module, a dew point temperature calculation module, a building load output module and parameter transfer relationship lines between modules.
[0028] In the established building load calculation model, the meteorological parameter import module imports the historical hourly dry-bulb temperature, relative humidity, solar azimuth, solar horizontal beam radiation, solar horizontal scattered radiation, and room temperature meteorological parameters obtained from the meteorological website, is connected to the zenith angle calculation module, outputs solar azimuth, solar horizontal beam radiation, and solar horizontal scattered radiation parameters, is connected to the effective sky temperature calculation module, outputs dry-bulb temperature, solar horizontal beam radiation, and solar horizontal scattered radiation parameters, is connected to the dew point temperature calculation module, outputs dry-bulb temperature and relative humidity parameters, is connected to the building load calculation module, outputs dry-bulb temperature, relative humidity, solar azimuth, and room temperature parameters; the solar zenith angle calculation module is connected to the building load calculation module, outputs solar zenith angle parameters; the effective sky temperature calculation module is connected to the building load calculation module, outputs effective sky temperature parameters; the dew point temperature calculation module is connected to the effective sky temperature calculation module, outputs dew point temperature parameters; the building load calculation module determines the building structure through the building envelope structure parameters, is connected to the building load output module, and outputs the building load.
[0029] The building load prediction model in the building load prediction part is a back propagation neural network used for building load prediction. The network consists of an input layer, three hidden layers, and an output layer, and each layer is fully connected.
[0030] The input layer of the network is the meteorological parameter node, the hidden layer is the feature extraction node, and the output layer is the building load node.
[0031] In the established neural network, each node in the input layer corresponds to a meteorological parameter value, and each node in the output layer corresponds to a building load value.
[0032] The input layer nodes are dry bulb temperature node, relative humidity node, solar azimuth node, solar horizontal beam radiation node, solar horizontal scattered radiation node, and room temperature node. The output layer is the building load value node.
[0033] The activation function of the input layer is set to GELU function, as shown in formula (1). The activation functions of the three hidden layers are GELU functions, and the number of nodes is 512, 256, and 128 respectively. The activation function of the output layer is GELU function, and the loss function is mean square error function, as shown in formula (2). The initial values of the bias and weight of each layer are random numbers.
[0034]
[0035]
[0036] In the above formula, x is the input value of each node, x in the input layer is the meteorological parameter plus the bias number, erf is the Gaussian error function, σ is the mean square error, N is the number of samples in the training batch, is the true value of the load in the batch, yi It is the load output value of the neural network in the batch. During training, the weights and biases of each layer are adjusted through error back propagation. The trained neural network is combined with the future meteorological parameters obtained from the meteorological website to output the building load forecast value.
[0037] The output building load is matched to the air conditioning load through the air conditioning load matching piecewise function, which is a single-valued piecewise function with the independent variable being the building load and the dependent variable being the household air conditioning load, as shown in formula (3).
[0038]
[0039] Independent variable Building load F b The number of value intervals is equal to the number of air conditioning loads n. Each building load value interval corresponds to an air conditioning load F a .
[0040] The building load calculated by the building load calculation model is combined with the historical hourly air conditioning load corresponding to the meteorological parameters to adjust the air conditioning load matching segmentation function, and determine the n+1 segmentation points P1, P2, ..., P of the independent variable building load. n ,P n+1 Take the value and get the piecewise function of air conditioning load matching between the corresponding building and air conditioner.
[0041] The building load prediction value output by the above-mentioned building load prediction neural network is combined with the air-conditioning load matching piecewise function to obtain the air-conditioning load prediction value.
[0042] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.
Claims
1. A household air conditioning load prediction method based on building load prediction and matching, characterized in that: The following steps are involved: S1: Obtain meteorological parameters, building envelope parameters and air conditioning load related parameters; S2: Use TRNSYS to build a building load calculation model for calculating building loads; S3: constructing an air conditioning load matching function for matching household air conditioning loads, wherein the independent variable of the air conditioning load matching function is the building load, and the dependent variable is a single-valued piecewise function of the household air conditioning load; S4: constructing a neural network for building load prediction, wherein the neural network is a back propagation neural network for building load prediction, the network consists of an input layer, three hidden layers, and an output layer, and each layer is fully connected; S5: Train and adjust the neural network constructed in step S4, where each node in the input layer corresponds to each parameter value of the historical meteorological parameters, and each node in the output layer corresponds to the building load calculated by the TRNSYS building load calculation model. After the network training and adjustment, each node in the input layer corresponds to each parameter value of the future meteorological parameters, and each node in the output layer outputs the predicted building load value. The predicted building load value is used as the input of the air-conditioning load matching function to obtain the predicted air-conditioning load value.
2. A household air conditioning load prediction method based on building load prediction and matching according to claim 1, characterized in that: The meteorological parameters in step S1 include historical and future hourly dry-bulb temperature, relative humidity, solar azimuth, solar horizontal beam radiation, solar horizontal scattered radiation, and room temperature obtained from a meteorological website; the building envelope parameters include building length, width, and height, building orientation, window-to-wall ratio in each direction, wall thickness, window thickness, wall thermal conductivity, and window thermal conductivity; the air-conditioning load includes the historical hourly actual air-conditioning load value corresponding to the meteorological parameters.
3. A household air conditioning load prediction method based on building load prediction and matching according to claim 1, characterized in that: The building load calculation model in step S2 is composed of a building load calculation module, a meteorological parameter import module, a solar zenith angle calculation module, an effective sky temperature calculation module, a dew point temperature calculation module, a building load output module and parameter transfer relationship lines between the modules.
4. A household air conditioning load prediction method based on building load prediction and matching according to claim 3, characterized in that: The meteorological parameter import module imports the historical meteorological parameters obtained from the meteorological website, is connected to the zenith angle calculation module, outputs the solar azimuth angle, the solar horizontal beam radiation, and the solar horizontal scattered radiation parameters, is connected to the effective sky temperature calculation module, outputs the dry bulb temperature, the solar horizontal beam radiation, and the solar horizontal scattered radiation parameters, is connected to the dew point temperature calculation module, outputs the dry bulb temperature and relative humidity parameters, is connected to the building load calculation module, outputs the dry bulb temperature, the solar azimuth angle, and the room temperature parameters; the solar zenith angle calculation module is connected to the building load calculation module, outputs the solar zenith angle parameters; the effective sky temperature calculation module is connected to the building load calculation module, outputs the effective sky temperature parameters; the dew point temperature calculation module is connected to the effective sky temperature calculation module, outputs the dew point temperature parameters, and is connected to the building load calculation module, outputs the relative humidity parameters.
5. A household air conditioning load prediction method based on building load prediction and matching according to claim 4, characterized in that: The building load calculation module determines the building structure through the building envelope structure parameters, is connected to the building load output module, and outputs the building load.
6. A household air conditioning load prediction method based on building load prediction and matching according to claim 1, characterized in that: In step S3, the number of value intervals of the independent variable building load is equal to the number of air-conditioning loads, and each building load value interval corresponds to one air-conditioning load.
7. A household air conditioning load prediction method based on building load prediction and matching according to claim 1, characterized in that: In step S4, the input layer of the network is a meteorological parameter node, the hidden layer is a feature extraction node, and the output layer is a building load node; in the established neural network, each node in the input layer corresponds to a meteorological parameter value, and the output layer node corresponds to a building load value.
8. A household air conditioning load prediction method based on building load prediction and matching according to claim 7, characterized in that: The input layer nodes are dry bulb temperature node, relative humidity node, solar azimuth node, solar horizontal beam radiation node, solar horizontal scattered radiation node, and room temperature node; the output layer nodes are building load value nodes.
9. A household air conditioning load prediction method based on building load prediction and matching according to claim 8, characterized in that: The activation function of the input layer is set to GELU function, as shown in formula (1), the activation functions of the three hidden layers are GELU functions, the number of nodes is 512, 256, and 128 respectively, the activation function of the output layer is GELU function, the loss function is mean square error function, as shown in formula (2), and the initial values of the bias and weight of each layer are random numbers; In the formula, x is the input value of each node, x in the input layer is the meteorological parameter plus the bias number, erf is the Gaussian error function; σ is the mean square error, N is the number of samples in the training batch, is the true value of the load in the batch, y i It is the load output value of the neural network in the batch. During training, the weights and biases of each layer are adjusted through error back propagation. The trained neural network is combined with the future meteorological parameters obtained from the meteorological website to output the building load forecast value.
