Air conditioning unit energy consumption prediction method and system, terminal and storage medium

Abstract

The invention relates to the technical field of artificial intelligence, and particularly provides an air conditioning unit energy consumption prediction method and system, a terminal and a storage medium. Converting the environment data and the operation data of the air conditioning unit into a plurality of time sequence data; fragment sequences with the same length are intercepted from the multiple pieces of time sequence data through time windows, the multiple fragment sequences are input into a pre-trained prediction model, the prediction model comprises a graph learning layer and a convolution layer, the convolution layer comprises a graph convolution module and a time convolution module, and a gating mechanism is arranged between nodes of the graph convolution module; and obtaining a predicted energy consumption value output by the prediction model. According to the method, the graph adjacency relation between the variables is adaptively learned, and the prediction model provided by the invention can accurately predict the energy consumption of the air conditioning unit in the absence of a predefined graph structure.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence technology, and specifically relates to a method, system, terminal and storage medium for predicting energy consumption of an air-conditioning unit. Background Art

[0002] The air conditioning system is the main component of energy consumption in buildings, accounting for about 40% of the total energy consumption of buildings. The system is mainly composed of chillers, cooling towers, refrigeration pumps, cooling pumps, and air conditioning terminals. In the entire air conditioning system, the energy consumption of the chiller accounts for the largest proportion, so previous studies on air conditioning systems have mainly focused on the energy consumption of the chiller. However, due to the strong coupling between the various devices within the air conditioning system, the degree of influence of each device on the performance of the entire air conditioning system varies. Therefore, when predicting the energy consumption of the air conditioning unit system, it is necessary to comprehensively consider a variety of complex factors. Summary of the invention

[0003] In view of the above-mentioned deficiencies in the prior art, the present invention provides an air-conditioning unit energy consumption prediction method, system, terminal and storage medium to solve the above-mentioned technical problems.

[0004] In a first aspect, the present invention provides a method for predicting energy consumption of an air conditioning unit, comprising: Use sensors to collect environmental data and air conditioning unit operation data; Converting the environmental data and the operating data of the air-conditioning unit into a plurality of time series data; Using a time window to extract fragment sequences of the same length from multiple time series data, the multiple fragment sequences are input into a pre-trained prediction model, wherein the prediction model includes a graph learning layer and a convolution layer, wherein the convolution layer includes a graph convolution module and a time convolution module, and a gating mechanism is provided between nodes of the graph convolution module; Obtain the predicted energy consumption value output by the prediction model.

[0005] In an optional embodiment, using sensors to collect environmental data and operating data of the air conditioning unit includes: Collecting environmental data using sensors, wherein the environmental data includes outdoor temperature and outdoor humidity; Sensors are used to collect the operating data of the air-conditioning unit. The operating data of the air-conditioning unit include the chilled water supply temperature, chilled return water temperature, chilled water supply flow, chilled return water flow, cooling water supply temperature, cooling water supply flow, cooling return water temperature, cooling return water flow, operating power, operating current, operating voltage, and power consumption of the chiller; the water pump frequency, water inlet pressure, water outlet pressure, operating power, operating current, operating voltage, and power consumption of the chiller; the water pump frequency, water inlet pressure, water outlet pressure, operating power, operating current, operating voltage and power consumption of the cooling pump; the fan operating frequency, outlet air temperature, operating power, operating current, operating voltage and power consumption at the air conditioning terminal.

[0006] In an optional implementation, the environmental data and the operating data of the air-conditioning unit are converted into a plurality of time series data, including: Set all sensors to collect data at the same frequency; Create a sequence for each variable, save the sensor detection value of the variable to the corresponding sequence, and obtain time series data of multiple variables.

[0007] In an optional embodiment, the prediction model includes: The graph learning layer is used to adaptively generate a graph adjacency matrix between variables, regularize the graph adjacency matrix through a ReLU activation function, and perform sparse processing on the processed graph adjacency matrix; The convolution layer includes m graph convolution modules and m time convolution modules, each graph convolution layer contains two hybrid jump propagation layers, the graph convolution modules are used as nodes, and the information selection between different nodes is realized by multiple gating mechanisms, and the gating mechanisms include update gates and reset gates; each graph convolution module is followed by a time convolution module, and the time convolution layer is composed of two one-dimensional expansion Inception layers, the output of one expansion Inception layer is processed by the tanh activation function to capture the key information in the signal; the other branch uses the sigmoid activation function as a gating mechanism to regulate the amount of information transmitted to the subsequent modules.

[0008] In a second aspect, the present invention provides an air conditioning unit energy consumption prediction system, comprising: A data acquisition module, used to collect environmental data and operating data of the air-conditioning unit using sensors; A data processing module, used for converting the environmental data and the operation data of the air-conditioning unit into a plurality of time series data; A data input module, used to extract fragment sequences of the same length from multiple time series data using a time window, and input the multiple fragment sequences into a pre-trained prediction model, wherein the prediction model includes a graph learning layer and a convolution layer, wherein the convolution layer includes a graph convolution module and a time convolution module, and a gating mechanism is provided between nodes of the graph convolution module; The result acquisition module is used to obtain the predicted energy consumption value output by the prediction model.

[0009] In an optional embodiment, the data acquisition module includes: A first collection unit, used to collect environmental data using a sensor, wherein the environmental data includes outdoor temperature and outdoor humidity; The second acquisition unit is used to use sensors to collect the operating data of the air-conditioning unit. The operating data of the air-conditioning unit include the chilled water supply temperature, chilled return water temperature, chilled water supply flow, chilled return water flow, cooling water supply temperature, cooling water supply flow, cooling return water temperature, cooling return water flow, operating power, operating current, operating voltage, and power consumption of the chiller; the water pump frequency, water inlet pressure, water outlet pressure, operating power, operating current, operating voltage, and power consumption of the freezing pump; the water pump frequency, water inlet pressure, water outlet pressure, operating power, operating current, operating voltage and power consumption of the cooling pump; the fan operating frequency, air outlet temperature, operating power, operating current, operating voltage and power consumption at the air conditioning terminal.

[0010] In an optional embodiment, the data processing module includes: The frequency setting unit is used to set all sensors to collect data at the same frequency; The sequence creation unit is used to create a sequence for each variable, save the sensor detection value of the variable to the corresponding sequence, and obtain time series data of multiple variables.

[0011] In an optional embodiment, the prediction model includes: The graph learning layer is used to adaptively generate a graph adjacency matrix between variables, regularize the graph adjacency matrix through a ReLU activation function, and perform sparse processing on the processed graph adjacency matrix; The convolution layer includes m graph convolution modules and m time convolution modules, each graph convolution layer contains two hybrid jump propagation layers, the graph convolution modules are used as nodes, and the information selection between different nodes is realized by multiple gating mechanisms, and the gating mechanisms include update gates and reset gates; each graph convolution module is followed by a time convolution module, and the time convolution layer is composed of two one-dimensional expansion Inception layers, the output of one expansion Inception layer is processed by the tanh activation function to capture the key information in the signal; the other branch uses the sigmoid activation function as a gating mechanism to regulate the amount of information transmitted to the subsequent modules.

[0012] In a third aspect, a terminal is provided, including: A memory, used for storing an energy consumption prediction program for an air-conditioning unit; A processor is used to implement the steps of the air conditioning unit energy consumption prediction method provided in the first aspect when executing the air conditioning unit energy consumption prediction program.

[0013] In a fourth aspect, a computer-readable storage medium is provided, on which an air-conditioning unit energy consumption prediction program is stored. When the air-conditioning unit energy consumption prediction program is executed by a processor, the steps of the air-conditioning unit energy consumption prediction method provided in the first aspect are implemented.

[0014] The beneficial effect of the present invention is that the air-conditioning unit energy consumption prediction method, system, terminal and storage medium provided by the present invention can clearly capture the relationship between the chilled water temperature and the power of the chiller, the relationship between the cooling water temperature and the outdoor environmental conditions, etc., so as to identify the key factors affecting energy consumption and their interaction. By adaptively learning the graph adjacency relationship between variables, the prediction model provided by the present invention can effectively reveal these complex relationships in the absence of a predefined graph structure; GRU is an improved recurrent neural network architecture, and its gating machine can determine the specific degree of participation of the information flow of each variable node when it is transmitted to the downstream processing link, which will further enhance the model's understanding of the trend of changes in the time dimension.

[0015] In addition, the invention has a reliable design principle, a simple structure and a very broad application prospect. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0017] Figure 1 is a schematic flow chart of a method according to an embodiment of the present invention.

[0018] Figure 2 is another schematic flow chart of a method according to an embodiment of the present invention.

[0019] Figure 3 is a schematic block diagram of a system according to an embodiment of the present invention.

[0020] Figure 4 A schematic diagram of the structure of a terminal provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings 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 should fall within the scope of protection of the present invention.

[0022] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0023] The air conditioning unit energy consumption prediction method provided in the embodiment of the present invention is executed by a computer device, and accordingly, the air conditioning unit energy consumption prediction system runs in the computer device.

[0024] Figure 1 is a schematic flow chart of a method according to an embodiment of the present invention. Figure 1 The execution subject may be an air conditioning unit energy consumption prediction system. According to different requirements, the order of the steps in the flow chart may be changed, and some may be omitted.

[0025] like Figure 1 As shown, the method includes: Step 110, using sensors to collect environmental data and operating data of the air conditioning unit; Step 120, converting the environmental data and the operating data of the air-conditioning unit into a plurality of time series data; Step 130, using a time window to respectively extract fragment sequences of the same length from a plurality of time series data, and inputting the plurality of fragment sequences into a pre-trained prediction model, wherein the prediction model comprises a graph learning layer and a convolution layer, wherein the convolution layer comprises a graph convolution module and a time convolution module, and a gating mechanism is provided between nodes of the graph convolution module; Step 140: Obtain the predicted energy consumption value output by the prediction model.

[0026] To facilitate understanding of the present invention, the energy consumption prediction method for an air-conditioning unit provided by the present invention is further described below based on the principle of the energy consumption prediction method for an air-conditioning unit of the present invention and in combination with the process of predicting energy consumption of an air-conditioning unit in an embodiment.

[0027] For details, please refer to Figure 2 , the energy consumption prediction methods of air-conditioning units include: S1. Use sensors to collect environmental data and operating data of air-conditioning units.

[0028] Environmental data are collected by sensors, and the environmental data include outdoor temperature and outdoor humidity; operating data of air-conditioning units are collected by sensors, and the operating data of air-conditioning units include the chilled water supply temperature, chilled return water temperature, chilled water supply flow, chilled return water flow, cooling water supply temperature, cooling water supply flow, cooling return water temperature, cooling return water flow, operating power, operating current, operating voltage, and power consumption of the chiller; the water pump frequency, water inlet pressure, water outlet pressure, operating power, operating current, operating voltage, and power consumption of the freezing pump; the water pump frequency, water inlet pressure, water outlet pressure, operating power, operating current, operating voltage and power consumption of the cooling pump; the fan operating frequency, outlet air temperature, operating power, operating current, operating voltage and power consumption at the air conditioning terminal.

[0029] Various types of sensor data refer to relevant data obtained from outdoor environment, core equipment of air conditioning system, inlet and outlet locations, etc., including but not limited to: outdoor temperature and humidity, chilled water supply and return water temperature and flow, cooling water inlet and outlet temperature and flow, chilled pump and cooling pump inlet and outlet water pressure and power, etc. The 40 core data are time-aligned, and the input parameters can be expressed as. tp consecutive timestamps and N The time series data between variables is shown in formula (1): (1) in, Defined as a specific timestamp The set of observation values ​​of all metrics at time , is shown in formula (2): (2) Among them, each data point Indicates The value of a variable at that point in time.

[0030] S2. Convert the environmental data and the operating data of the air-conditioning unit into a plurality of time series data.

[0031] Set all sensors to collect data at the same frequency; create a sequence for each variable, save the sensor detection value of the variable to the corresponding sequence, and obtain time series data of multiple variables.

[0032] In one example, the following steps are included: Select appropriate sensors according to monitoring requirements, such as temperature sensors, humidity sensors, voltage sensors, current sensors, etc. Connect the sensors to the data acquisition system (such as microcontrollers, single-chip microcomputers, computers, etc.). Through the programming or configuration interface, set all sensors to collect data at the same frequency (such as once per second, once per minute, etc.). This usually involves adjusting the sampling rate parameters of the sensor or setting the timing task of the data acquisition system.

[0033] Define a unique identifier (such as a variable name) for each variable to be monitored, and create a sequence (such as an array, list, database table, etc.) for each variable to store the time series data of the variable. At the beginning of the program, initialize the sequences of all variables to ensure that they can accommodate the expected number of data points.

[0034] Start the data acquisition system so that it triggers the sensors to sample data at the set frequency. Each time it is triggered, read the current detection value from each sensor. Record the timestamp of each data acquisition (optional, but helpful for subsequent data analysis). Append the read sensor data (possibly including the timestamp) to the sequence of the corresponding variable.

[0035] Check the validity of data regularly or in real time (e.g., whether it is within a reasonable range, whether there are outliers, etc.). Delete invalid or duplicate data points to ensure data quality. According to the amount of data and storage requirements, select the appropriate storage medium (e.g., memory, hard disk, cloud storage, etc.) and storage format (e.g., CSV, JSON, database, etc.).

[0036] S3. Using the time window, segment sequences of the same length are respectively extracted from multiple time series data, and the multiple segment sequences are input into a pre-trained prediction model to obtain a predicted energy consumption value.

[0037] The prediction model mainly includes a graph learning layer, m graph convolution modules, m time series convolution modules, and an output module. In addition to the main modules, it also includes residual connections and jump connections. The data from multiple sensors are used to construct time series samples using sliding window technology; then, the graph adjacency matrix describing the relationship between variables is adaptively generated in the graph learning layer, and the information of nodes and their neighbors is integrated in the graph convolution module; then, the data enters the time convolution module to capture multi-scale time features and manage time series information flow; finally, all features are sent to the output module to generate energy consumption forecasts for a period of time in the future.

[0038] The core modules of the prediction model include: The graph learning module in the present invention can adaptively generate a graph adjacency matrix between variables, thereby effectively capturing the implicit functional relationship between nodes. This adaptive process solves the problem of lack of predefined graph structure in traditional graph neural multivariate time series tasks. In addition, by using the ReLU activation function to regularize the graph adjacency matrix, the asymmetric characteristics of the graph adjacency matrix are realized, which helps to capture the unidirectional influence relationship between variables. Furthermore, by sparsely processing the graph adjacency matrix, the computational resource requirements of the model are significantly reduced, and the efficiency and scalability of the model are improved. The implementation method is shown in formulas (3)-(5): (3) (4) (5) in, E 1 and E 2 represents randomly initialized node embeddings that can be learned during training, which means that as the model training progresses, E 1 and E 2 will be automatically adjusted to optimize the prediction performance. In addition, if there is a graph adjacency matrix between nodes of different variables Z , you can E 1= E 2= Z。 , are model parameters, is a hyperparameter that controls node saturation. argtopk Returns the indices of the first k largest values ​​in a vector. Formula (5) uses ReLU The activation function transforms the graph adjacency matrix A Regularization, realizing the graph adjacency matrix A Furthermore, in order to realize the graph adjacency matrix A Sparsification, for the matrix A Each row in i , only keep the first k The elements corresponding to the maximum values ​​are set, and the remaining elements are set to zero. In this way, the graph adjacency matrix can be realized A The accuracy and efficiency of energy consumption prediction are improved by making the predictions sparse.

[0039] After the data is transformed by the 1×1 convolution module, the present invention uses the graph convolution module to integrate the information of the node and its neighboring nodes. In the specific implementation, each graph convolution layer contains two hybrid jump propagation layers for processing information transfer between variables. One of the hybrid jump propagation layers receives the graph adjacency matrix output from the graph learning layer. Aas input, and the other takes the graph adjacency matrix A The transposed form of A T The outputs of the two layers are then added together to capture and integrate the features of the information transmitted through each node in two directions, thereby reflecting the complex implicit structure of the data.

[0040] Among them, in the specific structure of the hybrid jump propagation layer, the information propagation and selection mechanisms between different variable nodes correspond to the horizontal and vertical paths respectively, thereby ensuring the bidirectional flow and effective aggregation of information in the graph network.

[0041] Specific implementation method: The information propagation between different variable nodes is shown in formula (6), where is a hyperparameter that controls the input node proportion. is the hidden state obtained in the previous layer. The standardized self-loop adjacency matrix between multivariate nodes is obtained by formula (7) to ensure that the information of each node not only interacts with its direct neighbors, but also fully considers its own state characteristics.

[0042] (6) (7) Among them, the self-loop matrix between multivariable nodes By the identity matrix With the graph adjacency matrix A The diagonal matrix is ​​formed by summing each column of . Specifically, The diagonal elements of represent the degree of each node (i.e., the sum of the weights of all edges connected to the node), including self-loop edges. The inverse matrix of and the graph adjacency matrix A and the identity matrix I By multiplying the sum of , we can achieve effective normalization of the graph structure.

[0043] Based on the above steps, an initial node information retention factor is added to the information propagation, so that during the propagation process, the node can not only maintain the local characteristics of the initial node, but also obtain updated neighbor information.

[0044] After that, the information selection between different nodes is realized by multiple GRU units. GRU consists of two gates, namely the update gate and the reset gate. This structure can use its gating mechanism to determine the specific degree of participation of each node and its information flow when it is passed to the downstream processing link. Specifically, the update gate is responsible for regulating the proportion of the old state retained and the degree to which the new state is added to the hidden state. Its function can be reflected by the following update expression. As shown in formula (8), the update gate Determines how many old states will be retained and how many candidate states will is added to the hidden state, To update the state of the door, I represents the identity matrix, Denotes the Hadamard product.

[0045] (8) GRU current candidate status The calculation formula is shown in formula (9): (9) In the formula, is the input vector, To reset the state of the door, The candidate set and and The weight parameter of the connection matrix multiplied by each other, tanh () represents the tangent S Type activation function.

[0046] Update Gate and reset gate The calculation formulas are shown in (10) and (11): (10) = ( ) (11) In the formula, ( ) represents the sigmoid function, , They are the weight parameters of the update gate and reset gate respectively.

[0047] Each graph convolution module is followed by a temporal convolution module, which is responsible for extracting high-level temporal features from the data exported by the previous module. The temporal convolution layer consists of two one-dimensional dilated Inception layers. The output of one of the dilated Inception layers is processed by the tanh activation function and used as a filter to capture key information in the signal; at the same time, the other branch uses the sigmoid activation function as a gating mechanism to regulate the amount of information passed to subsequent modules. This structural design can significantly enhance the model's ability to understand complex temporal dependencies.

[0048] Among them, the dilated Inception layer mainly combines one-dimensional convolution kernels of various sizes with dilated convolution technology. Different from the traditional Inception structure in the field of image processing, which often uses convolution kernels of different sizes such as 1x1, 3x3 and 5x5 to capture multi-scale features, we selected four specific sizes of one-dimensional convolution kernels: 1x2, 1x3, 1x6 and 1x7 according to the characteristics of one-dimensional time series data. Such a design can better adapt to and capture the inherent periodic patterns in time series, such as common periodic changes such as 7 days, 12 hours, 24 hours, 28 days and 60 minutes.

[0049] Specifically, the dilated convolution technique is used in this module to significantly expand the receptive field of the model without increasing the number of parameters or computational complexity, which is achieved by formula (12). As the number of network layers increases, the dilation index grows exponentially. This strategy enables the model to effectively identify and utilize implicit relationships in longer sequences even when the network depth remains relatively unchanged.

[0050] (12) in, is the receptive field size of the current layer, Represents the receptive field size of the previous layer, represents the size of the convolution kernel, s is the dilation exponent of the dilated convolution, layer The current number of layers.

[0051] Among them, the training method of the prediction model includes: Since the time intervals of the data collected by each sensor are exactly the same, no additional preprocessing is required for the 40 types of data. Next, the present invention will use sliding window technology to construct time series samples. Specifically, every 5 minutes is set as a time step, and the data of the past 1 hour (ie 12 time steps) is selected each time as the input sequence of a single sample. At the same time, the output target is defined as the energy consumption data for the next 3 time steps (ie the next 15 minutes), mainly including key indicators such as power consumption, operating power, operating current and operating voltage. Finally, the generated time series samples are divided into 75% for training and 25% for verification for model training and performance evaluation.

[0052] In order to improve the universality of the model, the constructed data set is selected for training using the cross-validation method. For example, at the beginning of the model training phase, 75% of the data set is selected as the training set and 25% of the data set is selected as the validation set for model training. The parameter configuration of MTGNN-GRU in the training phase is shown in the table: Table 1 MTGNN-GRU parameter configuration table

[0053] After training, the mean absolute error is used for the prediction results. MAPE To evaluate the model accuracy, when the calculated MAPE Less than the preset MAPE When , the energy consumption prediction model training is completed; MAPE The calculation formula of is shown in formula (13).

[0054] (13) In the formula, To predict the energy consumption data at the nth time point in the future, is the real data at the nth time point in the future.

[0055] The prediction and reasoning stage is the same as the prediction and reasoning stage of learning algorithms such as deep learning networks. The time series data collected by each sensor is input into the trained prediction model. The model will output the next few time steps (according to the seq_out_len setting, here it is 3 time steps, that is, the energy consumption data for the next 15 minutes is predicted, mainly including key indicators such as power consumption, operating power, operating current and operating voltage.

[0056] In some embodiments, the air conditioning unit energy consumption prediction system may include multiple functional modules composed of computer program segments. The computer program of each program segment in the air conditioning unit energy consumption prediction system may be stored in a memory of a computer device and executed by at least one processor to perform (see Figure 1 Description) Function of predicting energy consumption of air conditioning units.

[0057] In this embodiment, the air conditioning unit energy consumption prediction system can be divided into multiple functional modules according to the functions it performs, such as Figure 3 As shown. The functional modules of the system 300 may include: a data acquisition module 310, a data processing module 320, a data input module 330 and a result acquisition module 340. The module referred to in the present invention refers to a series of computer program segments that can be executed by at least one processor and can complete fixed functions, which are stored in a memory. In this embodiment, the functions of each module will be described in detail in subsequent embodiments.

[0058] A data acquisition module, used to collect environmental data and operating data of the air-conditioning unit using sensors; A data processing module, used for converting the environmental data and the operation data of the air-conditioning unit into a plurality of time series data; A data input module, used to extract fragment sequences of the same length from multiple time series data using a time window, and input the multiple fragment sequences into a pre-trained prediction model, wherein the prediction model includes a graph learning layer and a convolution layer, wherein the convolution layer includes a graph convolution module and a time convolution module, and a gating mechanism is provided between nodes of the graph convolution module; The result acquisition module is used to obtain the predicted energy consumption value output by the prediction model.

[0059] Optionally, as an embodiment of the present invention, the data acquisition module includes: A first collection unit, used to collect environmental data using a sensor, wherein the environmental data includes outdoor temperature and outdoor humidity; The second acquisition unit is used to use sensors to collect the operating data of the air-conditioning unit. The operating data of the air-conditioning unit include the chilled water supply temperature, chilled return water temperature, chilled water supply flow, chilled return water flow, cooling water supply temperature, cooling water supply flow, cooling return water temperature, cooling return water flow, operating power, operating current, operating voltage, and power consumption of the chiller; the water pump frequency, water inlet pressure, water outlet pressure, operating power, operating current, operating voltage, and power consumption of the freezing pump; the water pump frequency, water inlet pressure, water outlet pressure, operating power, operating current, operating voltage and power consumption of the cooling pump; the fan operating frequency, air outlet temperature, operating power, operating current, operating voltage and power consumption at the air conditioning terminal.

[0060] Optionally, as an embodiment of the present invention, the data processing module includes: The frequency setting unit is used to set all sensors to collect data at the same frequency; The sequence creation unit is used to create a sequence for each variable, save the sensor detection value of the variable to the corresponding sequence, and obtain time series data of multiple variables.

[0061] Optionally, as an embodiment of the present invention, the prediction model includes: The graph learning layer is used to adaptively generate a graph adjacency matrix between variables, regularize the graph adjacency matrix through a ReLU activation function, and perform sparse processing on the processed graph adjacency matrix; The convolution layer includes m graph convolution modules and m time convolution modules, each graph convolution layer contains two hybrid jump propagation layers, the graph convolution modules are used as nodes, and the information selection between different nodes is realized by multiple gating mechanisms, and the gating mechanisms include update gates and reset gates; each graph convolution module is followed by a time convolution module, and the time convolution layer is composed of two one-dimensional expansion Inception layers, the output of one expansion Inception layer is processed by the tanh activation function to capture the key information in the signal; the other branch uses the sigmoid activation function as a gating mechanism to regulate the amount of information transmitted to the subsequent modules.

[0062] Figure 4 The present invention provides a schematic diagram of the structure of a terminal 400 provided in an embodiment of the present invention. The terminal 400 can be used to execute the air-conditioning unit energy consumption prediction method provided in an embodiment of the present invention.

[0063] The terminal 400 may include: a processor 410, a memory 420 and a communication unit 430. These components communicate via one or more buses. Those skilled in the art will appreciate that the server structure shown in the figure does not limit the present invention, and it may be a bus structure or a star structure, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0064] The memory 420 can be used to store the execution instructions of the processor 410, and the memory 420 can be implemented by any type of volatile or non-volatile storage terminal or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. When the execution instructions in the memory 420 are executed by the processor 410, the terminal 400 can perform some or all of the steps in the following method embodiments.

[0065] The processor 410 is the control center of the storage terminal, and uses various interfaces and lines to connect various parts of the entire electronic terminal. It runs or executes software programs and / or modules stored in the memory 420, and calls data stored in the memory to perform various functions of the electronic terminal and / or process data. The processor can be composed of an integrated circuit (IC), for example, it can be composed of a single packaged IC, or it can be composed of multiple packaged ICs with the same or different functions. For example, the processor 410 can include only a central processing unit (CPU). In the embodiment of the present invention, the CPU can be a single computing core or multiple computing cores.

[0066] The communication unit 430 is used to establish a communication channel so that the storage terminal can communicate with other terminals, receive user data sent by other terminals or send user data to other terminals.

[0067] The present invention also provides a computer storage medium, wherein the computer storage medium may store a program, and when the program is executed, the program may include some or all of the steps in each embodiment provided by the present invention. The storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM).

[0068] Therefore, the present invention can clearly capture the relationship between chilled water temperature and chiller power, the relationship between cooling water temperature and outdoor environmental conditions, etc., so as to identify the key factors affecting energy consumption and their interactions. By adaptively learning the graph adjacency relationship between variables, the prediction model provided by the present invention can effectively reveal these complex relationships in the absence of a predefined graph structure; GRU is an improved recurrent neural network architecture, and its gating machine can determine the specific degree of participation of the information flow of each variable node when it is passed to the downstream processing link, which will further enhance the model's understanding of the changing trend in the time dimension. The technical effects that can be achieved by this embodiment can be found in the description above and will not be repeated here.

[0069] Those skilled in the art can clearly understand that the technology in the embodiments of the present invention can be implemented by means of software plus a necessary general hardware platform. Based on this understanding, the technical solution in the embodiments of the present invention, in essence or in other words, the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, and other media that can store program codes, including several instructions for enabling a computer terminal (which can be a personal computer, a server, or a second terminal, a network terminal, etc.) to execute all or part of the steps of the methods described in each embodiment of the present invention.

[0070] In this specification, the same or similar parts between the various embodiments can be referred to each other. In particular, for the terminal embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the description in the method embodiment.

[0071] In the several embodiments provided by the present invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are only schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of systems or modules, which can be electrical, mechanical or other forms.

[0072] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0073] In addition, each functional module in each embodiment of the present invention may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0074] Although the present invention has been described in detail with reference to the accompanying drawings and in combination with the preferred embodiments, the present invention is not limited thereto. Without departing from the spirit and essence of the present invention, a person of ordinary skill in the art may make various equivalent modifications or substitutions to the embodiments of the present invention, and these modifications or substitutions shall be within the scope of the present invention. Any person of ordinary skill in the art may easily think of changes or substitutions within the technical scope disclosed by the present invention, and these shall be within the scope of protection of the present invention.

Claims

1. A method for predicting energy consumption of an air conditioning unit, characterized in that: include: Use sensors to collect environmental data and air conditioning unit operation data; Converting the environmental data and the operating data of the air-conditioning unit into a plurality of time series data; Using a time window to extract fragment sequences of the same length from multiple time series data, the multiple fragment sequences are input into a pre-trained prediction model, wherein the prediction model includes a graph learning layer and a convolution layer, wherein the convolution layer includes a graph convolution module and a time convolution module, and a gating mechanism is provided between nodes of the graph convolution module; Obtain the predicted energy consumption value output by the prediction model.

2. The method according to claim 1, characterized in that Sensors are used to collect environmental data and air conditioning unit operation data, including: Collecting environmental data using sensors, wherein the environmental data includes outdoor temperature and outdoor humidity; Sensors are used to collect the operating data of the air-conditioning unit. The operating data of the air-conditioning unit include the chilled water supply temperature, chilled return water temperature, chilled water supply flow, chilled return water flow, cooling water supply temperature, cooling water supply flow, cooling return water temperature, cooling return water flow, operating power, operating current, operating voltage, and power consumption of the chiller; the water pump frequency, water inlet pressure, water outlet pressure, operating power, operating current, operating voltage, and power consumption of the chiller; the water pump frequency, water inlet pressure, water outlet pressure, operating power, operating current, operating voltage and power consumption of the cooling pump; the fan operating frequency, outlet air temperature, operating power, operating current, operating voltage and power consumption at the air conditioning terminal.

3. The method according to claim 1, characterized in that The environmental data and the operation data of the air-conditioning unit are converted into a plurality of time series data, including: Set all sensors to collect data at the same frequency; Create a sequence for each variable, save the sensor detection value of the variable to the corresponding sequence, and obtain time series data of multiple variables.

4. The method according to claim 3, characterized in that The prediction model includes: The graph learning layer is used to adaptively generate a graph adjacency matrix between variables, regularize the graph adjacency matrix through a ReLU activation function, and perform sparse processing on the processed graph adjacency matrix; The convolution layer includes m graph convolution modules and m time convolution modules, each graph convolution layer contains two hybrid jump propagation layers, the graph convolution modules are used as nodes, and the information selection between different nodes is realized by multiple gating mechanisms, and the gating mechanisms include update gates and reset gates; each graph convolution module is followed by a time convolution module, and the time convolution layer is composed of two one-dimensional expansion Inception layers, the output of one expansion Inception layer is processed by the tanh activation function to capture the key information in the signal; the other branch uses the sigmoid activation function as a gating mechanism to regulate the amount of information transmitted to the subsequent modules.

5. An air conditioning unit energy consumption prediction system, characterized in that: include: A data acquisition module, used to collect environmental data and operating data of the air-conditioning unit using sensors; A data processing module, used for converting the environmental data and the operation data of the air-conditioning unit into a plurality of time series data; A data input module, used to extract fragment sequences of the same length from multiple time series data using a time window, and input the multiple fragment sequences into a pre-trained prediction model, wherein the prediction model includes a graph learning layer and a convolution layer, wherein the convolution layer includes a graph convolution module and a time convolution module, and a gating mechanism is provided between nodes of the graph convolution module; The result acquisition module is used to obtain the predicted energy consumption value output by the prediction model.

6. The system according to claim 5, characterized in that The data acquisition module comprises: A first collection unit, used to collect environmental data using a sensor, wherein the environmental data includes outdoor temperature and outdoor humidity; The second acquisition unit is used to use sensors to collect the operating data of the air-conditioning unit. The operating data of the air-conditioning unit include the chilled water supply temperature, chilled return water temperature, chilled water supply flow, chilled return water flow, cooling water supply temperature, cooling water supply flow, cooling return water temperature, cooling return water flow, operating power, operating current, operating voltage, and power consumption of the chiller; the water pump frequency, water inlet pressure, water outlet pressure, operating power, operating current, operating voltage, and power consumption of the freezing pump; the water pump frequency, water inlet pressure, water outlet pressure, operating power, operating current, operating voltage and power consumption of the cooling pump; the fan operating frequency, air outlet temperature, operating power, operating current, operating voltage and power consumption at the air conditioning terminal.

7. The system according to claim 5, characterized in that The data processing module comprises: The frequency setting unit is used to set all sensors to collect data at the same frequency; The sequence creation unit is used to create a sequence for each variable, save the sensor detection value of the variable to the corresponding sequence, and obtain time series data of multiple variables.

8. The system according to claim 7, characterized in that The prediction model includes: The graph learning layer is used to adaptively generate a graph adjacency matrix between variables, regularize the graph adjacency matrix through a ReLU activation function, and perform sparse processing on the processed graph adjacency matrix; The convolution layer includes m graph convolution modules and m time convolution modules, each graph convolution layer contains two hybrid jump propagation layers, the graph convolution modules are used as nodes, and the information selection between different nodes is realized by multiple gating mechanisms, and the gating mechanisms include update gates and reset gates; each graph convolution module is followed by a time convolution module, and the time convolution layer is composed of two one-dimensional expansion Inception layers, the output of one expansion Inception layer is processed by the tanh activation function to capture the key information in the signal; the other branch uses the sigmoid activation function as a gating mechanism to regulate the amount of information transmitted to the subsequent modules.

9. A terminal, characterized in that: include: A memory, used for storing an energy consumption prediction program for an air conditioning unit; A processor is used to implement the steps of the air conditioning unit energy consumption prediction method as described in any one of claims 1 to 4 when executing the air conditioning unit energy consumption prediction program.

10. A computer-readable storage medium storing a computer program, characterized in that: The readable storage medium stores an air conditioning unit energy consumption prediction program, and when the air conditioning unit energy consumption prediction program is executed by the processor, the steps of the air conditioning unit energy consumption prediction method according to any one of claims 1 to 4 are implemented.

Images