A fault intelligent detection method and device for operation of a heat pump heating system

Abstract

The application relates to the technical field of fault detection, in particular to a kind of intelligent fault detection method and device for the operation of heat pump heating system, the method comprises the following steps: obtaining various operation data of heat pump heating system under normal conditions and various fault conditions;Various operation data is divided into multiple data segments;According to the data fluctuation characteristics of each data segment, each data segment is divided into noise data segment and normal data segment;According to the degree of influence of noise on each layer signal after wavelet decomposition of each noise data segment, the denoising threshold of each layer signal in each noise data segment is obtained, and the denoising threshold of each layer signal in normal data segment is set as a preset initial threshold, so as to denoise various operation data;According to the various operation data after denoising under various conditions, and the various operation data at the current time, the fault condition of heat pump heating system at the current time is obtained.The application improves the accuracy of intelligent fault detection by adaptively obtaining the denoising threshold.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fault detection, in particular to a fault intelligent detection method and device for operation of a heat pump heating system. BACKGROUND

[0002] Heat pump technology can effectively recover waste heat from factories by consuming a small amount of electricity, and can improve low-grade waste heat in industrial production to high-grade heat energy that can be directly used in production, which is one of the key technologies for realizing green transformation and energy saving and carbon reduction of industrial heating at present. At the same time, the stable operation of its heating system is the key to efficient use of resources.

[0003] During the operation of the heat pump heating system, the fault detection of the heat pump heating system has a high requirement for the accuracy of data, and low-quality data can cause false positives or false negatives. During the operation of the heat pump device, the frequency drive, high-power compressor, fan motor and water pump motor will generate a strong electromagnetic field during operation. When the sensor for fault detection works in a strong electromagnetic field environment, the signal will be disturbed by the high-frequency switching of power elements such as IGBT, causing the output data to be abnormal, and further causing the fault detection result of the heat pump heating system during operation to be misjudged. The filtering and denoising algorithm used in traditional data processing cannot effectively process the signal interference generated when the above situation occurs, so that the filtered data still cannot accurately represent the real running data, resulting in a low fault detection accuracy. SUMMARY

[0004] In order to solve the above technical problems, the purpose of the present application is to provide a fault intelligent detection method and device for operation of a heat pump heating system, and the technical solutions adopted are as follows:

[0005] In a first aspect, the present application provides a fault intelligent detection method for operation of a heat pump heating system, which comprises the following steps:

[0006] Obtaining various types of operation data of the heat pump heating system under normal conditions and various types of fault conditions;

[0007] Dividing the various types of operation data under various conditions into multiple data segments; according to the data fluctuation frequency and dispersion degree of each data segment in the various types of operation data, obtaining a non-noise interference factor of each data segment, so as to divide the data segments in the various types of operation data into noise data segments and normal data segments;

[0008] Extract the wavelet decomposition coefficients of each layer of signal after wavelet decomposition of each data segment; based on the energy intensity and energy concentration of all detail coefficients in the wavelet decomposition coefficients of each layer of signal in each noisy data segment, as well as the fluctuation regularity of all detail coefficients, obtain the noise influence weight of each layer of signal in each noisy data segment, and then combine it with the corresponding preset initial threshold to obtain the denoising threshold of each layer of signal in each noisy data segment; set the denoising threshold of each layer of signal in the normal data segment as the preset initial threshold;

[0009] Based on the denoising thresholds of each layer of signals in each data segment of various types of operational data, the denoised operational data of various types is obtained; then, based on the denoised operational data of various types under various conditions, as well as the operational data of the current moment, it is determined whether the heat pump heating system has a fault at the current moment, and the type of fault is obtained.

[0010] Preferably, the non-noise interference factor of each data segment refers to the ratio of the stationary variation coefficient of each data segment to the variance of all data in the corresponding data segment.

[0011] Preferably, the formula for calculating the stationary variation coefficient of each data segment is: In the formula, This represents the stationary variation coefficient of the a-th data segment; This is a sign discrimination function. When the input data is less than or equal to 0, its value is 0; when the input data is greater than 0, its value is 1. This represents the difference between the j-th data point and the (j+1)-th data point in the a-th data segment; This represents the difference between the (j+1)th and (j+2)th data points in the (a)th data segment; This represents the total number of data points in the data segment.

[0012] Preferably, the specific process of dividing the data segments in various types of operating data into noise data segments and normal data segments is as follows: data segments in various types of operating data with a non-noise interference factor less than a preset segmentation threshold are recorded as noise data segments, and data segments with a non-noise interference factor greater than or equal to the preset segmentation threshold are recorded as normal data segments.

[0013] Preferably, the method for obtaining the noise influence weight of each layer of signal in each noise data segment is as follows: calculating the mean instantaneous energy intensity of all detail coefficients of each layer of signal in each noise data segment; calculating the permutation entropy corresponding to the first-order difference sequence of the detail sequence of each layer of signal in each noise data segment; calculating the variance of the instantaneous energy intensity of all detail coefficients corresponding to each layer of signal in each noise data segment; the noise influence weight of each layer of signal in each noise data segment is positively correlated with the mean and the permutation entropy, and negatively correlated with the variance.

[0014] Preferably, the detail sequence of each layer of signal refers to the sequence composed of all detail coefficients in each layer of signal in temporal order.

[0015] Preferably, the formula for calculating the denoising threshold of each layer of signal in each noise data segment is as follows: In the formula, This represents the denoising threshold of the v-th layer signal in the k-th noisy data segment. This represents the normalized result of the noise influence weight of the v-th layer signal in the k-th noisy data segment; The preset initial threshold is the signal of layer v in the k-th noise data segment.

[0016] Preferably, the process of acquiring the various types of denoised operating data is as follows: removing wavelet decomposition coefficients smaller than their corresponding denoising thresholds from all layers of signals in all data segments of various types of operating data to obtain the denoised wavelet decomposition coefficients of each layer of signals in all data segments of various types of operating data; performing wavelet reconstruction on each data segment in this way; and splicing and fusing all reconstructed data segments according to their original order before segmentation to obtain the various types of denoised operating data.

[0017] Preferably, the specific process of determining whether the heat pump heating system has malfunctioned at the current moment and obtaining the category of the malfunction is as follows: Set corresponding category labels for various types of operating data under normal conditions and various malfunction conditions; construct a one-dimensional vector from all types of operating data at each sampling moment under each type of condition, denoted as the operating vector at each sampling moment under each type of condition; use the one-dimensional vector of all sampling moments under all conditions and its corresponding category label as input to the BP neural network, train the BP neural network, and output the trained malfunction detection model; input various types of operating data at the current moment into the malfunction detection model, output the category label at the current moment, and then determine whether the heat pump heating system has malfunctioned at the current moment based on the obtained category label, and obtain the category of the malfunction.

[0018] Secondly, embodiments of this application also provide an intelligent fault detection device for the operation of a heat pump heating system, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of the intelligent fault detection method for the operation of a heat pump heating system described in any one of the above-mentioned methods.

[0019] This application has at least the following beneficial effects:

[0020] This application addresses the problem that in detecting operational faults in heat pump heating systems, the collected operational data contains noise due to electromagnetic fields in the environment. Traditional wavelet threshold denoising algorithms filter noise by setting a fixed threshold, resulting in poor denoising performance and affecting the accuracy of fault detection. By analyzing the high-frequency irregular fluctuation characteristics of various operational data when subjected to EMI noise interference, a non-noise interference factor is constructed, which can distinguish between data segments affected by noise and normal data segments. By analyzing the data characteristics of the detail coefficients of each layer of signals after decomposition within the noisy data segment, a noise influence weight for each layer of signals is constructed, which can characterize the degree of noise influence on each layer of signals in each data segment. Therefore, the corresponding denoising threshold can be adaptively obtained according to the degree of noise influence on each layer of signals in each data segment, improving the accuracy of data denoising. This provides more accurate operational data for subsequent training of BP neural network models, thereby improving the accuracy of intelligent fault detection in heat pump heating systems. Attached Figure Description

[0021] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A flowchart illustrating the steps of an intelligent fault detection method for a heat pump heating system provided in one embodiment of this application;

[0023] Figure 2 This is a flowchart illustrating the process of obtaining the denoising threshold of each layer of signal in each noise data segment according to an embodiment of this application. Detailed Implementation

[0024] To further illustrate the technical means and effects adopted by this application to achieve the intended purpose of the invention, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a fault intelligent detection method and apparatus for the operation of a heat pump heating system proposed in this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0025] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0026] The following description, in conjunction with the accompanying drawings, details the specific scheme of the intelligent fault detection method and device for the operation of a heat pump heating system provided in this application.

[0027] Please see Figure 1 The diagram illustrates a flowchart of a fault intelligent detection method for the operation of a heat pump heating system according to an embodiment of this application. The method includes the following steps:

[0028] Step 1: Obtain various operating data of the heat pump heating system under normal conditions and under various fault conditions.

[0029] When performing fault detection on a heat pump heating system, it is crucial to consider that the operational data during the core circulation process of the heat pump plays a key role in determining whether a fault has occurred. Therefore, temperature sensors are installed on the pipe at the heat pump's main heating outlet to obtain the outlet water temperature in real time; temperature sensors are also installed on the pipe at the heat pump's main heating return outlet to obtain the return water temperature in real time; pressure sensors are installed on the pipe between the compressor's exhaust port and the four-way valve to obtain exhaust port pressure data; pressure sensors are installed on the return gas pipe before the compressor's suction port to obtain suction port pressure data in real time; temperature sensors are installed on the pipe wall close to the compressor's exhaust port to obtain the compressor's exhaust temperature in real time; and temperature sensors are installed on the pipe wall near the compressor's suction port to obtain the compressor's suction temperature in real time.

[0030] The system collects various operational data from the heat pump heating system within a timeframe of T minutes, under both normal and various fault conditions. These fault conditions include refrigerant leakage, four-way valve malfunction, and compressor malfunction. The operational data include outflow temperature, return temperature, exhaust pressure, suction pressure, compressor exhaust temperature, and compressor suction temperature. T is a preset data collection duration; in this embodiment, T is set to 60 minutes.

[0031] In this embodiment, the sampling frequency of various types of operational data is set to 10Hz.

[0032] Step 2: Divide the various types of operational data under various conditions into multiple data segments on an average basis; based on the data fluctuation frequency and dispersion of each data segment in various types of operational data, obtain the non-noise interference factor of each data segment, thereby dividing the data segments in various types of operational data into noisy data segments and normal data segments.

[0033] Furthermore, considering that the variable frequency drive, high-power compressor, fan motor, and water pump motor generate strong electromagnetic fields during the operation of the heat pump heating system, the signals collected by the sensor may be interfered with when operating in a strong electromagnetic field environment, resulting in severe noise fluctuations in the output data. Therefore, it is necessary to use a wavelet threshold denoising algorithm to filter and denoise the various operating data acquired under different conditions. The specific process is as follows:

[0034] The sensor signal noise in a heat pump system mainly originates from strong electromagnetic interference (EMI) generated by the inverter driver, high-power compressor, fan motor, and water pump motor. High-frequency switching of power components such as IGBTs in the inverter driver couples into the sensor signal lines via radiation and conduction, manifesting as random high-frequency noise superimposed on the signals. This generates high-frequency voltage spikes and current glitches, which are coupled into the signal lines of various sensors through spatial radiation or conduction, affecting the data collected by all types of sensors. The following analysis uses the i-th type of operating data under any given condition as an example.

[0035] Based on the above analysis, it can be seen that the high-frequency switching of power components such as IGBTs in the frequency converter driver will couple to the sensor signal line through radiation and conduction, which manifests as random high-frequency noise superimposed on the signal. These situations will generate high-frequency voltage spikes and current glitches. Analyzing the situation, at the time domain level, the original signal of the sensor changes relatively slowly and smoothly, and there is a strong correlation between data points. However, EMI noise is usually high-frequency and random, which manifests as rapid and irregular glitches and fluctuations superimposed on the signal. This leads to a larger difference between adjacent data points, a weakening of correlation, and an increase in entropy when affected.

[0036] Therefore, the i-th type of running data in any case is divided into multiple data segments of equal length N, where N is 20 in this embodiment. During the division process, if the number of data points in a data segment is less than N, the data is supplemented by padding with the mean.

[0037] As a preferred implementation, the stability coefficient of each data segment is obtained based on the data fluctuation frequency of each data segment in various types of operational data, which is used to characterize the stability of the change trend of each data segment in various types of operational data.

[0038] In this embodiment, the stationary change coefficient of the a-th data segment is denoted as... Its specific expression is: In the formula, This represents the stationary variation coefficient of the a-th data segment; This is a sign discrimination function. When the input data is less than or equal to 0, its value is 0; when the input data is greater than 0, its value is 1. This represents the difference between the j-th data point and the (j+1)-th data point in the a-th data segment; This represents the difference between the (j+1)th and (j+2)th data points in the (a)th data segment; N is the total number of data points in the data segment.

[0039] When spikes or glitch-like spikes occur, the signs of the data differences before and after the glitch location are different, so the ratio of their differences will be less than or equal to 0. Therefore, the more spikes there are in the a-th data segment, the more complex the calculated... The smaller the value will be; The smaller the value, the less stable the trend of data change within the a-th data segment.

[0040] Furthermore, taking the a-th data segment in the i-th type of operational data as an example, the dispersion of all data in the a-th data segment is calculated. The greater the dispersion, the greater the fluctuation of the data in the a-th data segment, and the more likely that the data segment is to be affected by EMI noise. The dispersion can be calculated using variance, standard deviation, or coefficient of variation; in this embodiment, variance is used for calculation.

[0041] As a preferred implementation, the non-noise interference factor of each data segment is obtained based on the steady-state variation coefficient and dispersion of each data segment in various types of operating data. This factor is used to characterize the possibility that the data in each data segment is not affected by EMI noise.

[0042] In this embodiment, the non-noise interference factor of the a-th data segment is denoted as... Its specific expression is: In the formula, This represents the non-noise interference factor of the a-th data segment; This represents the variance of all data in the a-th data segment; This represents the stationary variation coefficient of the a-th data segment; This is a preset constant used to prevent the denominator from being 0. Its value range is [0.01, 1], and in this embodiment it is taken as 0.1.

[0043] The greater the degree of EMI noise interference in the data segment a, the smaller the stationary coefficient and the larger the variance of that data segment. Therefore, the calculated... The smaller.

[0044] Furthermore, calculate the non-noise interference factor of all data segments in the i-th type of running data, use all non-noise interference factors as input to the Otsu thresholding method, output the segmentation threshold, and record it as the preset segmentation threshold. Data segments with non-noise interference factors less than the preset segmentation threshold are recorded as noisy data segments, and data segments with non-noise interference factors greater than or equal to the preset segmentation threshold are recorded as normal data segments.

[0045] Step 3: Extract the wavelet decomposition coefficients of each layer of signal after wavelet decomposition of each data segment; based on the energy intensity and energy concentration of all detail coefficients in the wavelet decomposition coefficients of each layer of signal in each noisy data segment, as well as the fluctuation regularity of all detail coefficients, obtain the noise influence weight of each layer of signal in each noisy data segment, and then combine it with the corresponding preset initial threshold to obtain the denoising threshold of each layer of signal in each noisy data segment; set the denoising threshold of each layer of signal in the normal data segment as the preset initial threshold.

[0046] The noisy data segments in the i-th type of running data are layered using wavelet decomposition. Specifically, the wavelet basis db4 is used to decompose the signal into L layers, where L is a preset number of layers. In this embodiment, L is set to 7. This yields the wavelet decomposition coefficients for each layer of the decomposed signal, which include approximation coefficients and detail coefficients. The detail coefficients refer to the high-frequency components of each layer of the signal. Wavelet decomposition is a well-known technique, and the specific process will not be elaborated further.

[0047] Next, further targeted processing was carried out on each noise data segment. Considering that when affected by EMI noise disturbance caused by the high-frequency switching of power components such as IGBTs in the frequency converter driver, the noise would cause the high-frequency components of each layer of signal to exhibit large amplitude, concentrated and irregular fluctuations.

[0048] Based on the above characteristics, the sequence of all detail coefficients in each layer of the signal in each noisy data segment, arranged in temporal order, is denoted as the detail sequence of each layer of the signal in each noisy data segment. This allows for the acquisition of the first-order difference sequence of the detail sequence of each layer of the signal. The permutation entropy of the first-order difference sequence is calculated. During the calculation, the embedding dimension is set to 3 and the delay to 1. The permutation entropy of the first-order difference sequence of the detail sequence of each layer of the signal is obtained. The permutation entropy can characterize the regular fluctuation of the first-order difference sequence. The larger the value, the more irregular and chaotic the fluctuation of the detail coefficients, indicating that the corresponding signal layer is more severely affected by noise. The calculation of permutation entropy is a well-known technique, and the specific process will not be elaborated further.

[0049] In a preferred embodiment, the noise impact weights of each layer of signals in each noise data segment are obtained based on the energy intensity and energy concentration of all detail coefficients in the wavelet decomposition coefficients of each layer of signals in each noise data segment, as well as the degree of fluctuation of all detail coefficients. These weights characterize the degree to which each layer of signals in each noise data segment is affected by EMI noise. The method for obtaining the noise impact weights is as follows: calculating the mean instantaneous energy intensity of all detail coefficients in each layer of signals in each noise data segment; calculating the permutation entropy corresponding to the first-order difference sequence of the detail sequence in each layer of signals in each noise data segment; calculating the variance of the instantaneous energy intensity of all detail coefficients corresponding to each layer of signals in each noise data segment. The noise impact weights of each layer of signals in each noise data segment are positively correlated with the mean and the permutation entropy, and negatively correlated with the variance. It should be noted that a positive correlation means that the dependent variable increases (decreases) as the independent variable increases (decreases), and a negative correlation means that the dependent variable decreases (increases) as the independent variable increases (decreases).

[0050] Preferably, in this embodiment, the noise influence weight of the v-th layer signal in the k-th noise data segment is denoted as... Its specific expression is: In the formula, This represents the noise impact weight of the v-th layer signal in the k-th noisy data segment. Let represent the mean instantaneous energy intensity of all detail coefficients of the v-th layer signal in the k-th noisy data segment. This represents the permutation entropy corresponding to the first-order difference sequence of the detail sequence of the v-th layer signal in the k-th noisy data segment. Let V represent the variance of the instantaneous energy intensity of all detail coefficients corresponding to the v-th layer signal in the k-th noisy data segment. These are preset constants. The calculation of the instantaneous energy intensity of each detail coefficient is a well-known technique, and the specific process will not be described in detail here.

[0051] The larger the value, the greater the energy of the detail coefficients corresponding to the v-th layer signal in the k-th noisy data segment, and the more severely the signal of that layer is affected by noise. The larger the value, the more irregular the fluctuation of the v-th layer signal in the k-th noisy data segment, and the more severely the signal of that layer is affected by noise. The smaller the value, the more concentrated the instantaneous energy intensity of the v-th layer signal in the k-th noise data segment, and the more severely the signal in that layer is affected by noise. Therefore, The larger the threshold, the more severely the signal at layer v in the k-th noisy data segment is affected by noise, and a larger threshold is needed for stronger noise filtering in the wavelet denoising process; conversely, a smaller threshold is needed for noise filtering.

[0052] In the wavelet denoising process, it is necessary to preset an initial threshold for each layer of signals in each data segment. In this embodiment, the preset initial threshold for each layer of signals in each data segment is set to... ,in, The standard deviation of the signals in each data segment is the standard deviation of the signals in each layer. This indicates the length of each layer of signal in each data segment. This represents a logarithmic function with the natural constant e as its base.

[0053] Furthermore, based on the noise impact weights of each layer of signals in each noisy data segment and a preset initial threshold, the denoising thresholds for each layer of signals in each noisy data segment are obtained. The process for obtaining the denoising thresholds for each layer of signals in each noisy data segment is as follows: Figure 2 As shown. In this embodiment, the denoising threshold of the v-th layer signal in the k-th noise data segment is denoted as... Its specific expression is: In the formula, This represents the denoising threshold of the v-th layer signal in the k-th noisy data segment. This represents the normalized result of the noise influence weight of the v-th layer signal in the k-th noisy data segment. The normalization method includes, but is not limited to, the sigmoid function and the tanh function. In this embodiment, the tanh function is used for normalization. The preset initial threshold value is the signal of the v-th layer in the k-th noise data segment. In this embodiment, it is set to a value of ,in, This represents the standard deviation of the signal in the v-th layer within the k-th noisy data segment. This represents the length of the v-th layer signal in the k-th noisy data segment. This represents a logarithmic function with the natural constant e as its base.

[0054] Furthermore, each normal data segment is also decomposed into L layers using wavelet basis db4, and the denoising threshold of each layer is set to its corresponding preset initial threshold.

[0055] Step 4: Based on the denoising threshold of each layer of signal in each data segment of various types of operating data, obtain the denoised operating data of various types; then, based on the denoised operating data of various situations and the operating data of the current moment, determine whether the heat pump heating system has a fault at the current moment, and obtain the type of fault.

[0056] Based on the denoising threshold of each layer of signals in each data segment, the wavelet decomposition coefficients of each layer of signals are denoised. If any wavelet decomposition coefficient is less than its corresponding denoising threshold, it is removed. Similarly, wavelet decomposition coefficients less than their corresponding denoising thresholds in all layers of signals in all data segments of the i-th type of running data are removed, resulting in the denoised wavelet decomposition coefficients of each layer of signals in all data segments of the i-th type of running data. Then, wavelet reconstruction is performed on each data segment using the wavelet decomposition coefficients of all layers of signals after denoising. All reconstructed data segments are then spliced ​​and fused according to the order before segmentation to obtain the denoised i-th type of running data.

[0057] Denoising was performed on each type of operational data under all conditions according to the steps described above, and the acquired data under each condition were labeled. Specifically, the category label for all types of operational data under normal conditions was set to 0, and the category labels for all types of operational data under refrigerant leakage fault, four-way valve fault, and compressor fault were set to 1, 2, and 3, respectively. All types of operational data at each sampling time under each condition were constructed into a one-dimensional vector, denoted as the operational vector at each sampling time under each condition. The one-dimensional vector of all sampling times under all conditions and its corresponding category label were used as input to a BP neural network. The input layer dimension was set to the data length of a one-dimensional vector, the activation function was set to the Softmax function, the loss function was set to the cross-entropy loss function, the optimizer was selected as the Adam optimizer, and the number of iterations was set to 100 rounds. The BP neural network was then trained, and the trained fault detection model was output. The trained fault detection model was connected to the terminal for real-time fault detection. The real-time acquired operational data at the current time was input into the fault detection model, and the category label at the current time was output. Then, based on the obtained category label, it was determined whether the heat pump heating system had a fault at the current time, and the fault category was obtained. The above methods can be used to more accurately detect the operation of heat pump heating systems.

[0058] Based on the same inventive concept as the above method, this application embodiment also provides an intelligent fault detection device for the operation of a heat pump heating system, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above-described intelligent fault detection methods for the operation of a heat pump heating system.

[0059] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, specific embodiments of this specification have been described above. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0060] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0061] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the principles of this application should be included within the protection scope of this application.

Claims

1. A method for intelligent fault detection in the operation of a heat pump heating system, characterized in that, The method includes the following steps: Acquire various operational data of the heat pump heating system under normal conditions and under various fault conditions; The various types of operational data under different conditions are divided into multiple data segments on an average basis. Based on the data fluctuation frequency and dispersion of each data segment in the various types of operational data, the non-noise interference factor of each data segment is obtained, thereby dividing the data segments in the various types of operational data into noisy data segments and normal data segments. Extract the wavelet decomposition coefficients of each layer of signal after wavelet decomposition of each data segment; based on the energy intensity and energy concentration of all detail coefficients in the wavelet decomposition coefficients of each layer of signal in each noisy data segment, as well as the fluctuation regularity of all detail coefficients, obtain the noise influence weight of each layer of signal in each noisy data segment, and then combine it with the corresponding preset initial threshold to obtain the denoising threshold of each layer of signal in each noisy data segment; set the denoising threshold of each layer of signal in the normal data segment as the preset initial threshold; Based on the denoising thresholds of each layer of signals in each data segment of various types of operational data, obtain the denoised operational data of various types; then, based on the denoised operational data of various situations and the operational data of the current moment, determine whether the heat pump heating system has a fault at the current moment and obtain the type of fault. The non-noise interference factor of each data segment refers to the ratio of the stationary variation coefficient of each data segment to the variance of all data in the corresponding data segment. The specific process of dividing the data segments in various types of operational data into noise data segments and normal data segments is as follows: data segments in various types of operational data with a non-noise interference factor less than a preset segmentation threshold are recorded as noise data segments, and data segments with a non-noise interference factor greater than or equal to the preset segmentation threshold are recorded as normal data segments. The formula for calculating the stationary variation coefficient of each data segment is as follows: In the formula, This represents the stationary variation coefficient of the a-th data segment; This is a sign discrimination function. When the input data is less than or equal to 0, its value is 0; when the input data is greater than 0, its value is 1. This represents the difference between the j-th data point and the (j+1)-th data point in the a-th data segment; This represents the difference between the (j+1)th and (j+2)th data points in the (a)th data segment; This represents the total number of data points in the data segment.

2. The intelligent fault detection method for the operation of a heat pump heating system as described in claim 1, characterized in that, The method for obtaining the noise influence weight of each layer of signal in each noise data segment is as follows: calculate the mean instantaneous energy intensity of all detail coefficients of each layer of signal in each noise data segment; calculate the permutation entropy corresponding to the first-order difference sequence of the detail sequence of each layer of signal in each noise data segment; calculate the variance of the instantaneous energy intensity of all detail coefficients corresponding to each layer of signal in each noise data segment; the noise influence weight of each layer of signal in each noise data segment is positively correlated with the mean and the permutation entropy, and negatively correlated with the variance.

3. The intelligent fault detection method for the operation of a heat pump heating system as described in claim 2, characterized in that, The detail sequence of each layer of signal refers to the sequence of all detail coefficients in each layer of signal arranged in temporal order.

4. The intelligent fault detection method for the operation of a heat pump heating system as described in claim 1, characterized in that, The formula for calculating the denoising threshold of each layer of signal in each noise data segment is as follows: In the formula, This represents the denoising threshold of the v-th layer signal in the k-th noisy data segment. This represents the normalized result of the noise influence weight of the v-th layer signal in the k-th noisy data segment; The preset initial threshold is the signal of layer v in the k-th noise data segment.

5. The intelligent fault detection method for the operation of a heat pump heating system as described in claim 1, characterized in that, The process of acquiring the various types of denoised operating data is as follows: remove the wavelet decomposition coefficients of all layers of signals in all data segments of various types of operating data that are less than their corresponding denoising thresholds, and obtain the denoised wavelet decomposition coefficients of each layer of signals in all data segments of various types of operating data. Then, perform wavelet reconstruction on each data segment respectively, and splice and fuse all reconstructed data segments according to their original order before segmentation to obtain the various types of denoised operating data.

6. The intelligent fault detection method for the operation of a heat pump heating system as described in claim 1, characterized in that, The specific process for determining whether the heat pump heating system has malfunctioned at the current moment and obtaining the type of malfunction is as follows: Set corresponding category labels for various types of operating data under normal conditions and various malfunction conditions; construct a one-dimensional vector from all types of operating data at each sampling moment under each condition, denoted as the operating vector at each sampling moment under each condition; use the one-dimensional vector of all sampling moments under all conditions and its corresponding category label as input to a BP neural network, train the BP neural network, and output the trained malfunction detection model; input various types of operating data at the current moment into the malfunction detection model, output the category label at the current moment, and then determine whether the heat pump heating system has malfunctioned at the current moment based on the obtained category label, and obtain the type of malfunction.

7. A fault intelligent detection device for the operation of a heat pump heating system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the intelligent fault detection method for the operation of a heat pump heating system as described in any one of claims 1-6.

Images