A method for monitoring and evaluating air conditioning energy efficiency based on real-time data acquisition

Abstract

This invention relates to the field of air conditioning system energy efficiency monitoring and industrial energy consumption assessment technology, and discloses a method for monitoring and assessing air conditioning energy efficiency based on real-time data acquisition. The method involves: acquiring instantaneous active power at the air conditioning power supply end and forming a power sequence according to a set frequency and duration; generating window data using equal-length sliding windows, setting the frequency resolution and discrete frequency points according to the window duration, and constructing basis functions; calculating the amplitude and average value of the frequency projection coefficients; reconstructing the power using the average value and finite-frequency components and taking the residuals; collecting supply airflow and inlet / outlet air temperatures to calculate heat, and constructing energy efficiency indicators by combining them with responsive high-frequency spectra; measuring the differences between adjacent windows using spectral feature vectors and adaptively dividing operating conditions; aggregating energy efficiency and deviation within each operating condition, and marking energy efficiency anomaly windows.

Description

Technical Field

[0001] This invention relates to the field of air conditioning system energy efficiency monitoring and industrial energy consumption assessment technology, specifically to an air conditioning energy efficiency monitoring and assessment method based on real-time data acquisition. Background Technology

[0002] In existing technologies, energy efficiency assessments of factory workshop air conditioning systems often involve using electricity meters or building automation systems to collect electricity consumption data for the air conditioning circuits. This data, combined with parameters such as operating time and rated cooling capacity, is used to calculate energy consumption per unit time or per unit output. Some systems introduce simple time-series analysis or threshold alarms to detect short-term power fluctuations exceeding limits. However, these methods are mostly based on average power over minutes or longer periods, making it difficult to reflect the true energy consumption behavior of equipment under conditions of frequent start-stop cycles, sudden load changes, and other multi-condition switching. They only provide an overall result of "how much electricity is consumed," failing to reveal the subtle changes in energy consumption structure and energy efficiency levels.

[0003] In factory workshops, production processes often involve multiple stages, shifts, and overlapping loads, resulting in numerous short-term spikes and irregular fluctuations in the power curves of air conditioning systems. Traditional time-domain-based energy efficiency assessment methods typically treat these high-frequency fluctuations simply as noise or transient disturbances, failing to distinguish between "stable baseline loads" and "high-frequency energy consumption caused by switching operating conditions or improper control" from a spectral structure perspective. This makes it difficult to identify the ineffective energy consumption hidden behind the total power consumption. Furthermore, most existing methods conduct overall assessments on a daily, shift, or production line basis, lacking independent energy efficiency evaluations of individual operating conditions or segments within fine-grained time windows. This can easily lead to statistical averaging and masking of efficient and inefficient operating conditions. On the other hand, many energy efficiency monitoring schemes only focus on electrical energy consumption, insufficiently collecting thermal parameters such as airflow and inlet / outlet temperature difference, or using them only as environmental monitoring indicators without strictly correlating them with electrical power data at the same time scale. Consequently, it is impossible to establish a correspondence between "cooling and heating output" and "energy input" within the same time window, making it difficult to obtain dimensionless energy efficiency indicators with physical meaning. In addition, the determination of operating condition changes in existing technologies mostly relies on empirical thresholds or manual rules, which are not closely related to physical quantities such as the response time constant of air conditioning equipment and the resolution of the acquisition system. This makes it difficult to unify standards between different projects and is not conducive to objectively comparing energy efficiency under complex operating conditions.

[0004] Therefore, this case aims to propose a method for monitoring and evaluating the energy efficiency of air conditioners based on real-time data acquisition. First, the solution synchronously acquires the electrical and thermal signals of the air conditioner operation through a high-frequency power acquisition module and temperature and flow sensors, forming a multi-dimensional data sequence. Then, a series of spectrum analysis algorithms, such as sliding time window division, adaptive frequency resolution setting, and discrete frequency point projection, are used to extract the power spectrum characteristics and energy efficiency indicators within each time window. Next, by quantifying and clustering the differences in the spectrum structure of adjacent windows, different operating conditions are automatically identified. Finally, the average energy efficiency and deviation are statistically analyzed within each operating condition segment, and abnormally low energy efficiency points are marked in real time. Summary of the Invention

[0005] This invention provides a method for monitoring and evaluating the energy efficiency of air conditioners based on real-time data acquisition, which helps to solve the problems mentioned in the background art.

[0006] This invention provides the following technical solution: a method for monitoring and evaluating the energy efficiency of air conditioners based on real-time data acquisition, comprising:

[0007] The instantaneous active power is collected at the power input terminal of the air conditioner, the sampling frequency and sampling duration are set, and a power sequence is formed according to the time interval;

[0008] The power sequence is divided into equal-length sliding time windows according to a preset number of sampling points. The number of sampling points and duration of each time window are obtained to generate power data segments for the time windows.

[0009] Within each time window, the window-level frequency resolution is set according to the window duration to limit the analysis frequency range. Discrete frequency points are generated according to the window-level frequency resolution, and a set of frequency basis functions is constructed.

[0010] The projection coefficients of each frequency are calculated based on the power data segments within the time window and the set of frequency basis functions to obtain the frequency amplitude and the average value of the power samples.

[0011] The power sequence is reconstructed based on the power sample average value and finite frequency components. The reconstructed residual sequence is calculated and the spectral residual data is obtained.

[0012] Collect supply air volumetric flow rate and inlet and outlet air temperature, calculate heat transfer, and construct a spectrum energy efficiency index by combining high-frequency responsive spectrum set;

[0013] A spectral feature vector is constructed based on the frequency amplitude distribution of each time window, the spectral structure difference between adjacent time windows is calculated, and the operating condition segment is divided according to the spectral difference threshold.

[0014] Aggregate the spectral energy efficiency index within each operating condition segment, obtain the average energy efficiency index and energy efficiency deviation index, and output the time window of abnormally low energy efficiency values.

[0015] Optionally, the air conditioner power input terminal collects instantaneous active power, sets the sampling frequency and sampling duration, and forms a power sequence according to time intervals, specifically including:

[0016] A high-frequency power acquisition module is configured at the power input terminal of the workshop air conditioner so that the high-frequency power acquisition module continuously outputs sampled values ​​representing the instantaneous active power of the air conditioner power supply during air conditioner operation.

[0017] Set the sampling frequency, and set the sampling time interval between adjacent power sample values ​​according to the reciprocal of the sampling frequency, while keeping the sampling time interval constant;

[0018] Set the total sampling duration parameter, and obtain the number of sampling points based on the ratio of the total sampling duration to the sampling time interval. When the calculated number of sampling points is less than one sampling point, set the number of sampling points to one sampling point to ensure that there is at least one discrete sampling point.

[0019] Starting from the sampling start time, the continuous time of each sampling time is obtained in sequence according to the sampling time interval, and the sampling point number is assigned in sequence to establish a one-to-one correspondence between the sampling point number and the continuous time moment;

[0020] The instantaneous active power sample values ​​at each sampling time are recorded as discrete power samples. A power sequence is constructed by arranging all discrete power samples according to the sampling point number, and the discrete power samples are limited to non-negative real numbers.

[0021] Optionally, the step of dividing the power sequence into equal-length sliding time windows according to a preset number of sampling points, obtaining the number of sample points and duration of each time window, and generating a power data segment for the time window specifically includes:

[0022] Set the number of sampling points included in each sliding time window, so that the duration of the time window is equal to the product of the sampling time interval and the number of sampling points, and slide the window on the power sequence according to the number of sampling points;

[0023] Based on the total number of sampling points in the power sequence and the number of sampling points in the time window, obtain the number of time windows covering the entire sampling duration, and number the time windows in chronological order.

[0024] For each time window, consecutive sampling points corresponding to the time window number are extracted from the power sequence in chronological order. When the number of remaining sampling points in the last time window is less than the preset number of sampling points, the number of sampling points in the time window is set as the number of remaining sampling points.

[0025] For each time window, the actual duration is calculated based on the number of sampling points and the sampling time interval within the time window. The power values ​​of each sampling point within the time window are recorded to form a power data segment for the time window.

[0026] All power data segments within a time window are combined into a time window data sequence according to their time window numbers.

[0027] Optionally, within each time window, the step of setting a window-level frequency resolution based on the window duration, limiting the analysis frequency range, generating discrete frequency points according to the window-level frequency resolution, and constructing a set of frequency basis functions specifically includes:

[0028] For each time window, set the window-level frequency resolution based on the actual duration of the time window, and limit the interval between adjacent analysis frequencies;

[0029] Set the upper limit of the power signal analysis frequency and compare it with the Nyquist limit corresponding to half of the sampling frequency, and select the smaller value as the upper limit of the effective analysis frequency.

[0030] Within the range from zero frequency to the upper limit of the effective analysis frequency, discrete frequency points are divided according to window-level frequency resolution; no frequency points are set when the upper limit of the effective analysis frequency is zero; when the upper limit of the effective analysis frequency is greater than zero, the number of frequency components used in the time window and the value of each frequency are obtained.

[0031] Each frequency value is converted into an angular frequency according to the proportional relationship between frequency and angular velocity, forming an angular frequency index set associated with the time window;

[0032] Based on the angular frequency index set, cosine basis functions and sine basis functions are constructed at continuous time points to obtain a frequency basis function set consisting of a set of mutually orthogonal basis functions.

[0033] Optionally, the step of calculating the projection coefficients of each frequency based on the power data segment and the set of frequency basis functions within a time window to obtain the frequency amplitude and the average value of the power samples specifically includes:

[0034] For each time window, based on the starting sampling point number in the power sequence and the sampling point number within the time window, calculate the continuous time corresponding to each sampling point and establish the mapping relationship between the sampling point number and the continuous time.

[0035] Within each time window, for each frequency basis function, the power value of each sampling point in the power data segment of the time window is multiplied point by point with the cosine basis function value and sine basis function value at the corresponding continuous time moment, and the product of all sampling points is summed within the time window. Then, normalization is performed according to the number of sampling points in the time window to obtain the cosine projection coefficient and sine projection coefficient corresponding to each frequency component.

[0036] Within each time window, the frequency amplitude of each frequency component is calculated based on the sum of the squares of the cosine projection coefficient and the sine projection coefficient, forming a frequency amplitude sequence covering all frequency components.

[0037] Within each time window, the power values ​​of all sampled points in the power data segment of the time window are arithmetically averaged to obtain the average power sample value of the time window, and the average power sample value is used as the power baseline component.

[0038] Optionally, the step of reconstructing the power sequence based on the power sample average and finite frequency components, calculating the reconstructed residual sequence, and obtaining spectral residual data specifically includes:

[0039] For each time window, the power baseline component is combined with the cosine projection coefficient, sine projection coefficient and corresponding frequency basis function of each frequency component within the time window, and the reconstructed power value of each sampling point is calculated according to the sampling time order to form the time window power reconstruction sequence.

[0040] Within each time window, the power data segment of the time window is subtracted from the power reconstruction sequence of the time window point by point according to the sampling point to obtain the reconstruction residual sequence.

[0041] Optionally, the process of collecting the supply air volumetric flow rate and inlet / outlet air temperature, calculating heat transfer, and constructing a spectral energy efficiency index by combining a high-frequency responsive spectrum set specifically includes:

[0042] Within each time window, the average volumetric flow rate of the supply air, the average inlet air temperature, and the average outlet air temperature are collected, and the average temperature difference of the time window is obtained by the difference between the inlet and outlet temperatures.

[0043] Obtain the air volumetric specific heat capacity parameter, and calculate the total heat based on the air volumetric specific heat capacity, average air volumetric flow rate, average temperature difference and actual duration of the time window;

[0044] Obtain the shortest response time parameter of the air conditioning system. Based on the relationship between the shortest response time and the window-level frequency resolution, calculate the frequency component index corresponding to the lowest frequency that the system can respond to for each time window, and divide the frequency component corresponding to the lowest frequency and the frequency components above it into a high-frequency responsive spectrum set.

[0045] Within each time window, when there are no frequency components, the energy consumption of the set of high-frequency responsive spectra is set to zero, and the spectral energy efficiency index is also set to zero; when there are frequency components, the amplitude of the set of high-frequency responsive spectra is combined with the duration of the time window to calculate the high spectral energy consumption.

[0046] Within each time window, when the high-frequency energy consumption is greater than zero, the ratio of total heat to high-frequency energy consumption is used as the spectrum energy efficiency index; when the high-frequency energy consumption is equal to zero, the spectrum energy efficiency index is set to zero.

[0047] Optionally, the step of constructing a spectral feature vector based on the frequency amplitude distribution of each time window, calculating the spectral structure difference between adjacent time windows, and dividing the operating condition segment according to the spectral difference threshold specifically includes:

[0048] Within each time window, the frequency amplitudes of all frequency components are combined in frequency order to form a spectral feature vector;

[0049] For any two different time windows, the larger value is selected as the common bandwidth based on the window-level frequency resolution of the two, and adjacent frequency bands are divided according to the common bandwidth within the interval from zero frequency to the upper limit of the effective analysis frequency.

[0050] For each time window, the frequency amplitudes belonging to each frequency band are summed to obtain the band-level amplitude of each frequency band; when there are no frequency components in a certain frequency band, the band-level amplitude of that frequency band is set to zero.

[0051] For any two time windows, calculate the normalization ratio of the difference in band amplitude to the sum of the two band amplitudes and the minimum resolvable power value of the high-frequency power acquisition module in all frequency bands, and sum them in all frequency bands to obtain the spectral distance value representing the degree of difference in the spectral structure of the two time windows.

[0052] In chronological order, the spectral distance values ​​of adjacent time windows are statistically analyzed. When there are at least two time windows, the arithmetic mean of the spectral distance values ​​of all adjacent time windows is calculated and the arithmetic mean is used as the spectral difference threshold. When there is only one time window, the spectral difference threshold is set to zero.

[0053] For adjacent time windows, when the spectral distance value is greater than the spectral difference threshold, the next time window is taken as the starting time window of the new operating condition segment; when the spectral distance value is less than or equal to the spectral difference threshold, the next time window is divided into the same operating condition segment as the previous time window.

[0054] Optionally, the step of aggregating spectral energy efficiency indicators within each operating condition segment, obtaining average energy efficiency indicators and energy efficiency deviation indicators, and outputting time windows with abnormally low energy efficiency values ​​specifically includes:

[0055] All time windows are divided into multiple working condition segments according to the working condition identification results, and a set of windows containing the time window number of each working condition segment is established.

[0056] Within each operating condition segment, the spectral energy efficiency index of all time windows in the window set is arithmetically averaged to obtain the average energy efficiency index of the operating condition segment.

[0057] Within each operating condition segment, the arithmetic mean is calculated on the absolute difference between the spectral energy efficiency index and the average energy efficiency index for all time windows in the window set to obtain the energy efficiency deviation index within the operating condition segment.

[0058] Within each operating condition segment, when the spectral energy efficiency index of a certain time window is less than the average energy efficiency index of the operating condition segment minus the energy efficiency deviation index, the time window is marked as an abnormally low energy efficiency time window; when the spectral energy efficiency index of a certain time window is not less than the average energy efficiency index of the operating condition segment minus the energy efficiency deviation index, the time window is marked as a normal energy efficiency time window, and the operating condition segment remains unchanged.

[0059] The present invention has the following beneficial effects:

[0060] 1. This solution configures a high-frequency power acquisition module at the air conditioner power input end, ensuring the integrity and accuracy of instantaneous active power data. By setting an adjustable sampling frequency and sampling duration, it achieves adaptive adjustment of the number of sampling points to ensure that even the fewest sampling points can form an effective sequence. Compared with traditional single sampling or low-frequency sampling methods, this design improves the resolution of time-domain data and ensures the system's real-time response capability to sudden power changes. At the same time, based on the real-time power sequence construction process, it solves the data loss or overload problem caused by unreasonable sampling frequency settings in the industry, providing a solid data foundation for subsequent window division and spectrum analysis, and improving the monitoring system's ability to capture dynamic energy efficiency changes.

[0061] 2. The proposed solution divides the power sequence into equal-length sliding windows based on a preset number of sampling points. This ensures consistency in data length across windows while allowing the last window to adaptively adjust based on remaining sampling points, avoiding data waste. Compared to traditional fixed-window or non-sliding-window methods, this approach maintains statistical continuity while enabling fine-grained capture of short-term energy efficiency fluctuations through sliding. Furthermore, the number of sampling points and duration of each window are precisely calculated and recorded, providing accurate data for subsequent frequency resolution setting and feature extraction. This enhances the applicability and reliability of the monitoring algorithm across different time scales.

[0062] 3. This scheme dynamically sets the fundamental frequency interval based on the actual duration of each window, and combines it with Nyquist limiting constraints to select the optimal upper limit of the analysis frequency. Discrete frequency points are divided from zero frequency to the upper limit with a fixed step size, thereby generating a set of orthogonal sine and cosine basis functions. Unlike existing techniques that use static fixed basis functions or only consider a single frequency component, this method can automatically adapt the basis function resolution to the data characteristics of each window, ensuring the precision of spectral decomposition and anti-aliasing capability. At the same time, the establishment of the angular frequency index sequence provides a unified frequency domain reference framework for the subsequent projection of spectral data and the calculation of energy efficiency indicators, improving the accuracy and interpretability of spectral analysis.

[0063] 4. By projecting and normalizing the power samples within each window onto the sine and cosine basis functions, this method can simultaneously obtain the cosine and sine projection coefficients of each frequency component. Then, it integrates the comprehensive spectral amplitude through the amplitude function and calculates the average value of the power samples as the baseline component. Unlike the conventional mode that only uses the amplitude spectrum or ignores the baseline component, this method can completely preserve the DC component and harmonic characteristics of the signal, providing richer spectral information for the multi-scale analysis of power signals, thereby enhancing the accuracy of subsequent energy efficiency reconstruction and residual analysis.

[0064] 5. The proposed method reconstructs the power sequence through a linear combination of baseline components and finite harmonic components, and calculates the residual sequence with the measured values ​​to quantify the reconstruction error distribution. This method can establish a quantitative correspondence between the frequency domain and the time domain, and verify the spectral interpretability of actual power fluctuations. Compared with traditional direct residual methods or techniques based solely on time-domain statistics, the spectral information contained in the reconstructed residual sequence of this method has more physical meaning, which can be used to detect power anomalies or system faults in real time, and provide a reliable error basis for subsequent energy efficiency assessments, thereby reducing the false alarm rate and missed alarm rate of the monitoring system.

[0065] 6. Combining the average volumetric flow rate of the air supply at each window with the inlet and outlet temperature difference, the scheme calculates the window-level heat transfer and dynamically determines the set of high-frequency responsive spectra using the ratio of the system's shortest response time to frequency resolution. Based on this, a spectral energy efficiency index is constructed. Unlike single-dimensional methods that rely solely on thermodynamic calculations or spectral ratios, this scheme integrates thermodynamics and spectral characteristics to form a multi-physics coupled energy efficiency evaluation index. This effectively compensates for the blind spots in energy efficiency monitoring that rely solely on heat or electrical power data, and achieves a fine characterization of the transient energy efficiency of the air conditioning system.

[0066] 7. This scheme constructs a spectral feature vector and accumulates the difference in amplitude of the band level after division by common frequency band between adjacent windows with the normalized distance under uniform power resolution to form a spectral structure difference metric. Thus, the time window is automatically divided into different operating condition segments according to a preset threshold. Compared with traditional time-domain operating condition identification or simple statistical value threshold division, this method combines spectral structure information, which can more accurately identify the changes in the operating condition of the system under different loads and operating modes, significantly improve the precision of operating condition segmentation, and provide a more reliable basis for dynamic energy efficiency assessment and energy saving optimization.

[0067] 8. Within each operating condition segment, this scheme calculates the arithmetic mean and absolute deviation of the spectral energy efficiency indicators for all windows, and identifies inefficiency anomaly time windows accordingly. This method is based on statistical comparison within the same operating condition rather than a globally fixed threshold, and can adapt to differences in energy efficiency levels under different operating conditions. Compared with general thresholds or empirical methods, the judgment logic combining intra-segment averaging and deviation can more effectively distinguish between normal fluctuations and genuine energy efficiency anomalies, helping to take timely and targeted energy-saving or maintenance measures to improve system operating efficiency and reliability. Attached Figure Description

[0068] Figure 1 This is a schematic diagram of the process of the present invention. Detailed Implementation

[0069] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0070] Example, refer to Figure 1 A method for monitoring and evaluating the energy efficiency of air conditioners based on real-time data acquisition, comprising:

[0071] The instantaneous active power is collected at the power input terminal of the air conditioner, the sampling frequency and sampling duration are set, and a power sequence is formed according to the time interval;

[0072] The power sequence is divided into equal-length sliding time windows according to a preset number of sampling points. The number of sampling points and duration of each time window are obtained to generate power data segments for the time windows.

[0073] Within each time window, the window-level frequency resolution is set according to the window duration to limit the analysis frequency range. Discrete frequency points are generated according to the window-level frequency resolution, and a set of frequency basis functions is constructed.

[0074] The projection coefficients of each frequency are calculated based on the power data segments within the time window and the set of frequency basis functions to obtain the frequency amplitude and the average value of the power samples.

[0075] The power sequence is reconstructed based on the power sample average value and finite frequency components. The reconstructed residual sequence is calculated and the spectral residual data is obtained.

[0076] Collect supply air volumetric flow rate and inlet and outlet air temperature, calculate heat transfer, and construct a spectrum energy efficiency index by combining high-frequency responsive spectrum set;

[0077] A spectral feature vector is constructed based on the frequency amplitude distribution of each time window, the spectral structure difference between adjacent time windows is calculated, and the operating condition segment is divided according to the spectral difference threshold.

[0078] Aggregate the spectral energy efficiency index within each operating condition segment, obtain the average energy efficiency index and energy efficiency deviation index, and output the time window of abnormally low energy efficiency values.

[0079] By constructing a complete process from high-frequency data acquisition at the input end to operating condition segmentation and energy efficiency assessment, a full-link dynamic monitoring of energy efficiency during the operation of the air conditioning system is achieved. First, instantaneous active power is directly acquired at the air conditioning power input end and a power sequence is formed according to a configurable sampling frequency and duration, avoiding the slow response to short-term power fluctuations inherent in traditional empirical models or offline detection methods. Next, equal-length sliding time window segmentation and adaptive frequency resolution settings are introduced. By projecting the power data segments of each window using frequency domain basis functions, accurate extraction of different frequency components in the power signal is achieved, overcoming the limitation of weak separation of harmonic and periodic components in ordinary time-domain statistics. Finally, through residual sequence reconstruction analysis, minute deviations between the signal and the model are identified, providing early detection of changes in operating conditions and equipment anomalies. The system achieves higher sensitivity in detection. Simultaneously, it combines thermal information such as airflow and inlet / outlet temperature difference with high-frequency power consumption components to form a spectral energy efficiency index, allowing electrical and thermal analyses to complement each other and overcoming the limitation of single power analysis in reflecting thermal efficiency. Furthermore, the solution automatically divides operating condition segments using spectral feature vectors and adjacent window difference measurements, replacing the empirical practice of manually setting operating condition thresholds and enhancing the adaptability and robustness of operating condition identification. Finally, based on the statistical aggregation and deviation calculation of the spectral energy efficiency index within each operating condition segment, it automatically marks low-value windows of abnormal energy efficiency, achieving precise location and real-time early warning of energy efficiency anomalies. Overall, this claim integrates multi-dimensional real-time data, spectral analysis, and thermal calculation, improving the timeliness and accuracy of energy efficiency monitoring and providing a reliable basis for energy conservation, emission reduction, and intelligent operation and maintenance.

[0080] The air conditioner power input terminal collects instantaneous active power, sets the sampling frequency and sampling duration, and forms a power sequence according to time intervals, specifically including:

[0081] A high-frequency power acquisition module is configured at the power input terminal of the workshop air conditioner so that the high-frequency power acquisition module continuously outputs sampled values ​​representing the instantaneous active power of the air conditioner power supply during air conditioner operation.

[0082] Set the sampling frequency, and set the sampling time interval between adjacent power sample values ​​according to the reciprocal of the sampling frequency, while keeping the sampling time interval constant;

[0083] Set the total sampling duration parameter, and obtain the number of sampling points based on the ratio of the total sampling duration to the sampling time interval. When the calculated number of sampling points is less than one sampling point, set the number of sampling points to one sampling point to ensure that there is at least one discrete sampling point.

[0084] Starting from the sampling start time, the continuous time of each sampling time is obtained in sequence according to the sampling time interval, and the sampling point number is assigned in sequence to establish a one-to-one correspondence between the sampling point number and the continuous time moment;

[0085] The instantaneous active power sample values ​​at each sampling time are recorded as discrete power samples. A power sequence is constructed by arranging all discrete power samples according to the sampling point number, and the discrete power samples are limited to non-negative real numbers.

[0086] Further specific implementation steps include:

[0087] A high-frequency power acquisition module is configured at the power input terminal of the workshop air conditioner. The module outputs instantaneous active power. The unit is kilowatt; among which, For continuous time intervals The instantaneous active power of the air conditioner power supply measured above; It is a continuous-time variable;

[0088] Set the system sampling frequency to Based on this, the sampling time interval is calculated as follows: ;

[0089] Set the total sampling time to Then the number of sampling points is ;

[0090] Record each sampling moment: , ;in, For the first Each discrete sampling point corresponds to a continuous time interval; The sampling point number;

[0091] The sampled power sequence is constructed as follows: , ;in, To be at the sampling time Discrete active power values ​​collected from the device; It is the set of nonnegative real numbers;

[0092] The power sequence was obtained: ;in, It is an ordered sequence consisting of all discrete power samples.

[0093] By configuring a high-frequency power acquisition module at the power input of the workshop air conditioner and automatically calculating the number of sampling points based on a preset sampling frequency and total duration, this solution solves the problem of adapting to dynamic changes in the number of sampling points in the on-site installation environment. A minimum sampling point guarantee mechanism ensures effective power data is obtained even with extremely short sampling durations, avoiding data loss or analysis errors due to insufficient sampling time. Real-time mapping of sampling point numbers to continuous time points enables accurate calibration of time-series data. This method not only meets the stability requirements of high-frequency real-time sampling in industrial settings but also provides a precise time reference for subsequent sliding window segmentation and spectrum projection, eliminating analysis errors caused by uneven sampling or data fragmentation, thereby improving the reliability and completeness of energy efficiency monitoring.

[0094] The step of dividing the power sequence into equal-length sliding time windows according to a preset number of sampling points, obtaining the number of sampling points and duration of each time window, and generating a power data segment for the time window specifically includes:

[0095] Set the number of sampling points included in each sliding time window, so that the duration of the time window is equal to the product of the sampling time interval and the number of sampling points, and slide the window on the power sequence according to the number of sampling points;

[0096] Based on the total number of sampling points in the power sequence and the number of sampling points in the time window, obtain the number of time windows covering the entire sampling duration, and number the time windows in chronological order.

[0097] For each time window, consecutive sampling points corresponding to the time window number are extracted from the power sequence in chronological order. When the number of remaining sampling points in the last time window is less than the preset number of sampling points, the number of sampling points in the time window is set as the number of remaining sampling points.

[0098] For each time window, the actual duration is calculated based on the number of sampling points and the sampling time interval within the time window. The power values ​​of each sampling point within the time window are recorded to form a power data segment for the time window.

[0099] All power data segments within a time window are combined into a time window data sequence according to their time window numbers.

[0100] Further specific implementation steps include:

[0101] Set the number of sampling points contained in each sliding window to The duration of a single window is ;

[0102] Total number of windows is ;

[0103] Construct the first Power sequence within a window: , , ;in, Number the windows; For the first The number of sampling points contained in each actual window; For the first The actual duration of each window; For the first In the window, the first Power values ​​of each sample point; The location number within the window;

[0104] Forming a window data sequence: ;in, For the first A set of power sequences within a window.

[0105] By dividing the power sequence into equal-length sliding time windows according to a preset number of points, and dynamically adjusting the last window with insufficient points, this scheme solves the problems of statistical bias and time-domain discontinuity caused by uneven window lengths. Strictly defining window numbers and point numbers ensures clear data boundaries within each window, providing an accurate starting point for spectrum analysis and operating condition switching. The generated "time window power data segments" unify the data structure and simplify subsequent algorithm calls. This mechanism balances data smoothness and real-time performance, enabling timely capture of power characteristics even during rapid operating condition switching, avoiding the loss of critical information during sudden changes in operating conditions, as is common in traditional non-overlapping or fixed-window schemes, thus improving the timeliness and accuracy of monitoring.

[0106] Within each time window, a window-level frequency resolution is set based on the window duration to limit the analysis frequency range. Discrete frequency points are generated according to the window-level frequency resolution, and a set of frequency basis functions is constructed. Specifically, this includes:

[0107] For each time window, set the window-level frequency resolution based on the actual duration of the time window, and limit the interval between adjacent analysis frequencies;

[0108] Set the upper limit of the power signal analysis frequency and compare it with the Nyquist limit corresponding to half of the sampling frequency, and select the smaller value as the upper limit of the effective analysis frequency.

[0109] Within the range from zero frequency to the upper limit of the effective analysis frequency, discrete frequency points are divided according to window-level frequency resolution; no frequency points are set when the upper limit of the effective analysis frequency is zero; when the upper limit of the effective analysis frequency is greater than zero, the number of frequency components used in the time window and the value of each frequency are obtained.

[0110] Each frequency value is converted into an angular frequency according to the proportional relationship between frequency and angular velocity, forming an angular frequency index set associated with the time window;

[0111] Based on the angular frequency index set, cosine basis functions and sine basis functions are constructed at continuous time points to obtain a frequency basis function set consisting of a set of mutually orthogonal basis functions.

[0112] Further specific implementation steps include:

[0113] Set the window-level base frequency interval to ;in, For the first The resolution of a window in the frequency domain;

[0114] Set the upper limit for the analysis frequency to And construct an effective upper bound to satisfy the Nyquist constraint, specifically: ;in, To simultaneously consider the Nyquist constraint The effective frequency upper limit obtained thereafter;

[0115] The number of frequency components and frequency sequence segments supported by this window are as follows:

[0116] , , ;

[0117] in, To make the interval According to the Window frequency step When divided equally, the theoretical number of complete frequency steps that can be accommodated; For the first The number of harmonic frequency components actually used within each window; For the first The first window The frequency value of each frequency component;

[0118] The angular frequency sequence for constructing this window is as follows: , ;in, For the first The first window The angular frequency corresponding to each frequency component;

[0119] Construct the time basis function as follows:

[0120] Cosine basis: ,

[0121] Sine base: ;in, For the first The first window The cosine basis functions are continuous-time The basis functions on the projection power signal are used; For the first The first window A sinusoidal basis function.

[0122] By automatically calculating the fundamental frequency interval based on the duration of each window and setting an effective frequency upper limit while satisfying the Nyquist constraint, this scheme overcomes the limitation that fixed frequency resolution cannot adapt to different window lengths. The dynamically divided discrete frequency points fit the window characteristics, balancing frequency resolution and computational efficiency; the constructed set of angular frequency indices and mutually orthogonal basis functions ensure the orthogonality and completeness of the projection calculation. Compared with the spectral leakage and insufficient resolution caused by the sensitivity of traditional Fourier analysis to window length, the adaptive resolution and frequency step size in this scheme can accurately capture the main harmonic components within each time window, and reduce computational complexity through a unified set of frequency basis functions, thus enabling efficient execution on embedded or edge devices.

[0123] The calculation of frequency projection coefficients based on power data segments within a time window and a set of frequency basis functions to obtain frequency amplitude and average power sample values ​​specifically includes:

[0124] For each time window, based on the starting sampling point number in the power sequence and the sampling point number within the time window, calculate the continuous time corresponding to each sampling point and establish the mapping relationship between the sampling point number and the continuous time.

[0125] Within each time window, for each frequency basis function, the power value of each sampling point in the power data segment of the time window is multiplied point by point with the cosine basis function value and sine basis function value at the corresponding continuous time moment, and the product of all sampling points is summed within the time window. Then, normalization is performed according to the number of sampling points in the time window to obtain the cosine projection coefficient and sine projection coefficient corresponding to each frequency component.

[0126] Within each time window, the frequency amplitude of each frequency component is calculated based on the sum of the squares of the cosine projection coefficient and the sine projection coefficient, forming a frequency amplitude sequence covering all frequency components.

[0127] Within each time window, the power values ​​of all sampled points in the power data segment of the time window are arithmetically averaged to obtain the average power sample value of the time window, and the average power sample value is used as the power baseline component.

[0128] Further specific implementation steps include:

[0129] For windows nuclear frequency The sampling time point was constructed as follows ;in, For the global timeline corresponding to the first The first window The continuous time points of each sampling point;

[0130] Calculate the cosine projection coefficients: ;in, For the first In the window corresponding to the first Cosine projection coefficients of each frequency component;

[0131] Calculate the sinusoidal projection coefficients: ;in, For the first In the window corresponding to the first Sine projection coefficients of each frequency component;

[0132] The frequency amplitude function is constructed as follows: ;in, For the first The first window The combined amplitude of each frequency component;

[0133] The baseline components are constructed as follows: ;in, For the first The average value of all power samples within a window.

[0134] This scheme solves the problem of the uninterpretability of spectral amplitude in FFT by independently calculating the cosine and sine projection coefficients for each discrete frequency component, constructing the frequency amplitude in the form of a sum of squares, and extracting the window power baseline by arithmetic mean. This projection calculation clearly corresponds to the energy contribution of each frequency component and removes the DC component and energy drift of the power signal through the baseline component. Unlike traditional spectral analysis, which struggles to quantify the baseline effect, this method can clearly separate the baseline and harmonic components, providing an accurate basis for subsequent residual judgment. Simultaneously, it simplifies the algorithm flow, significantly reducing redundant calculations and memory usage, making it more suitable for real-time deployment.

[0135] The process of reconstructing the power sequence based on the power sample average and finite frequency components, calculating the reconstructed residual sequence, and obtaining spectral residual data specifically includes:

[0136] For each time window, the power baseline component is combined with the cosine projection coefficient, sine projection coefficient and corresponding frequency basis function of each frequency component within the time window, and the reconstructed power value of each sampling point is calculated according to the sampling time order to form the time window power reconstruction sequence.

[0137] Within each time window, the power data segment of the time window is subtracted from the power reconstruction sequence of the time window point by point according to the sampling point to obtain the reconstruction residual sequence.

[0138] Further specific implementation steps include:

[0139] Calculate the first The reconstructed power sequence for each window is as follows:

[0140] ;

[0141] in, To use the baseline component and a finite number of harmonic components to the first The first window The estimated value is obtained by reconstructing the power of each sampling point;

[0142] Constructing and reconstructing residuals: , ;in, For the first The first window The difference between the actual power and the reconstructed power at each sampling point.

[0143] By reconstructing the time-domain power curve by superimposing baseline components with finite frequency components and obtaining the residual sequence through point-by-point difference with the measured signal, this method overcomes the limitation of traditional methods that rely solely on frequency domain indicators and cannot grasp dynamic deviations. The residual sequence intuitively reflects model fitting errors and signal abrupt changes, facilitating the capture of transient events such as equipment startup and shutdown, and sudden changes in airflow. This method eliminates the difficulty in quantifying anomalies due to direct observation of power waveforms and provides highly sensitive data support for further spectral residual analysis; in actual operation, it can promptly identify subtle anomalies and improve early warning capabilities.

[0144] The process involves collecting the supply air volumetric flow rate and inlet / outlet air temperature, calculating heat transfer, and constructing a spectral energy efficiency index using a high-frequency responsive spectrum set. Specifically, this includes:

[0145] Within each time window, the average volumetric flow rate of the supply air, the average inlet air temperature, and the average outlet air temperature are collected, and the average temperature difference of the time window is obtained by the difference between the inlet and outlet temperatures.

[0146] Obtain the air volumetric specific heat capacity parameter, and calculate the total heat based on the air volumetric specific heat capacity, average air volumetric flow rate, average temperature difference and actual duration of the time window;

[0147] Obtain the shortest response time parameter of the air conditioning system. Based on the relationship between the shortest response time and the window-level frequency resolution, calculate the frequency component index corresponding to the lowest frequency that the system can respond to for each time window, and divide the frequency component corresponding to the lowest frequency and the frequency components above it into a high-frequency responsive spectrum set.

[0148] Within each time window, when there are no frequency components, the energy consumption of the set of high-frequency responsive spectra is set to zero, and the spectral energy efficiency index is also set to zero; when there are frequency components, the amplitude of the set of high-frequency responsive spectra is combined with the duration of the time window to calculate the high spectral energy consumption.

[0149] Within each time window, when the high-frequency energy consumption is greater than zero, the ratio of total heat to high-frequency energy consumption is used as the spectrum energy efficiency index; when the high-frequency energy consumption is equal to zero, the spectrum energy efficiency index is set to zero.

[0150] Further specific implementation steps include:

[0151] Collect the average airflow within the current window. Inlet air temperature With air outlet temperature The temperature difference is calculated as follows: ;in, For the first The average volumetric airflow rate of the air supplied by the air conditioner within each window; For the first The average air temperature at the air conditioning inlet of each window; For the first The average air temperature at the air conditioner outlet in each window; For the first The average temperature difference within each window;

[0152] Obtain the specific heat capacity of air by volume, denoted as . ;

[0153] The heat output per unit window is: ;in, For the first The total heat transfer corresponding to the air cooling effect of the air conditioner within each window;

[0154] The shortest response time of the air conditioning system is denoted as: ;

[0155] The frequency exclusion index is calculated based on the shortest response time and window frequency resolution of the air conditioning system, and zero-harmonic cases are handled, specifically as follows:

[0156] like Then let , ;in, For the first The energy consumption corresponding to the high-frequency spectrum components within each window; For the first Spectral energy efficiency indicators for each window;

[0157] like Then let ,

[0158] get ;in, In the first The harmonic index is determined within each window to be the lowest frequency that the system can respond to.

[0159] like Then let ;like Then let .

[0160] By synchronously collecting supply air volumetric flow rate and inlet / outlet air temperature difference, and combining this with air specific heat capacity, the total cooling heat within the window is calculated. This is then compared with the ratio of high-frequency responsive spectrum energy consumption to construct a spectrum energy efficiency index. This solution addresses the disconnect between power and heat assessment. This index reflects the energy efficiency impact of different harmonic energy consumption on cooling capacity, offering more granularity than the traditional COP which only focuses on the overall ratio. Especially in multi-mode, fast-start scenarios, this method can reveal the specific frequency band contribution to energy efficiency degradation under high-frequency operating conditions, providing direct evidence for precise energy-saving control and inverter parameter optimization, thus enhancing the guiding value of monitoring.

[0161] The process of constructing a spectral feature vector based on the frequency amplitude distribution of each time window, calculating the spectral structure difference between adjacent time windows, and dividing the operating condition segments according to the spectral difference threshold specifically includes:

[0162] Within each time window, the frequency amplitudes of all frequency components are combined in frequency order to form a spectral feature vector;

[0163] For any two different time windows, the larger value is selected as the common bandwidth based on the window-level frequency resolution of the two, and adjacent frequency bands are divided according to the common bandwidth within the interval from zero frequency to the upper limit of the effective analysis frequency.

[0164] For each time window, the frequency amplitudes belonging to each frequency band are summed to obtain the band-level amplitude of each frequency band; when there are no frequency components in a certain frequency band, the band-level amplitude of that frequency band is set to zero.

[0165] For any two time windows, calculate the normalization ratio of the difference in band amplitude to the sum of the two band amplitudes and the minimum resolvable power value of the high-frequency power acquisition module in all frequency bands, and sum them in all frequency bands to obtain the spectral distance value representing the degree of difference in the spectral structure of the two time windows.

[0166] In chronological order, the spectral distance values ​​of adjacent time windows are statistically analyzed. When there are at least two time windows, the arithmetic mean of the spectral distance values ​​of all adjacent time windows is calculated and the arithmetic mean is used as the spectral difference threshold. When there is only one time window, the spectral difference threshold is set to zero.

[0167] For adjacent time windows, when the spectral distance value is greater than the spectral difference threshold, the next time window is taken as the starting time window of the new operating condition segment; when the spectral distance value is less than or equal to the spectral difference threshold, the next time window is divided into the same operating condition segment as the previous time window.

[0168] Further specific implementation steps include:

[0169] The spectral eigenvectors are constructed as follows: ;in, For the first The spectral feature vector of each window;

[0170] Calculate the distance between the aligned frequency band and the spectral structure of any two windows, specifically as follows:

[0171] S701, First construct the common bandwidth of the two windows: Based on this, the interval The average is: There are several adjacent frequency bands; among them... Indices for the two different windows being compared; To be in two windows A uniform bandwidth is used when performing spectrum alignment; Based on common bandwidth The number of frequency bands divided into intervals;

[0172] S702. For each window, converge its spectral amplitude by band: , ;in, Frequency band number; In the first In the window, the first The sum of spectral amplitudes across all frequency bands;

[0173] If the summation index set is empty, then let ;

[0174] S703. Construct a distance function between two windows using the amplitude of the band, specifically: ;in, For window With window A measure of the difference in spectral structure between them; This represents the minimum resolvable power of the power acquisition module.

[0175] First calculate ;in, This is the spectral difference threshold used to determine whether adjacent windows belong to the same operating condition;

[0176] like Then determine the window The starting point is a new operating condition section;

[0177] when , window Return to Window The same operating conditions.

[0178] By constructing spectral feature vectors and aligning adjacent windows with a common bandwidth for band-level convergence, and then calculating spectral structure differences by accumulating normalized differences and adaptively setting operating condition switching thresholds, this solution solves the problem of objectively determining operating condition switching points. Without the need for manual threshold presets, the system can automatically calculate the difference threshold based on historical adjacent differences, achieving accurate division of operating condition segments. Compared with traditional empirical methods or single-indicator judgments, this method is more robust and adaptable to noise and environmental interference; simultaneously, band-level statistics and spectral alignment processing improve anti-interference capabilities, ensuring that segment division under complex operating conditions still accurately reflects the system's operating status.

[0179] The process of aggregating spectral energy efficiency indicators within each operating condition segment, obtaining average energy efficiency indicators and energy efficiency deviation indicators, and outputting time windows with abnormally low energy efficiency values ​​specifically includes:

[0180] All time windows are divided into multiple working condition segments according to the working condition identification results, and a set of windows containing the time window number of each working condition segment is established.

[0181] Within each operating condition segment, the spectral energy efficiency index of all time windows in the window set is arithmetically averaged to obtain the average energy efficiency index of the operating condition segment.

[0182] Within each operating condition segment, the arithmetic mean is calculated on the absolute difference between the spectral energy efficiency index and the average energy efficiency index for all time windows in the window set to obtain the energy efficiency deviation index within the operating condition segment.

[0183] Within each operating condition segment, when the spectral energy efficiency index of a certain time window is less than the average energy efficiency index of the operating condition segment minus the energy efficiency deviation index, the time window is marked as an abnormally low energy efficiency time window; when the spectral energy efficiency index of a certain time window is not less than the average energy efficiency index of the operating condition segment minus the energy efficiency deviation index, the time window is marked as a normal energy efficiency time window, and the operating condition segment remains unchanged.

[0184] Further specific implementation steps include:

[0185] Group all windows into Each working condition section, each working condition Contains a collection of windows The average energy efficiency under the operating conditions is: ;in, This represents the total number of working condition sections identified during the entire observation period; Index for operating condition sections; To be classified as the first A set of window indexes for each working condition; For set The number of windows contained; For the first Average energy efficiency index for each operating condition section;

[0186] The deviation is: ;in, For the first The average absolute deviation of energy efficiency within each operating condition section;

[0187] like Then the window This is identified as an abnormally low energy efficiency window.

[0188] when Then the window It is determined to be within the normal energy efficiency window, and its corresponding operating condition classification remains unchanged.

[0189] By calculating the arithmetic mean and mean absolute deviation of spectral energy efficiency indicators within each operating condition segment, and using the difference between the two as the anomaly detection threshold, this scheme solves the problem that a globally fixed threshold is difficult to adapt to differences in different operating conditions. This method takes into account the normal fluctuation range within the operating conditions, avoiding the risk of false alarms or missed alarms; when the window energy efficiency is below the threshold, it automatically marks an abnormally low value window, providing an objective quantitative basis for maintenance decisions. Compared with traditional methods that rely solely on empirical thresholds or single-moment comparisons, this strategy considers the overall level and fluctuation characteristics within the operating conditions, thereby achieving more robust anomaly identification and improving operation and maintenance efficiency and energy-saving effects.

[0190] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0191] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for monitoring and evaluating the energy efficiency of air conditioners based on real-time data acquisition, characterized in that, include: The instantaneous active power is collected at the power input terminal of the air conditioner, and the sampling frequency and sampling duration are set to form a power sequence according to the time interval. The power sequence is divided into equal-length sliding time windows according to a preset number of sampling points. The number of sampling points and duration of each time window are obtained to generate power data segments for the time windows. Within each time window, the window-level frequency resolution is set according to the window duration to limit the analysis frequency range. Discrete frequency points are generated according to the window-level frequency resolution, and a set of frequency basis functions is constructed, specifically including: For each time window, set the window-level frequency resolution based on the actual duration of the time window, and limit the interval between adjacent analysis frequencies; Set the upper limit of the power signal analysis frequency and compare it with the Nyquist limit corresponding to half of the sampling frequency, and select the smaller value as the upper limit of the effective analysis frequency. Within the range from zero frequency to the upper limit of the effective analysis frequency, discrete frequency points are divided according to window-level frequency resolution; no frequency points are set when the upper limit of the effective analysis frequency is zero; when the upper limit of the effective analysis frequency is greater than zero, the number of frequency components used in the time window and the value of each frequency are obtained. Each frequency value is converted into an angular frequency according to the proportional relationship between frequency and angular velocity, forming an angular frequency index set associated with the time window; Based on the angular frequency index set, cosine basis functions and sine basis functions are constructed at continuous time points to obtain a frequency basis function set consisting of a set of mutually orthogonal basis functions; The calculation of frequency projection coefficients based on power data segments within a time window and a set of frequency basis functions to obtain frequency amplitude and average power sample values ​​specifically includes: For each time window, based on the starting sampling point number in the power sequence and the sampling point number within the time window, calculate the continuous time corresponding to each sampling point and establish the mapping relationship between the sampling point number and the continuous time. Within each time window, for each frequency basis function, the power value of each sampling point in the power data segment of the time window is multiplied point by point with the cosine basis function value and sine basis function value at the corresponding continuous time moment, and the product of all sampling points is summed within the time window. Then, normalization is performed according to the number of sampling points in the time window to obtain the cosine projection coefficient and sine projection coefficient corresponding to each frequency component. Within each time window, the frequency amplitude of each frequency component is calculated based on the sum of the squares of the cosine projection coefficient and the sine projection coefficient, forming a frequency amplitude sequence covering all frequency components. Within each time window, the power values ​​of all sampling points in the power data segment of the time window are arithmetically averaged to obtain the average power sample value of the time window, and the average power sample value is used as the power baseline component. The projection coefficients of each frequency are calculated based on the power data segments within the time window and the set of frequency basis functions to obtain the frequency amplitude and the average value of the power samples. The power sequence is reconstructed based on the power sample average value and finite frequency components. The reconstructed residual sequence is calculated and the spectral residual data is obtained. Collect supply air volume flow rate and inlet and outlet air temperature, calculate heat transfer, and construct a spectrum energy efficiency index by combining high-frequency responsive spectrum set; A spectral feature vector is constructed based on the frequency amplitude distribution of each time window, the spectral structure difference between adjacent time windows is calculated, and the operating condition segments are divided according to the spectral difference threshold. Aggregate the spectral energy efficiency index within each operating condition segment, obtain the average energy efficiency index and the energy efficiency deviation index, and output the time window of abnormally low energy efficiency values.

2. The method for monitoring and evaluating air conditioning energy efficiency based on real-time data acquisition according to claim 1, characterized in that, The air conditioner power input terminal collects instantaneous active power, sets the sampling frequency and sampling duration, and forms a power sequence according to time intervals, specifically including: A high-frequency power acquisition module is configured at the power input terminal of the workshop air conditioner so that the high-frequency power acquisition module continuously outputs sampled values ​​representing the instantaneous active power of the air conditioner power supply during air conditioner operation. Set the sampling frequency, and set the sampling time interval between adjacent power sample values ​​according to the reciprocal of the sampling frequency, while keeping the sampling time interval constant; Set the total sampling duration parameter, and obtain the number of sampling points based on the ratio of the total sampling duration to the sampling time interval. When the calculated number of sampling points is less than one sampling point, set the number of sampling points to one sampling point to ensure that there is at least one discrete sampling point. Starting from the sampling start time, the continuous time of each sampling time is obtained in sequence according to the sampling time interval, and the sampling point number is assigned in sequence to establish a one-to-one correspondence between the sampling point number and the continuous time moment; The instantaneous active power sample values ​​at each sampling time are recorded as discrete power samples. A power sequence is constructed by arranging all discrete power samples according to the sampling point number, and the discrete power samples are limited to non-negative real numbers.

3. The method for monitoring and evaluating air conditioning energy efficiency based on real-time data acquisition according to claim 2, characterized in that, The step of dividing the power sequence into equal-length sliding time windows according to a preset number of sampling points, obtaining the number of sampling points and duration of each time window, and generating a power data segment for the time window specifically includes: Set the number of sampling points included in each sliding time window, so that the duration of the time window is equal to the product of the sampling time interval and the number of sampling points, and slide the window on the power sequence according to the number of sampling points; Based on the total number of sampling points in the power sequence and the number of sampling points in the time window, obtain the number of time windows covering the entire sampling duration, and number the time windows in chronological order. For each time window, consecutive sampling points corresponding to the time window number are extracted from the power sequence in chronological order. When the number of remaining sampling points in the last time window is less than the preset number of sampling points, the number of sampling points in the time window is set as the number of remaining sampling points. For each time window, the actual duration is calculated based on the number of sampling points and the sampling time interval within the time window. The power values ​​of each sampling point within the time window are recorded to form a power data segment for the time window. All power data segments within a time window are combined into a time window data sequence according to their time window numbers.

4. The method for monitoring and evaluating air conditioning energy efficiency based on real-time data acquisition according to claim 1, characterized in that, The process of reconstructing the power sequence based on the power sample average and finite frequency components, calculating the reconstructed residual sequence, and obtaining spectral residual data specifically includes: For each time window, the power baseline component is combined with the cosine projection coefficient, sine projection coefficient and corresponding frequency basis function of each frequency component within the time window, and the reconstructed power value of each sampling point is calculated according to the sampling time order to form the time window power reconstruction sequence. Within each time window, the power data segment of the time window is subtracted from the power reconstruction sequence of the time window point by point according to the sampling point to obtain the reconstruction residual sequence.

5. The method for monitoring and evaluating air conditioning energy efficiency based on real-time data acquisition according to claim 4, characterized in that, The process involves collecting the supply air volumetric flow rate and inlet / outlet air temperature, calculating heat transfer, and constructing a spectral energy efficiency index using a high-frequency responsive spectrum set. Specifically, this includes: Within each time window, the average volumetric flow rate of the supply air, the average inlet air temperature, and the average outlet air temperature are collected, and the average temperature difference of the time window is obtained by the difference between the inlet and outlet temperatures. Obtain the air volumetric specific heat capacity parameter, and calculate the total heat based on the air volumetric specific heat capacity, average air volumetric flow rate, average temperature difference and actual duration of the time window; Obtain the shortest response time parameter of the air conditioning system. Based on the relationship between the shortest response time and the window-level frequency resolution, calculate the frequency component index corresponding to the lowest frequency that the system can respond to for each time window, and divide the frequency component corresponding to the lowest frequency and the frequency components above it into a high-frequency responsive spectrum set. Within each time window, when there are no frequency components, the energy consumption of the set of high-frequency responsive spectra is set to zero, and the spectral energy efficiency index is also set to zero; when there are frequency components, the amplitude of the set of high-frequency responsive spectra is combined with the duration of the time window to calculate the high spectral energy consumption. Within each time window, when the high-frequency energy consumption is greater than zero, the ratio of total heat to high-frequency energy consumption is used as the spectrum energy efficiency index; when the high-frequency energy consumption is equal to zero, the spectrum energy efficiency index is set to zero.

6. The method for monitoring and evaluating air conditioning energy efficiency based on real-time data acquisition according to claim 5, characterized in that, The process of constructing a spectral feature vector based on the frequency amplitude distribution of each time window, calculating the spectral structure difference between adjacent time windows, and dividing the operating condition segments according to the spectral difference threshold specifically includes: Within each time window, the frequency amplitudes of all frequency components are combined in frequency order to form a spectral feature vector; For any two different time windows, the larger value is selected as the common bandwidth based on the window-level frequency resolution of the two, and adjacent frequency bands are divided according to the common bandwidth within the interval from zero frequency to the upper limit of the effective analysis frequency. For each time window, the frequency amplitudes belonging to each frequency band are summed to obtain the band-level amplitude of each frequency band; when there are no frequency components in a certain frequency band, the band-level amplitude of that frequency band is set to zero. For any two time windows, calculate the normalization ratio of the difference in band amplitude to the sum of the two band amplitudes and the minimum resolvable power value of the high-frequency power acquisition module in all frequency bands, and sum them in all frequency bands to obtain the spectral distance value representing the degree of difference in the spectral structure of the two time windows. In chronological order, the spectral distance values ​​of adjacent time windows are statistically analyzed. When there are at least two time windows, the arithmetic mean of the spectral distance values ​​of all adjacent time windows is calculated and the arithmetic mean is used as the spectral difference threshold. When there is only one time window, the spectral difference threshold is set to zero. For adjacent time windows, when the spectral distance value is greater than the spectral difference threshold, the next time window is taken as the starting time window of the new operating condition segment; when the spectral distance value is less than or equal to the spectral difference threshold, the next time window is divided into the same operating condition segment as the previous time window.

7. The method for monitoring and evaluating air conditioning energy efficiency based on real-time data acquisition according to claim 6, characterized in that, The process of aggregating spectral energy efficiency indicators within each operating condition segment, obtaining average energy efficiency indicators and energy efficiency deviation indicators, and outputting time windows with abnormally low energy efficiency values ​​specifically includes: All time windows are divided into multiple working condition segments according to the working condition identification results, and a set of windows containing the time window number of each working condition segment is established. Within each operating condition segment, the spectral energy efficiency index of all time windows in the window set is arithmetically averaged to obtain the average energy efficiency index of the operating condition segment. Within each operating condition segment, the arithmetic mean is calculated on the absolute difference between the spectral energy efficiency index and the average energy efficiency index for all time windows in the window set to obtain the energy efficiency deviation index within the operating condition segment. Within each operating condition segment, when the spectral energy efficiency index of a certain time window is less than the average energy efficiency index of the operating condition segment minus the energy efficiency deviation index, the time window is marked as an abnormally low energy efficiency time window; when the spectral energy efficiency index of a certain time window is not less than the average energy efficiency index of the operating condition segment minus the energy efficiency deviation index, the time window is marked as a normal energy efficiency time window, and the operating condition segment remains unchanged.

Images