Self-adaptive temperature control method for heating system based on data acquisition

Abstract

The invention relates to the technical field of building thermal response characteristic analysis and adaptive temperature, and discloses a heating system adaptive temperature control method based on data acquisition. Seasonal decomposition processing is carried out on multi-period collected data to separate a long-term degradation trend, track fitting and time extrapolation are adopted to predict and control mismatch moments, a parameter sensitivity mapping matrix is established to realize parameter adaptive adjustment, and the temperature control problem of a heating system in a building thermal characteristic slow degradation scene is solved.

Description

Technical Field

[0001] This invention relates to the field of building thermal response characteristic analysis and adaptive temperature technology, and more specifically, to an adaptive temperature control method for heating systems based on data acquisition. Background Technology

[0002] During the long-term operation of a heating system, the building envelope will slowly degrade in thermal performance due to factors such as material aging, moisture absorption of the insulation layer, and decreased sealing of doors and windows. After several years, this will lead to a significant mismatch between the original temperature control parameters and the actual thermal characteristics of the building.

[0003] Traditional heating temperature control methods operate with fixed parameters, and the main technical problem is that short-term fluctuations in the collected data (including weather changes, differences in usage habits, and seasonal temperature cycles) can mask the long-term degradation trend of building thermal response characteristics. Traditional methods cannot distinguish between normal seasonal fluctuations and thermal response characteristic degradation signals from temperature data collected over multiple periods. They can only passively adjust control parameters after the temperature control effect has deteriorated significantly, and cannot achieve active adaptive temperature control based on data acquisition. Summary of the Invention

[0004] This invention provides an adaptive temperature control method for heating systems based on data acquisition, which solves the technical problem of temperature control parameter mismatch caused by the slow degradation of building thermal characteristics in related technologies.

[0005] This invention provides an adaptive temperature control method for a heating system based on data acquisition, comprising: Indoor temperature response data of buildings under standard thermal excitation test is collected, and feature extraction and dimensionality reduction are performed on the temperature response data to generate a time-series thermal response fingerprint vector database. The fingerprint vectors in the time-series thermal response fingerprint vector database are subjected to time-series decomposition processing to separate periodic fluctuation components and trend components. After removing the periodic fluctuation components, a deseasonalized fingerprint trend sequence is generated. Trajectory fitting is performed on the deseasoned fingerprint trend sequence to predict the evolution function of each feature dimension over time. Time extrapolation is performed based on a preset temperature control mismatch threshold to calculate the time when the fingerprint feature is expected to exceed the mismatch threshold, and the expected mismatch time point and future fingerprint prediction value are generated. The difference between the predicted future fingerprint value and the initial baseline fingerprint vector is calculated. The difference is then input into the parameter sensitivity mapping matrix to calculate the temperature control parameter correction increment, and an adaptive parameter adjustment scheme is generated and output. The parameter sensitivity mapping matrix is ​​used to characterize the sensitivity coefficient of each temperature control parameter to changes in each fingerprint feature dimension.

[0006] Furthermore, the time-series decomposition process employs the STL decomposition algorithm, which uses iterative local weighted regression to separate the components of each dimension of the fingerprint vector into seasonal, trend, and residual terms.

[0007] Furthermore, before performing trajectory fitting on the deseasoned fingerprint trend sequence, the method further includes: A sliding window statistical analysis was performed on the deseasoned fingerprint trend sequence to calculate the mean and variance of each feature dimension within each window; The changes in the mean and variance of each dimension between adjacent windows are statistically analyzed, the slope of the change trend is calculated, and a fingerprint degradation trend feature set is generated.

[0008] Furthermore, the trajectory fitting employs a multinomial regression algorithm to fit the time evolution of each feature dimension separately, thereby obtaining a multinomial function for each dimension as a function of time.

[0009] Furthermore, the parameter sensitivity mapping matrix is ​​obtained by: conducting disturbance tests on each temperature control parameter during the initial commissioning phase of the heating system, recording the corresponding thermal response fingerprint changes, and using the least squares method to fit and establish a linear mapping relationship between fingerprint changes and parameter changes.

[0010] Furthermore, before generating and outputting the adaptive parameter tuning scheme, the following steps are also included: Calculate the time interval between the current moment and the predicted mismatch time point; The time intervals are compared with a preset adaptive adjustment lead amount; If the time interval is less than or equal to the adaptive adjustment lead amount, the temperature control parameter adaptive adjustment process is triggered.

[0011] Furthermore, the output adaptive parameter adjustment scheme includes: Calculate the deviation between the target parameter and the current temperature control operating parameter in the adaptive parameter adjustment scheme; The deviation is linearly decomposed over time into multiple fine-tuning increments; The parameter fine-tuning instructions are output sequentially at preset time intervals to smoothly transition the current parameter to the target parameter.

[0012] Furthermore, it also includes: Continuously collect indoor temperature response data and extract the current thermal response fingerprint; Calculate the deviation between the current actual fingerprint and the predicted fingerprint of the fitted trajectory at the corresponding time. If the deviation exceeds the preset trajectory deviation tolerance, the trajectory fitting function is updated using the current actual fingerprint data, the expected mismatch time point and future fingerprint prediction value are recalculated, and the adaptive parameter adjustment scheme is updated.

[0013] Furthermore, the dimensionality reduction process employs principal component analysis, retaining principal components whose cumulative contribution rate exceeds a preset threshold as the dimensional components of the fingerprint vector.

[0014] This invention provides an adaptive temperature control system for a heating system based on data acquisition, comprising: The fingerprint generation module is used to collect indoor temperature response data of buildings under standard thermal excitation tests, perform feature extraction and dimensionality reduction on the temperature response data, and generate a time-series thermal response fingerprint vector database. The seasonal decomposition module is used to perform time-series decomposition processing on the fingerprint vectors in the time-series thermal response fingerprint vector database, separate the periodic fluctuation component and the trend component, and generate a deseasonalized fingerprint trend sequence. The trajectory prediction module is used to perform trajectory fitting on the deseasoned fingerprint trend sequence, extrapolate the time based on the temperature-controlled mismatch threshold, and generate the expected mismatch time point and future fingerprint prediction value. The parameter adjustment module is used to calculate the difference between the predicted future fingerprint value and the initial baseline fingerprint vector, input the difference into the parameter sensitivity mapping matrix to calculate the temperature control parameter correction increment, and generate and output an adaptive parameter adjustment scheme.

[0015] The beneficial effects of this invention are as follows: This invention establishes a time-series thermal response fingerprint vector database and performs seasonal decomposition processing, abstracting building thermal response characteristics into quantifiable and traceable fingerprint vectors. It separates periodic fluctuation components and long-term trend components from multi-period data acquisition, overcoming the problem of short-term data fluctuations masking long-term degradation trends. By performing trajectory fitting and time extrapolation prediction in the fingerprint feature space, it can predict when fingerprint features exceed the temperature control mismatch threshold, overcoming the lag problem of traditional methods that can only passively respond after the control effect deteriorates. A quantitative relationship between fingerprint changes and temperature control parameter correction is established through a parameter sensitivity mapping matrix, enabling adaptive calculation of parameter correction based on the predicted thermal characteristic degradation. Therefore, this invention solves the technical problem of heating systems being unable to achieve active adaptive temperature control based on data acquisition in scenarios of slow building thermal characteristic degradation, achieving the technical effect of actively predicting parameter mismatch moments and adaptively adjusting control parameters. Attached Figure Description

[0016] Figure 1 This is a flowchart of the adaptive temperature control method for a heating system based on data acquisition, as described in this invention. Figure 2 This is a bar chart comparing the initial characteristic parameters of multiple measurement points according to the present invention; Figure 3 This is a three-year fingerprint trend sequence evolution line chart of the present invention; Figure 4 This is a histogram of the fingerprint feature dimension degradation rate according to the present invention; Figure 5 This is a line graph showing the trajectory fitting and mismatch prediction of the present invention; Figure 6 This is the parameter sensitivity mapping matrix heatmap of the present invention; Figure 7 This is a bar chart of the parameter adaptive adjustment scheme of the present invention; Figure 8 This is a line graph illustrating the closed-loop adaptive control monitoring of the present invention. Detailed Implementation

[0017] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, some features described in the examples may be combined in other examples.

[0018] At least one embodiment of the present invention discloses an adaptive temperature control method for a heating system based on data acquisition, such as... Figure 1 As shown, it includes the following steps: Step 1: Collect building thermal response data and extract features to generate a time-series thermal response fingerprint vector database.

[0019] Indoor temperature response data of the building is collected by multi-point temperature sensors in the heating system under standard thermal excitation tests conducted at fixed intervals. The standard thermal excitation test refers to thermally exciting the building according to a preset heating power step change pattern, with each sensor synchronously recording the temperature response curve. Feature extraction is performed on the collected response data, including the temperature rise time constant, steady-state temperature deviation, consistency index of temperature response at each sensor point, and thermal inertia coefficient.

[0020] Furthermore, the temperature rise time constant is obtained by curve fitting the step response curve of a single measuring point, using a first-order system model. Fit the response curve, where The equilibrium temperature after the step response stabilizes (usually a measurement taken at least half an hour after the thermal excitation ends). The initial temperature at the moment the thermal excitation begins. It is a time constant. The base of the natural logarithm (exponential function) is used to minimize the error between the fitted curve and the actual response data using the least squares method. Minimum is used to determine, where the summation symbol is used to determine. The sampling time index is indicated; the steady-state temperature deviation is the difference between the measured temperature and the set temperature at each measuring point after the thermal excitation has stabilized; the consistency index of the temperature response at each measuring point is defined as the coefficient of variation (standard deviation divided by mean) of the temperature rise time constant at each measuring point, and the consistency index reflects the uniformity of the thermal response characteristics among multiple measuring points; the thermal inertia coefficient is the ratio of the thermal excitation power to the steady-state temperature rise (i.e., ... The ratio of ).

[0021] Data preprocessing is performed on the extracted multidimensional features. Since each feature has different physical dimensions (temperature rise time constant in units of time, steady-state temperature deviation in units of temperature, consistency index as a dimensionless relative value, thermal inertia coefficient in units of time, etc.) and numerical ranges, directly performing principal component analysis can lead to features with larger dimensions excessively influencing the dimensionality reduction results. Therefore, it is necessary to standardize each feature using the Z-score standardization method, normalizing each feature to a standard normal distribution with a mean of 0 and a standard deviation of 1, thus eliminating the impact of differences in dimensions and numerical ranges on the dimensionality reduction calculation.

[0022] The standardized multidimensional features are subjected to dimensionality reduction processing, and the features at each time point are compressed into fingerprint vectors of fixed dimensions. These vectors are then stored in chronological order to generate a time-series thermal response fingerprint vector database.

[0023] It should be noted that the above dimensionality reduction process can be performed using principal component analysis, retaining principal components whose cumulative contribution rate exceeds a preset threshold as the components of each dimension of the fingerprint vector.

[0024] Furthermore, the preset threshold is determined based on the building type and control precision requirements, and is usually set to a value between 85% and 95% to ensure that the retained principal components can cover the main information of the original features.

[0025] Taking the heating system of an office building as an example, a standard thermal excitation test was conducted during the initial commissioning phase, with the heating power increased from 0W to 5000W. Multi-point temperature sensors were used to simultaneously collect indoor temperature response data at three measuring points: the building center, the east side, and the west side. After feature extraction and standardization, the characteristic parameters obtained in the first year are shown in the table below: Table 1 Original Acquisition Feature Parameters: The fingerprint vector after Z-score normalization is The five dimensions correspond to the standardized time constant, steady-state temperature deviation, consistency index, thermal inertia coefficient, and comprehensive characteristic index, respectively. This vector is stored as the initial baseline fingerprint vector in the thermal response fingerprint database.

[0026] Figure 2The comparison of initial characteristic parameters of the three measuring points (center, east side, and west side) in step 1 is shown, including time constant, steady-state temperature deviation, and thermal inertia coefficient.

[0027] Step 2: Perform seasonal decomposition on the time-series thermal response fingerprint vector database to separate the periodic fluctuation component and the trend component, and generate a deseasonalized fingerprint trend sequence.

[0028] A temporal decomposition process is performed on the fingerprint vectors at each time point in the temporal thermal response fingerprint vector database. The temporal decomposition algorithm is used to decompose each dimension of the fingerprint vector, separating them into periodic fluctuation components, trend components, and residual components. The periodic fluctuation components reflect the regular fingerprint fluctuations caused by the annual heating season alternation and climate cycle changes, while the trend components reflect the long-term, slow changes in building thermal response characteristics. The periodic fluctuation components are removed, and the trend components of each dimension are retained and recombine to form a fingerprint vector, generating a deseasonalized fingerprint trend sequence.

[0029] It should be noted that the above time series decomposition algorithm is the STL decomposition algorithm. The STL decomposition algorithm separates the seasonal term, trend term and residual term of the time series through iterative local weighted regression, and is suitable for long-term series data with obvious periodic characteristics.

[0030] Furthermore, the seasonal cycle parameter in the STL decomposition algorithm is set to the number of data points for a heating season, which is typically 150 to 180 days. If the data collection interval is one week, the seasonal cycle parameter is set to 21 to 26 data points.

[0031] STL decomposition was performed on the three-year fingerprint vector sequence of the office building. Taking the time constant dimension as an example, the original fingerprint time series contained obvious annual periodic fluctuations (increasing when heating starts in winter and decreasing in spring and summer) and long-term trend components (the time constant increases year by year due to slow degradation of thermal performance). The STL decomposition algorithm separated each dimension of the fingerprint vector into a periodic term, a trend term, and a remainder term, where the trend term reflects the long-term changes in the building's thermal response characteristics. After removing the periodic component, the deseasonalized fingerprint trend sequence more clearly showed the continuous degradation characteristics of thermal properties.

[0032] The trend components of each dimension of the standardized fingerprint vector change as shown in the table below: Table 2 Deseasonalized fingerprint trend sequence (partial data): These trend components show that the fingerprint slowly drifts along a specific direction in the feature space, reflecting the thermal performance degradation of the building envelope due to aging.

[0033] Figure 3 It shows the trend evolution of five fingerprint feature dimensions over three years.

[0034] Step 3: Perform sliding window statistical analysis on the deseasonalized fingerprint trend sequence, calculate the slope of the change trend of each fingerprint feature dimension, and generate a fingerprint degradation trend feature set.

[0035] Set a sliding window width and perform sliding window statistical analysis on the deseasoned fingerprint trend sequence. For the fingerprint vector sequence within each window, calculate the mean of each feature dimension. and variance ,in Indicates the feature dimension number. Calculates the change in the mean of each dimension between adjacent windows and calculates the slope of the mean change trend. Slope of the trend of variance The slope reflects the degradation rate and stability of fingerprint features. Combining the mean slope and variance slope of each dimension generates a fingerprint degradation trend feature set. ,in The dimension of the fingerprint vector, and the index in the set. This represents the number of each dimension in the fingerprint feature space.

[0036] Furthermore, the width of the sliding window is determined based on the data acquisition frequency and the heating season cycle. The window width should cover the number of data points for at least one complete heating season to ensure that the statistics within the window can effectively reflect the heat response characteristics of that period and avoid the statistics being affected by short-term fluctuations due to an excessively small window.

[0037] Furthermore, the method for calculating the slope of the mean change trend is as follows: Let the means of two adjacent windows be respectively... and ,in and Corresponding to the first The and the first The fingerprint vector within the nth window Dimensional average, Indicates the window sequence number; the corresponding time points are respectively and ,in and If the midpoint of two adjacent windows is the slope of the mean change trend... Similarly, the slope of the variance trend .

[0038] The office building underwent sliding window analysis on deseasonalized fingerprint trend sequences, with a window width of 24 data points (corresponding to an approximately six-month collection period) and a window step size of 4 data points. The changes in each dimension between adjacent windows were statistically analyzed, and the slope of the degradation trend was calculated. The degradation trend feature set was obtained by calculating the features for each of the five fingerprint feature dimensions, reflecting the direction and rate of fingerprint drift throughout the feature space.

[0039] The calculation results of the slope of the mean change trend between adjacent windows are shown in the table below: Table 3. Fingerprint degradation trend feature set (average slope for each year): These slopes indicate that the building's thermal properties degrade at a relatively stable rate year by year, with the time constant increasing the fastest (0.090 per year), reflecting the gradual deterioration of the building's thermal insulation performance.

[0040] Figure 4 Showing the slope of the mean change trend (year) of the five fingerprint feature dimensions. ) and the slope of the variance change.

[0041] Step 4: Fit the fingerprint degradation trend feature set to a trajectory in the fingerprint feature space, estimate the fingerprint drift direction and rate, calculate the expected time when the fingerprint features exceed the temperature control mismatch threshold, and generate the expected mismatch time point and future fingerprint prediction value.

[0042] Using deseasoned fingerprint trend sequences as sample points, trajectory fitting is performed in the fingerprint feature space. A multinomial regression algorithm is then used to fit the temporal evolution of each feature dimension, obtaining the multinomial function of each dimension changing over time. ,in This represents the time variable. The drift direction vector and drift rate of the fingerprint in the feature space are calculated based on the fitting function.

[0043] Furthermore, at the current moment The drift direction vector is calculated as follows: ,in polynomial function At any moment The first derivative of a vector, and the indices of its elements. The fingerprint feature dimension number is represented by the drift rate, which is the Euclidean norm of the drift direction vector. The summation symbol in Traverse from 1 to ( (where is the dimension of the fingerprint vector).

[0044] Furthermore, the polynomial regression algorithm employs the least squares method for fitting, that is, for each feature dimension... Solve for the sum of squared residuals The smallest polynomial coefficients, where The summation symbol represents the total number of data points in the deseasoned fingerprint trend sequence. Traverse from 1 to Indicates the data point index. For the first The first data point The fingerprint feature value; the order of the polynomial is determined based on the number of sample data points and the fitting residual. The order that minimizes the fitting error of the validation set is selected by the K-fold cross-validation method (usually K=5). Usually, a second-order or third-order polynomial is selected to balance the fitting accuracy and the risk of overfitting.

[0045] Preset temperature control mismatch threshold vector ,in For the first The maximum allowable deviation component for each fingerprint feature dimension, and its index in the set. This represents the fingerprint feature dimension number. The fitted function is extrapolated over time to calculate the moment when each feature dimension first exceeds its corresponding threshold component, i.e., to solve for the condition. minimum moment ,in The first in the initial fingerprint eigenvalues ​​of dimension; Take the minimum value of each dimension exceeding the time limit as the expected mismatch time point: index in the set This indicates the corresponding fingerprint feature dimension number. Substitute the values ​​into the fitting function to calculate the predicted fingerprint values ​​at the expected mismatch time. Indices in a set This indicates the fingerprint feature dimension number.

[0046] Furthermore, the extrapolation interval of the time extrapolation is limited by the reliability of the fitting function. The extrapolation time should not exceed 50% of the historical data time span to ensure that the extrapolation prediction of the fitting function does not deviate excessively from the actual thermal degradation process.

[0047] Furthermore, the threshold vector The components of each dimension are determined based on the temperature control performance index. When the deviation of a certain dimension of the fingerprint feature exceeds the corresponding threshold component, it will cause the room temperature overshoot to exceed the set limit or the temperature response time to exceed the allowable range. The threshold components of each dimension are determined through parameter sensitivity tests during the initial debugging phase.

[0048] Trajectory fitting was performed on the deseasonalized fingerprint trend sequence of this office building. Using three years of fingerprint data points as samples, second-order polynomial regression was used to fit each dimension. Based on the degradation slope of each dimension, the polynomial function was... The coefficients are shown in the table below: Table 4. Coefficients of the trajectory fitting polynomial (partial dimensions): At the present moment The drift direction vector is calculated at the year (the 10th month of the third year). The first derivative of each dimension is... ,therefore The drift rate is Year .

[0049] Preset temperature control mismatch threshold vector This corresponds to the allowable deviation for each dimension. By solving the fitting function and extrapolating, the moment when each feature dimension first exceeds the threshold is calculated. For dimension 1 (where the time constant grows the fastest), the equation is solved... ,get Year. Based on predictions from various dimensions, the minimum mismatch time is [year]. Year. Substituting this into the fitting function, the predicted fingerprint value at the mismatch time is... .

[0050] Figure 5 This demonstrates the polynomial-fitted trajectory for dimension 1 (time constant).

[0051] Step 5: Calculate the interval between the current time and the expected mismatch time point, compare it with the preset adaptive adjustment lead, determine whether to trigger the temperature control parameter adaptive adjustment process, and generate an adaptive adjustment trigger signal.

[0052] Get the current time Calculate the time interval between the current moment and the expected mismatch time. .

[0053] Preset adaptive lead adjustment The lead time indicates how long before the anticipated mismatch should be initiated to ensure a smooth transition. The time interval is compared to the lead time; if... Then an adaptive adjustment trigger signal is generated, initiating the temperature control parameter adaptive adjustment process; if If the current control parameters are not met, the system will continue to operate and wait for reassessment in the next cycle.

[0054] Furthermore, the adaptive adjustment of lead amount The lead time should be determined based on the transition period of parameter adjustment and the uncertainty of trajectory prediction. It should be greater than the parameter transition period to ensure that the adjustment process is completed before mismatch occurs. At the same time, the influence of prediction error should be taken into account to leave a safety margin. It is usually set to 1.5 to 2 times the parameter transition period.

[0055] Step 6: Calculate the difference between the future fingerprint prediction value and the initial baseline fingerprint vector, input the difference into the parameter sensitivity mapping matrix to calculate the correction increment corresponding to each temperature control parameter, and generate an adaptive parameter adjustment scheme.

[0056] Obtain the initial baseline fingerprint vector The initial baseline fingerprint vector is the building thermal response fingerprint collected and calibrated during the initial commissioning phase of the heating system. The difference vector between the predicted future fingerprint value and the initial baseline fingerprint vector is calculated along each feature dimension. .

[0057] Input the difference vector into the parameter sensitivity mapping matrix. In the middle, calculate the temperature control parameter correction increment vector. ,in for 3D matrix The number of temperature control parameters. The dimension of the fingerprint vector, matrix elements Indicates the first The control parameter affects the first... The sensitivity coefficient for changes in each fingerprint feature dimension. The temperature control parameters include proportional gain, integral time constant, derivative time constant, and feedforward compensation coefficient. The correction increment vector is added to the current control parameters to generate an adaptive parameter adjustment scheme.

[0058] It should be noted that the above parameter sensitivity mapping matrix The method for obtaining the data is as follows: During the initial commissioning phase of the heating system, small-amplitude perturbation tests are conducted on each temperature control parameter, and the corresponding thermal response fingerprint changes are recorded. A linear mapping relationship between the fingerprint changes and parameter changes is established using the least squares method. When establishing the mapping relationship, since the fingerprint features have already been converted to dimensionless standardized values ​​through Z-score standardization in step 1, the matrix elements... The dimensional information of the fingerprint features has been absorbed during the fitting process, resulting in a dimensionless difference vector. The corrected increment vector is obtained by multiplying the changes from the standardized fingerprint by the matrix. Each component automatically has the correct dimensions of the corresponding control parameters (proportional gain is dimensionless, integral and derivative time constants are in time units, and feedforward compensation coefficients are dimensionless), ensuring the consistency of dimensional calculations.

[0059] Furthermore, the specific operation procedure for the small-amplitude disturbance test is as follows: The proportional gain, integral time constant, derivative time constant, and feedforward compensation coefficient are each disturbed individually, with the disturbance amplitude for each parameter being 5% to 10% of its set value. Under the disturbance state, the heating system is kept running for at least one complete heating cycle (one week). Thermal response fingerprint data is collected within this cycle, and the fingerprint change before and after the disturbance is calculated. and the change in the corresponding parameters ; Establish a least squares optimization problem, and solve it to make smallest matrix element ,in The summation symbol represents the number of temperature control parameters. Traverse from 1 to Indicates the parameter index. The number of dimensions of the fingerprint features. Indicates the first A unit change in the parameter affects the first... The influence coefficient of 3D fingerprint features.

[0060] In this embodiment of the application, in order to obtain accurate parameter correction amount when the fingerprint feature is greatly offset, the parameter sensitivity mapping matrix can be in the form of piecewise linear mapping, that is, different mapping matrices are used in different fingerprint offset intervals to adapt to the nonlinear characteristics of parameter sensitivity changing with the working point.

[0061] Furthermore, the fingerprint offset interval is based on the difference vector. Euclidean norm The fingerprint deviation is divided into several equal intervals from zero to the mismatch threshold. Each interval corresponds to a mapping matrix calibrated by parameter perturbation test within that deviation range, ensuring the accuracy of parameter correction at different deviation levels.

[0062] After the office building determines the trigger parameter adjustment in year 3.2, it calculates the difference between the future fingerprint prediction value and the initial baseline fingerprint vector. The initial baseline fingerprint vector is... The predicted fingerprint value at the time of mismatch is Therefore, the difference vector is .

[0063] The parameter sensitivity mapping matrix established during the initial commissioning phase of the building is as follows: The matrix (4 control parameters, 5-dimensional fingerprint features) is shown in the table below. Table 5 Parameter sensitivity mapping matrix : Through matrix operations The temperature control parameter correction increment vector is calculated as follows: These correspond to the correction increments for proportional gain, integral time, derivative time, and feedforward compensation coefficient, respectively.

[0064] The current control parameters are (These are proportional gain, integral time (minutes), derivative time (minutes), and feedforward compensation coefficient, respectively), the generated adaptive parameter adjustment scheme is as follows: .

[0065] Figure 6 The sensitivity coefficient matrix between four temperature control parameters and five fingerprint feature dimensions is displayed.

[0066] In addition to the steps described above, the following steps are also included: Step 7: Calculate the deviation between the adaptive parameter adjustment scheme and the current temperature control operating parameters, decompose it into multiple fine-tuning steps in a linear time manner, use a progressive switching algorithm to smoothly transition the current parameters to the target parameters, and output the temperature control parameter fine-tuning instruction sequence.

[0067] Get the current temperature control operating parameter vector Calculate the deviation vector between the adaptive parameter adjustment scheme and the current parameters. ,in The target parameters are determined for the adaptive parameter adjustment scheme. The parameter transition period is set. and fine-tuning the number of steps The deviation vector is linearly decomposed over time into A fine-tuning increment .

[0068] Furthermore, the parameter transition period Based on the requirements for building thermal inertia and room temperature control stability, the transition period should be long enough to avoid room temperature oscillations caused by excessively rapid parameter changes; it is typically set to 2 to 5 times the building temperature response time constant. The number of fine-tuning steps... The transition period and the sampling period of the control system are determined to ensure that each fine-tuning increment is small enough to maintain a smooth transition.

[0069] Furthermore, the building temperature response time constant is obtained by fitting historically collected step response data, using the first-order system model from step 1. By fitting the temperature response curve of the most recent heating season, the time constant can be obtained. This is the building temperature response time constant.

[0070] according to The parameter fine-tuning instructions are output sequentially at time intervals. Each instruction increments the current parameter by a fine-tuning increment. After a few fine-tuning steps, the current parameters smoothly transition to the target parameters.

[0071] Before outputting the parameter fine-tuning instructions, it is necessary to verify whether the parameter values ​​after each fine-tuning step meet the constraints. ,in and For the first The algorithm defines the lower and upper bounds of the allowable range for each control parameter. If a fine-tuning increment in a certain step causes a parameter to exceed its allowable range, the increment should be adjusted to bring it precisely to the range boundary. This gradual switching algorithm avoids room temperature fluctuations caused by sudden parameter changes, ensuring stable operation of the temperature control system during parameter adjustments.

[0072] Figure 7 This demonstrates an adjustment scheme for four control parameters to smoothly transition from their current values ​​to target values.

[0073] Step 8: Continuously collect indoor temperature data to monitor the deviation between the actual fingerprint change and the predicted trajectory. If the deviation exceeds the preset tolerance, correct the trajectory prediction and update the adaptive adjustment plan, and output the trajectory correction signal to realize closed-loop adaptive temperature control.

[0074] During and after parameter adjustment, indoor temperature response data is continuously collected at fixed intervals, and the current thermal response fingerprint is extracted using the method in step 1. The deviation between the current actual fingerprint and the predicted fingerprint of the fitted trajectory in step 4 at the corresponding time is calculated. .

[0075] Preset trajectory deviation tolerance vector The index of the set This indicates the fingerprint feature dimension number, and it determines whether the deviation of each dimension exceeds the corresponding tolerance. If a dimension deviation exceeds the tolerance, a trajectory correction signal is output. The trajectory fitting function is updated using the current actual fingerprint data, the expected mismatch time point and future fingerprint prediction values ​​are recalculated, and the adaptive parameter adjustment scheme is updated accordingly. Through the above closed-loop correction mechanism, the prediction deviation is continuously corrected, realizing dynamic tracking of the building thermal characteristic degradation trajectory and continuous adaptive optimization of temperature control parameters.

[0076] Furthermore, the trajectory deviation tolerance vector The components of each dimension are determined based on the prediction error of the fitting function and the measurement accuracy of the sensor. The tolerance should be greater than the sum of the sensor noise level and the fitting residual by one standard deviation to avoid frequent corrections due to measurement noise and model uncertainty. At the same time, it should be less than 50% of the mismatch threshold to ensure timely detection of trajectory deviations.

[0077] Furthermore, the update method for the trajectory fitting function is as follows: to ensure the fitted trajectory better tracks the latest changing trends of the building's thermal properties, newly acquired data points should be given greater weight. Specifically, for each data point in the sample set... Assign weights ,in It is the base of the natural logarithm (exponential function). This is the attenuation coefficient (usually taken as 0.01). To determine the current time, data points closer to the current time are given greater weight. Newly collected fingerprint data points are appended to the existing sample set, and weights are applied. A new multinomial weighted regression is then performed on the weighted sample set to obtain the updated fitting function. The updated fitting function is used to recalculate the moment when each feature dimension first exceeds the threshold, thus obtaining the updated predicted mismatch time point. and future fingerprint prediction values .

[0078] Figure 8 This demonstrates the process of monitoring the deviation between the actual fingerprint and the predicted trajectory.

[0079] This implementation method establishes a time-series thermal response fingerprint vector database and performs seasonal decomposition processing, abstracting building thermal response characteristics into quantifiable and traceable fingerprint vectors. It separates periodic fluctuation components and long-term trend components from multi-period data acquisition, overcoming the factor that short-term data fluctuations mask long-term degradation trends. By performing trajectory fitting and time extrapolation prediction in the fingerprint feature space, it can predict when fingerprint features exceed the temperature control mismatch threshold, overcoming the lag factor of traditional methods that can only passively respond after the control effect deteriorates. A quantitative relationship between fingerprint changes and temperature control parameter correction is established through a parameter sensitivity mapping matrix, enabling adaptive calculation of parameter correction based on the predicted thermal characteristic degradation. Therefore, this implementation method solves the problem that heating systems cannot achieve data-based adaptive temperature control in scenarios of slow building thermal characteristic degradation.

[0080] The embodiments of the present invention have been described above. However, the embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments under the guidance of the present embodiments, and all of them are within the protection scope of the present embodiments.

Claims

1. An adaptive temperature control method for a heating system based on data acquisition, characterized in that, Includes the following steps: Indoor temperature response data of buildings under standard thermal excitation test is collected, and feature extraction and dimensionality reduction are performed on the temperature response data to generate a time-series thermal response fingerprint vector database. The fingerprint vectors in the time-series thermal response fingerprint vector database are subjected to time-series decomposition processing to separate periodic fluctuation components and trend components. After removing the periodic fluctuation components, a deseasonalized fingerprint trend sequence is generated. Trajectory fitting is performed on the deseasoned fingerprint trend sequence to predict the evolution function of each feature dimension over time. Time extrapolation is performed based on a preset temperature control mismatch threshold to calculate the time when the fingerprint feature is expected to exceed the mismatch threshold, and the expected mismatch time point and future fingerprint prediction value are generated. The difference between the predicted future fingerprint value and the initial baseline fingerprint vector is calculated. The difference is then input into the parameter sensitivity mapping matrix to calculate the temperature control parameter correction increment, and an adaptive parameter adjustment scheme is generated and output. The parameter sensitivity mapping matrix is ​​used to characterize the sensitivity coefficient of each temperature control parameter to changes in each fingerprint feature dimension.

2. The adaptive temperature control method for a heating system based on data acquisition according to claim 1, characterized in that, The time-series decomposition process employs the STL decomposition algorithm, which uses iterative local weighted regression to separate the components of each dimension of the fingerprint vector into seasonal, trend, and residual terms.

3. The adaptive temperature control method for a heating system based on data acquisition according to claim 1, characterized in that, Before performing trajectory fitting on the deseasoned fingerprint trend sequence, the method further includes: A sliding window statistical analysis was performed on the deseasoned fingerprint trend sequence to calculate the mean and variance of each feature dimension within each window; The changes in the mean and variance of each dimension between adjacent windows are statistically analyzed, the slope of the change trend is calculated, and a fingerprint degradation trend feature set is generated.

4. The adaptive temperature control method for a heating system based on data acquisition according to claim 1, characterized in that, The trajectory fitting uses a multinomial regression algorithm to fit the time evolution of each feature dimension separately, obtaining a multinomial function of each dimension changing with time.

5. The adaptive temperature control method for a heating system based on data acquisition according to claim 1, characterized in that, The parameter sensitivity mapping matrix is ​​obtained by performing disturbance tests on each temperature control parameter during the initial commissioning phase of the heating system, recording the corresponding thermal response fingerprint changes, and using the least squares method to fit and establish a linear mapping relationship between fingerprint changes and parameter changes.

6. The adaptive temperature control method for a heating system based on data acquisition according to claim 1, characterized in that, Before generating and outputting the adaptive parameter tuning scheme, the following steps are also included: Calculate the time interval between the current moment and the predicted mismatch time point; The time intervals are compared with a preset adaptive adjustment lead amount; If the time interval is less than or equal to the adaptive adjustment lead amount, the temperature control parameter adaptive adjustment process is triggered.

7. The adaptive temperature control method for a heating system based on data acquisition according to claim 1, characterized in that, The output adaptive parameter adjustment scheme includes: Calculate the deviation between the target parameter and the current temperature control operating parameter in the adaptive parameter adjustment scheme; The deviation is linearly decomposed over time into multiple fine-tuning increments; The parameter fine-tuning instructions are output sequentially at preset time intervals to smoothly transition the current parameter to the target parameter.

8. The adaptive temperature control method for a heating system based on data acquisition according to claim 1, characterized in that, Also includes: Continuously collect indoor temperature response data and extract the current thermal response fingerprint; Calculate the deviation between the current actual fingerprint and the predicted fingerprint of the fitted trajectory at the corresponding time. If the deviation exceeds the preset trajectory deviation tolerance, the trajectory fitting function is updated using the current actual fingerprint data, the expected mismatch time point and future fingerprint prediction value are recalculated, and the adaptive parameter adjustment scheme is updated.

9. The adaptive temperature control method for a heating system based on data acquisition according to claim 1, characterized in that, The dimensionality reduction process employs principal component analysis, retaining principal components whose cumulative contribution rate exceeds a preset threshold as the components of each dimension of the fingerprint vector.

10. A data acquisition-based adaptive temperature control system for a heating system, used to execute the data acquisition-based adaptive temperature control method for a heating system as described in any one of claims 1 to 9, characterized in that, include: The fingerprint generation module is used to collect indoor temperature response data of buildings under standard thermal excitation tests, perform feature extraction and dimensionality reduction on the temperature response data, and generate a time-series thermal response fingerprint vector database. The seasonal decomposition module is used to perform time-series decomposition processing on the fingerprint vectors in the time-series thermal response fingerprint vector database, separate the periodic fluctuation component and the trend component, and generate a deseasonalized fingerprint trend sequence. The trajectory prediction module is used to perform trajectory fitting on the deseasoned fingerprint trend sequence, extrapolate the time based on the temperature-controlled mismatch threshold, and generate the expected mismatch time point and future fingerprint prediction value. The parameter adjustment module is used to calculate the difference between the predicted future fingerprint value and the initial baseline fingerprint vector, input the difference into the parameter sensitivity mapping matrix to calculate the temperature control parameter correction increment, and generate and output an adaptive parameter adjustment scheme.

Images