Building energy-saving optimization method and system based on big data analysis

Abstract

The present invention relates to the field of energy-saving optimization of data analysis, and specifically to a building energy-saving optimization method and system based on big data analysis. The method obtains the operating data of a cooling water system and a chilled water system and generates a standardized time series data set in combination with the ambient wet-bulb temperature. The final evaporation temperature prediction value and the final condensation pressure prediction value are obtained through long-short-term memory network analysis, and the data are analyzed together with preset hard constraints. If the final evaporation temperature prediction value and the final condensation pressure prediction value both meet the preset hard constraints, a comprehensive energy efficiency value is obtained through energy efficiency analysis calculation; the minimum comprehensive energy efficiency value that meets the hard constraints is iteratively solved through an interior point method, and the corresponding frequencies of the cooling water pump and the chilled water pump are output as the optimal frequency instruction sequence and sent to the frequency conversion controller; the method can effectively strengthen the collaborative analysis of the cooling water system and the chilled water system, comprehensively consider the dynamic coupling relationship between the various parameters, and improve the overall energy efficiency.

Description

Technical Field

[0001] The present invention relates to the field of energy-saving optimization through data analysis, and specifically to a building energy-saving optimization method and system based on big data analysis. Background Art

[0002] With the continuous expansion of modern buildings and the dramatic increase in energy consumption, building energy conservation has become a critical issue that needs to be addressed. Within the field of building energy conservation, heating, ventilation, and air conditioning (HVAC) systems account for over 40% of a building's total energy consumption. Cooling and chilled water systems are crucial components of air conditioning systems, so coordinated control of these systems is crucial for energy-saving optimization.

[0003] However, traditional cooling and chilled water system controls typically rely on preset fixed parameters or simple feedback control strategies, such as PID (proportional-integral-derivative) control. These strategies struggle to cope with dynamic system changes, failing to accurately adjust pump frequency in real time and, consequently, failing to maximize system energy efficiency. Current control methods often overlook the interplay between these parameters, as system dynamics are complexly coupled to the operating parameters of the cooling and chilled water systems.

[0004] Therefore, how to achieve more refined coordinated analysis of cooling water systems and chilled water systems, comprehensively analyze the dynamic coupling relationship and uncertainty factors between system parameters, and improve the overall energy efficiency of the system has become a technical problem that needs to be solved urgently. Summary of the Invention

[0005] (1) Technical problems to be solved

[0006] The purpose of the present invention is to provide a building energy-saving optimization method and system based on big data analysis to solve the problem that the building's air-conditioning system does not fully consider the coordinated control of the cooling water system and the chilled water system, resulting in the inability to adjust the frequency of the cooling water pump and the chilled water pump in real time and accurately when the system changes, thereby failing to effectively improve the system energy efficiency and achieve building energy saving.

[0007] (2) Technical solution

[0008] To achieve the above objectives, the present invention provides, on the one hand, a building energy-saving optimization method based on big data analysis, the method comprising:

[0009] S1. Real-time acquisition of the cooling water tower outlet water temperature, cooling water flow, cooling water pump operating frequency, and condensing pressure of the cooling water system, as well as the chilled water supply and return water temperature difference, temperature recovery rate, chilled water pump operating frequency, and evaporation temperature of the chilled water system. Aligning the data with the ambient wet-bulb temperature results in unified timestamps, and then using the sliding window linear interpolation method to fill in missing values ​​to generate a standardized time series dataset.

[0010] S2. The standardized time series data set is analyzed through a long short-term memory network to analyze the hysteresis effect of cooling water flow changes on the chilled water evaporation temperature and the conduction effect of the chilled water temperature recovery rate on the condensing pressure, and the final evaporation temperature prediction value and the final condensing pressure prediction value are obtained.

[0011] S3. Perform data analysis on the final evaporation temperature prediction value and the final condensation pressure prediction value and the preset hard constraint conditions to obtain an analysis result. If the analysis result shows that the final evaporation temperature prediction value and the final condensation pressure prediction value both meet the preset hard constraint conditions, a comprehensive energy efficiency value is obtained through energy efficiency analysis calculation.

[0012] S4. Iteratively solve the minimum comprehensive energy efficiency value that meets the hard constraints through the interior point method, output the frequencies of the cooling water pump and the refrigeration pump corresponding to the minimum comprehensive energy efficiency value as the optimal frequency instruction sequence and send it to the frequency conversion controller.

[0013] Furthermore, the method of analyzing the hysteresis effect of cooling water flow rate change on chilled water evaporation temperature and the conduction effect of chilled water temperature recovery rate on condensation pressure of the standardized time series data set through a long short-term memory network to obtain the final evaporation temperature prediction value and the final condensation pressure prediction value includes:

[0014] The cooling water flow time series data and the chilled water evaporation temperature time series data are obtained and time-lag cross-correlation analysis is performed to obtain the peak lag time of the lag effect of cooling water flow change on chilled water evaporation temperature. The historical cooling water flow data within the lag time window is extracted and input into the long short-term memory network to obtain the preliminary evaporation temperature prediction value. The sensitivity coefficient of the cooling water pump frequency change to the chilled water evaporation temperature is obtained through automatic differentiation calculation.

[0015] The chilled water temperature recovery rate time series data and the condensing pressure time series data are obtained, and the transmission direction of the chilled water temperature recovery rate to the condensing pressure is determined by statistical analysis methods, and the lag time step of the influence of the chilled water temperature recovery rate on the condensing pressure is obtained. According to the lag time step and the ambient wet-bulb temperature, the input eigenvector of the long-short-term memory network is constructed, and the preliminary condensing pressure prediction value is output. The transmission coefficient of the chilled water temperature recovery rate to the condensing pressure is calculated.

[0016] The sensitivity coefficient and the conduction coefficient are subjected to multi-factor collaborative prediction analysis to obtain a final evaporation temperature prediction value and a final condensation pressure prediction value.

[0017] Furthermore, the method of obtaining the final evaporation temperature prediction value and the final condensation pressure prediction value by performing a multi-factor collaborative prediction analysis on the sensitivity coefficient and the conductance coefficient includes:

[0018] The cooling water pump frequency change and its sensitivity coefficient are obtained, and combined with the cumulative correction term of the cooling water flow rate within the lag time window, the preliminary evaporation temperature prediction value is corrected to obtain the final evaporation temperature prediction value.

[0019] The chilled water temperature recovery rate and its conduction coefficient are obtained, and a real-time correction term for the ambient wet-bulb temperature is introduced to dynamically adjust the initial condensing pressure prediction value to obtain the final condensing pressure prediction value.

[0020] Furthermore, the method of obtaining cooling water flow time series data and chilled water evaporation temperature time series data and performing time lag cross-correlation analysis to obtain the peak lag time of the lag effect of cooling water flow change on chilled water evaporation temperature, and extracting historical cooling water flow data within the lag time window and inputting it into the long short-term memory network includes:

[0021] The cooling water flow rate and the chilled water evaporation temperature are normalized and the correlation coefficient of the preset sliding window is calculated.

[0022] The peak lag time corresponding to the maximum value of the cross-correlation coefficient was obtained, and the statistical significance of the peak lag time was verified by bootstrap resampling, and the lag time window of the influence of cooling water flow rate change on chilled water evaporation temperature was obtained.

[0023] Furthermore, the method of verifying the statistical significance of the peak lag time by resampling using the bootstrap method and obtaining the lag time window of the effect of the cooling water flow rate change on the chilled water evaporation temperature includes:

[0024] The cooling water flow and chilled water evaporation temperature are phase-randomized to generate a preset number of alternative time series and calculate the correlation coefficient. The peak lag time is extracted and the distribution histogram of the peak lag time is obtained through statistical analysis. The distribution histogram is then used to obtain the lag time window according to the preset confidence interval boundaries.

[0025] Furthermore, the method for obtaining the comprehensive energy efficiency value through energy efficiency analysis and calculation includes:

[0026] The comprehensive energy efficiency value is obtained by calculating the comprehensive energy efficiency function; the comprehensive energy efficiency function is established within a preset rolling time window based on the cooling water pump frequency energy consumption item, the refrigeration pump frequency energy consumption item and the temperature deviation penalty item; the temperature deviation penalty adopts an asymmetric penalty strategy, applying a square penalty to the cooling tower outlet water temperature higher than the set value, and applying a linear penalty to the cooling tower outlet water temperature lower than the set value.

[0027] Furthermore, the method of iteratively solving the minimum comprehensive energy efficiency value that satisfies the hard constraints by the interior point method and outputting the frequencies of the cooling water pump and the refrigeration pump corresponding to the minimum comprehensive energy efficiency value as the optimal frequency instruction sequence includes:

[0028] Based on the optimal frequency instruction sequence obtained at the previous time, a guessed frequency instruction sequence for the current time is generated through linear extrapolation; the guessed frequency instruction sequence is dynamically iterated in combination with the gradient direction and momentum factor to obtain a guessed solution value; the guessed solution value includes the evaporation temperature, condensation pressure, and frequency change rate; the momentum factor dynamically increases with the number of iterations.

[0029] The guessed solution value is mapped to the feasible domain that satisfies the hard constraints, and the evaporation temperature is corrected with a safety margin. When the residual of the detection iteration is less than the preset residual convergence threshold, it is judged to be converged. At this time, the comprehensive energy efficiency value reaches the minimum value, and the optimal frequency instruction sequence is output. If it is judged to be non-converged, the iteration is continued until it is less than the preset residual convergence threshold.

[0030] Furthermore, when the residual of the detection iteration is less than a preset residual convergence threshold, the method of determining convergence includes:

[0031] The residual includes the original residual and the dual residual. The original residual evaluates the degree to which the guessed solution satisfies the hard constraints, and the dual residual evaluates the degree of gradient convergence. When both the original residual and the dual residual are less than the preset residual convergence threshold, it is judged to be converged.

[0032] Furthermore, the hard constraints are:

[0033] The hard constraints include a dynamic threshold of evaporation temperature, an upper limit of condensation pressure, and a frequency change rate limit.

[0034] The real-time lower limit of the evaporation temperature dynamic threshold is obtained by querying the corresponding saturation temperature through the refrigerant physical property table of the current condensing pressure and subtracting the compressor suction superheat setting value; the real-time upper limit of the evaporation temperature is set by the maximum evaporation temperature allowed by safety.

[0035] The upper limit of the condensing pressure is dynamically adjusted according to 85% of the rated pressure, and a sinusoidal fluctuation term is introduced to match the impact of day and night ambient temperature changes on the heat dissipation efficiency. The phase offset of the sine function is determined by fitting the time of occurrence of the peak ambient temperature on that day using the least squares method.

[0036] The frequency change rate is limited to ensure that the frequency change amount of the cooling water pump and the refrigeration pump in adjacent control cycles does not exceed a preset safe frequency change threshold.

[0037] On the other hand, based on the same inventive concept, the present invention also provides a building energy-saving optimization system based on big data analysis, the system comprising: a standardized time series data set generation module, a prediction value generation module, a comprehensive energy efficiency value analysis module, and an optimal frequency instruction sequence analysis module, wherein the modules are sequentially connected to each other in communication;

[0038] The standardized time series dataset generation module is used to obtain the cooling water tower outlet water temperature, cooling water flow, cooling water pump operating frequency, condensing pressure of the cooling water system in real time, as well as the chilled water supply and return water temperature difference, temperature recovery rate, chilled water pump operating frequency, and evaporation temperature of the chilled water system. After aligning the unified timestamps with the ambient wet-bulb temperature, the module uses the sliding window linear interpolation method to fill in the missing values ​​and generate a standardized time series dataset.

[0039] The prediction value generation module is used to analyze the hysteresis effect of cooling water flow changes on chilled water evaporation temperature and the conduction effect of chilled water temperature recovery rate on condensation pressure through the long short-term memory network on the standardized time series data set, and obtain the final evaporation temperature prediction value and the final condensation pressure prediction value.

[0040] The comprehensive energy efficiency value analysis module is used to perform data analysis on the final evaporation temperature prediction value and the final condensation pressure prediction value with the preset hard constraints to obtain the analysis results. If the analysis results show that the final evaporation temperature prediction value and the final condensation pressure prediction value both meet the preset hard constraints, the comprehensive energy efficiency value is obtained through energy efficiency analysis calculation.

[0041] The optimal frequency instruction sequence analysis module is used to iteratively solve the minimum comprehensive energy efficiency value that meets the hard constraints through the interior point method, output the frequencies of the cooling water pump and the refrigeration pump corresponding to the minimum comprehensive energy efficiency value as the optimal frequency instruction sequence and send it to the frequency conversion controller.

[0042] (3) Beneficial effects

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] 1. By combining long-short-term memory networks with time-delay cross-correlation analysis and bootstrap resampling verification, the hysteresis effect of cooling water flow on evaporation temperature and the transmission mechanism of chilled water temperature recovery rate on condensing pressure are accurately quantified. This can dynamically capture the nonlinear coupling relationship of multiple parameters, reduce the prediction error of evaporation temperature and condensing pressure, and provide reliable input for optimization control.

[0045] 2. Generate the initial guess solution through linear extrapolation, and adjust the search direction in combination with the dynamic incremental momentum factor, effectively reducing the number of iterations, shortening the time to generate the optimal frequency instruction, and effectively reducing the comprehensive energy consumption of the cooling water pump and the refrigeration pump.

[0046] 3. The evaporation temperature threshold is dynamically adjusted according to the real-time condensing pressure, and a sinusoidal fluctuation term is introduced into the condensing pressure upper limit to match the changes in heat dissipation efficiency during the day and night, as well as the frequency change rate limit, which can actively adapt to environmental disturbances such as the day and night temperature difference. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 This is a flowchart of the building energy-saving optimization method based on big data analysis according to Example 1 of the present invention.

[0048] Figure 2 This is a schematic diagram of the module composition of the building energy-saving optimization system based on big data analysis in Example 2 of the present invention. DETAILED DESCRIPTION

[0049] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0050] Before giving examples, it is necessary to explain the application scenarios of the present invention. This invention is a building energy-saving optimization method and system based on big data analysis. For example, a commercial complex is equipped with a modern central air-conditioning system, which includes a cooling water system and a chilled water system. Due to the large building area and the large number of users, energy consumption has always been a key issue for management. During the peak summer season, the commercial complex's air conditioning energy consumption accounts for more than 45% of the total energy consumption. Traditional fixed parameter control methods cannot effectively cope with load changes, resulting in energy waste.

[0051] Example 1: Figure 1 As shown, this embodiment provides a building energy-saving optimization method based on big data analysis, the method comprising:

[0052] S1. Real-time acquisition of the cooling water system's cooling tower outlet water temperature, cooling water flow, cooling water pump operating frequency, and condensing pressure, as well as the chilled water supply and return water temperature difference, temperature recovery rate, chilled water pump operating frequency, and evaporation temperature. Aligning these values ​​with the ambient wet-bulb temperature results in unified timestamps, and then using a sliding window linear interpolation method to fill in missing values, generating a standardized time series dataset.

[0053] For example, during energy-saving optimization of the air conditioning system in a commercial complex, operating data for the cooling water and chilled water systems was collected every five minutes. At 2:00 PM on one particular day, the data recorded included a cooling tower outlet water temperature of 28.5°C, a cooling water flow rate of 420 cubic meters per hour, a cooling water pump operating frequency of 42 Hz, a condensing pressure of 850 kPa, a chilled water supply / return temperature difference of 5.2°C, a temperature recovery rate of 0.8°C per hour, a chilled water pump operating frequency of 38 Hz, an evaporating temperature of 5.8°C, and an ambient wet-bulb temperature of 24.3°C. It was discovered that cooling water flow data at three points between 2:15 PM and 2:25 PM was missing due to a sensor failure. A sliding window linear interpolation method was used, based on the valid data from 1:45 PM to 2:45 PM. Based on the data of 415 cubic meters per hour at 2:10 PM and 425 cubic meters per hour at 2:30 PM, the data for the missing points were calculated to be 418, 420, and 423 cubic meters per hour, respectively. After the data was completed, it was normalized. The cooling tower outlet water temperature was measured using a platinum resistance temperature sensor (such as a PT100), the cooling water flow rate was measured using an ultrasonic flow meter, the cooling water pump operating frequency and the chilled water pump operating frequency were directly read through the inverter communication interface, the condensing pressure was measured using a pressure sensor (range 0-2 MPa), the chilled water supply and return temperature difference was calculated using the difference between two platinum resistance temperature sensors, and the chilled water temperature recovery rate was calculated using a differential algorithm to determine the rate of change of the temperature difference. The evaporation temperature was measured using a temperature sensor at the evaporator outlet, and the ambient wet-bulb temperature was measured using a dry-bulb and wet-bulb humidity sensor (such as an HMP155). Due to clock errors or transmission delays, the data arrival times of different sensors are not synchronized. The timestamp of the ambient wet-bulb temperature is used as a benchmark to interpolate and align other parameters.

[0054] S2. Using a long short-term memory network, the standardized time series dataset is analyzed to determine the hysteresis effect of cooling water flow rate changes on chilled water evaporation temperature and the conduction effect of chilled water temperature recovery rate on condensation pressure, thereby obtaining the final evaporation temperature and final condensation pressure prediction values.

[0055] S3. Performing data analysis on the final evaporation temperature prediction value and the final condensation pressure prediction value and the preset hard constraint conditions to obtain an analysis result. If the analysis result shows that the final evaporation temperature prediction value and the final condensation pressure prediction value both meet the preset hard constraint conditions, then performing energy efficiency analysis to obtain a comprehensive energy efficiency value.

[0056] S4. The minimum comprehensive energy efficiency value that satisfies the hard constraints is iteratively solved using the interior point method. The frequencies of the cooling water pump and the refrigeration pump corresponding to the minimum comprehensive energy efficiency value are output as the optimal frequency instruction sequence and sent to the variable frequency controller. The comprehensive energy efficiency value is calculated within a preset rolling time window, and the minimum comprehensive energy efficiency value is iteratively solved using the interior point method. If convergence conditions are met, the optimal frequency instruction is output and sent to the variable frequency controller for execution.

[0057] The method of analyzing the hysteresis effect of cooling water flow rate change on chilled water evaporation temperature and the conduction effect of chilled water temperature recovery rate on condensation pressure by using a long short-term memory network to analyze the standardized time series data set, and obtaining the final evaporation temperature prediction value and the final condensation pressure prediction value includes:

[0058] The cooling water flow rate time series data and the chilled water evaporation temperature time series data were obtained and time-lag cross-correlation analysis was performed to obtain the peak lag time of the lag effect of the cooling water flow rate change on the chilled water evaporation temperature. The historical cooling water flow rate data within the lag time window was extracted and input into the long-short-term memory network to obtain the preliminary evaporation temperature prediction value. The sensitivity coefficient of the cooling water pump frequency change to the chilled water evaporation temperature was obtained through automatic differentiation calculation.

[0059] Obtaining chilled water temperature recovery rate time series data and condensing pressure time series data, and determining the transmission direction of the chilled water temperature recovery rate to the condensing pressure through statistical analysis methods, and obtaining the lag time step of the influence of the chilled water temperature recovery rate on the condensing pressure, constructing the input feature vector of the long short-term memory network based on the lag time step and the ambient wet-bulb temperature, outputting a preliminary condensing pressure prediction value, and calculating the transmission coefficient of the chilled water temperature recovery rate to the condensing pressure;

[0060] The sensitivity coefficient and the conduction coefficient are subjected to multi-factor collaborative prediction analysis to obtain a final evaporation temperature prediction value and a final condensation pressure prediction value.

[0061] For example, when analyzing the impact of cooling water flow on chilled water evaporation temperature, nearly six hours of cooling water flow and chilled water evaporation temperature data were collected, and a 180-minute sliding window was selected to calculate the correlation coefficient. Historical cooling water flow data within this time window was extracted and input into a long-short-term memory network (LSTM), resulting in a preliminary evaporation temperature prediction of 5.5°C. Automatic differentiation calculations revealed a sensitivity coefficient of 0.08°C / Hz for changes in cooling water pump frequency to chilled water evaporation temperature. Statistical analysis of the time series data of chilled water temperature recovery rate and condensing pressure confirmed a positive correlation between the two. The lag time step for the effect of chilled water temperature recovery rate on condensing pressure was 10 minutes. The 10-minute lag chilled water temperature recovery rate data and the current ambient wet-bulb temperature of 24.3°C were used as input feature vectors for the LSTM network. The output was a preliminary condensing pressure prediction of 855 kPa, and the transfer coefficient of the chilled water temperature recovery rate to condensing pressure was calculated to be 25 kPa / (°C / hour).

[0062] The method of obtaining the final evaporation temperature prediction value and the final condensation pressure prediction value by using the sensitivity coefficient and the conduction coefficient through multi-factor collaborative prediction analysis includes:

[0063] The cooling water pump frequency change and its sensitivity coefficient are obtained, and combined with the cumulative correction term of the cooling water flow rate within the lag time window, the preliminary evaporation temperature prediction value is corrected to obtain the final evaporation temperature prediction value;

[0064] The chilled water temperature recovery rate and its conduction coefficient are obtained, and a real-time correction term for the ambient wet-bulb temperature is introduced to dynamically adjust the initial condensing pressure prediction value to obtain the final condensing pressure prediction value.

[0065] For example, in a multi-factor collaborative forecast analysis, the cooling water pump frequency increased from 40 Hz to 42 Hz over the past 10 minutes, a change of +2 Hz. Multiplying this frequency change by the sensitivity coefficient of 0.08°C / Hz yields a theoretical change in evaporation temperature of -0.16°C due to the frequency change. A negative value indicates a decrease in evaporation temperature due to an increase in frequency. Combined with the cumulative correction term of +0.2°C for cooling water flow within the lag time window, reflecting the recent increase in cooling water flow above the historical average, the initial evaporation temperature forecast of 5.5°C was revised to 5.5-0.16+0.2=5.54°C, rounded to the nearest 5.5°C. In practice, a safety margin was taken into account, resulting in a final evaporation temperature forecast of 5.3°C. The current chilled water temperature recovery rate is 0.8°C / hour, 0.6°C / hour higher than the historical average. This difference of 0.2°C / hour multiplied by the conductivity coefficient of 25 kPa / (°C / hour) yields a correction of +5 kPa. Considering the current ambient wet-bulb temperature of 24.3°C, 0.4°C lower than the previous forecast, a wet-bulb temperature correction of -10 kPa was introduced, adjusting the initial condensing pressure forecast of 855 kPa to 855 + 5 - 10 = 850 kPa. Taking into account the actual operating status of the unit, the final condensing pressure forecast of 865 kPa was obtained, ensuring the accuracy of the forecast and the safety of system operation.

[0066] The method of obtaining cooling water flow time series data and chilled water evaporation temperature time series data and performing time lag cross-correlation analysis to obtain the peak lag time of the lag effect of cooling water flow change on chilled water evaporation temperature, and extracting historical cooling water flow data within the lag time window and inputting it into the long short-term memory network includes:

[0067] Normalize the cooling water flow rate and the chilled water evaporation temperature and calculate the correlation coefficient of the preset sliding window;

[0068] The peak lag time corresponding to the maximum value of the cross-correlation coefficient was obtained, and the statistical significance of the peak lag time was verified by bootstrap resampling, and the lag time window of the influence of cooling water flow rate change on chilled water evaporation temperature was obtained.

[0069] For example, in the time-lagged cross-correlation analysis of cooling water flow and chilled water evaporation temperature, the two sets of data were first normalized, mapping the cooling water flow data range from the original 390-450 cubic meters / hour to the 0-1 interval, and mapping the chilled water evaporation temperature from 5.2-6.4°C to the 0-1 interval. The normalized data eliminated the dimensional difference, facilitating subsequent analysis. Within the preset 180-minute sliding window, the cross-correlation coefficient of the two sets of time series data was calculated. The results showed that the cross-correlation coefficient was 0.42 at a lag of 0 minutes, 0.53 at a lag of 5 minutes, 0.65 at a lag of 10 minutes, and reached a maximum of 0.72 at a lag of 15 minutes. It dropped to 0.61 at a lag of 20 minutes and 0.48 at a lag of 25 minutes. Based on the trend of the cross-correlation coefficient with lag time, it was determined that the peak lag time corresponding to the maximum cross-correlation coefficient of 0.72 was 15 minutes.

[0070] The method of verifying the statistical significance of the peak lag time by resampling using the bootstrap method to obtain the lag time window of the effect of the cooling water flow rate change on the chilled water evaporation temperature includes:

[0071] The cooling water flow and chilled water evaporation temperature are phase-randomized to generate a preset number of alternative time series and calculate the correlation coefficient. The peak lag time is extracted and the distribution histogram of the peak lag time is obtained through statistical analysis. The distribution histogram is then used to obtain the lag time window according to the preset confidence interval boundaries.

[0072] For example, to verify the statistical significance of the peak lag time, a bootstrap resampling verification was used. First, the original data of cooling water flow and chilled water evaporation temperature were Fourier transformed, retaining the amplitude information but randomizing the phase angle, and then inversely transformed to generate 1,000 sets of alternative time series. The cross-correlation coefficient was calculated for each set of alternative sequences, the peak lag time was extracted, and the probability distribution of the peak lag time was obtained. The statistical results showed that the peak lag time of 957 of the 1,000 groups of samples fell within the range of 12-18 minutes, and the resulting distribution histogram showed a clear peak at 15 minutes. Based on the 95% confidence interval (corresponding to ±2 standard deviations), the lag time window was determined to be 12-18 minutes.

[0073] The method for obtaining the comprehensive energy efficiency value through energy efficiency analysis and calculation includes:

[0074] The comprehensive energy efficiency value is obtained by calculating the comprehensive energy efficiency function; the comprehensive energy efficiency function is established within a preset rolling time window based on the cooling water pump frequency energy consumption item, the refrigeration pump frequency energy consumption item and the temperature deviation penalty item; the temperature deviation penalty adopts an asymmetric penalty strategy, applying a square penalty to the cooling tower outlet water temperature higher than the set value, and applying a linear penalty to the cooling tower outlet water temperature lower than the set value.

[0075] For example, a 30-minute rolling time window is used to establish a comprehensive energy efficiency function. The comprehensive energy efficiency function consists of three main components: cooling water pump frequency energy consumption, chiller pump frequency energy consumption, and temperature deviation penalty. Cooling water pump power is approximately proportional to the cube of the frequency. Measured data shows that a frequency of 42 Hz corresponds to an energy consumption of 6.5 kW, 38 Hz to 5.3 kW, and 34 Hz ​​to 4.2 kW. A chiller pump frequency of 38 Hz corresponds to an energy consumption of 5.8 kW, 34 Hz ​​to 4.6 kW, and 30 Hz to 3.5 kW. An asymmetric penalty strategy is used for temperature deviations: When the cooling tower outlet water temperature exceeds the setpoint of 28°C (for example, the actual measured temperature is 28.5°C), a quadratic penalty strategy is applied: k × (28.5 - 28)² = 0.25 kW, where k is the penalty factor of 1 kW / °C², resulting in a penalty of 0.25 kW. When the temperature falls below the setpoint (for example, the actual measured temperature is 27.8°C), a linear penalty of 0.2 × (28 - 27.8) = 0.04 kW is applied, with a penalty factor of 0.2 kW / °C. This asymmetric penalty strategy reflects the fact that excessively high temperatures have a greater impact on cooling efficiency. The resulting overall energy efficiency is 12.55 kW, which includes 6.5 kW of energy consumed by the cooling water pump, 5.8 kW of energy consumed by the chiller pump, and a 0.25 kW temperature deviation penalty.

[0076] The method of iteratively solving the minimum comprehensive energy efficiency value that satisfies the hard constraints by the interior point method and outputting the frequencies of the cooling water pump and the refrigeration pump corresponding to the minimum comprehensive energy efficiency value as the optimal frequency instruction sequence includes:

[0077] Based on the optimal frequency instruction sequence obtained at the previous time, a guessed frequency instruction sequence for the current time is generated through linear extrapolation; the guessed frequency instruction sequence is dynamically iterated in combination with the gradient direction and momentum factor to obtain a guessed solution value; the guessed solution value includes the evaporation temperature, condensation pressure, and frequency change rate; the momentum factor dynamically increases with the number of iterations.

[0078] The guessed solution value is mapped to the feasible domain that satisfies the hard constraints, and the evaporation temperature is corrected with a safety margin. When the residual of the detection iteration is less than the preset residual convergence threshold, it is judged to be converged. At this time, the comprehensive energy efficiency value reaches the minimum value, and the optimal frequency instruction sequence is output. If it is judged to be non-converged, the iteration is continued until it is less than the preset residual convergence threshold.

[0079] For example, during the iterative solution process using the interior point method, the optimal frequency command sequence at the previous time point, 14:55 (41 Hz for the cooling water pump and 37 Hz for the chilled water pump), was first derived through linear extrapolation based on the trend over the past three cycles. The predicted frequency command sequence for 15:00 was: 40 Hz for the cooling water pump and 36 Hz for the chilled water pump. The momentum factor was initially set to 0.1 and increased by 0.05 with each iteration, capped at a maximum of 0.4. In the first iteration, the predicted frequencies were input into the prediction model, resulting in a predicted solution: evaporating temperature 5.4°C, condensing pressure 860 kPa, and a frequency change rate of -1 Hz / cycle for the cooling water pump and -1 Hz / cycle for the chilled water pump. This predicted solution was within the hard constraints: evaporating temperature lower limit 4.5°C, upper limit 7°C; condensing pressure upper limit 980 kPa; and frequency change rate limit ±3 Hz / cycle. To increase operational safety margin, the evaporating temperature was adjusted to 5.5°C by adding a 0.1°C safety margin. The corresponding comprehensive energy efficiency value was calculated to be 12.2 kW, with a primal residual of 0.05 and a dual residual of 0.08, both exceeding the preset residual convergence threshold of 0.01. Iterations continued, and after five iterations, the optimal solution was obtained: a cooling water pump speed of 39 Hz and a chilled water pump speed of 36 Hz, corresponding to an evaporating temperature of 5.6°C and a condensing pressure of 845 kPa. The comprehensive energy efficiency value was 11.8 kW, a primal residual of 0.008, and a dual residual of 0.009, meeting the convergence criteria.

[0080] The method of determining convergence when the residual of the detection iteration is less than a preset residual convergence threshold includes:

[0081] The residual includes the original residual and the dual residual. The original residual evaluates the degree to which the guessed solution satisfies the hard constraints, and the dual residual evaluates the degree of gradient convergence. When both the original residual and the dual residual are less than the preset residual convergence threshold, it is judged to be converged.

[0082] Exemplarily, in the process of judging iterative convergence, dual residual indicators are used: primal residual and dual residual. The primal residual evaluates the degree of satisfaction of the guessed solution to the hard constraints, expressed as the maximum constraint violation; the dual residual evaluates the degree of gradient convergence, expressed as the norm of the gradient vector. The first iteration obtained a primal residual of 0.05, which mainly came from the evaporation temperature being close to the lower limit constraint; the dual residual was 0.08, indicating that the gradient had not yet converged. After the second iteration, the primal residual dropped to 0.03 and the dual residual dropped to 0.05; after the third iteration, the primal residual was 0.015 and the dual residual was 0.025; after the fourth iteration, the primal residual was 0.011 and the dual residual was 0.013; after the fifth iteration, the primal residual dropped to 0.008 and the dual residual dropped to 0.009, both of which were less than the preset residual convergence threshold of 0.01. At this point, the iterations are considered converged, and the calculated comprehensive energy efficiency value of 11.8 kilowatts reaches its minimum. The optimal frequency instruction sequence is output: 39 Hz for the cooling water pump and 36 Hz for the chilled water pump. Statistical analysis of tests shows that convergence is achieved within an average of 4-6 iterations, with calculation time under 2 seconds, meeting real-time control requirements.

[0083] The hard constraints are:

[0084] The hard constraints include a dynamic threshold of evaporation temperature, an upper limit of condensation pressure, and a frequency change rate limit;

[0085] The real-time lower limit of the evaporation temperature dynamic threshold is obtained by querying the corresponding saturation temperature through the refrigerant physical property table of the current condensing pressure and subtracting the compressor suction superheat setting value; the real-time upper limit of the evaporation temperature is set by the maximum evaporation temperature allowed by safety.

[0086] The upper limit of the condensing pressure is dynamically adjusted according to 85% of the rated pressure, and a sinusoidal fluctuation term is introduced to match the impact of day and night ambient temperature changes on the heat dissipation efficiency. The phase offset of the sine function is determined by fitting the time of occurrence of the peak ambient temperature on that day using the least squares method.

[0087] The frequency change rate is limited to ensure that the frequency change amount of the cooling water pump and the refrigeration pump in adjacent control cycles does not exceed a preset safe frequency change threshold.

[0088] For example, hard constraints are key to ensuring safe system operation. The experimentally set dynamic evaporation temperature thresholds include a lower limit for the real-time evaporation temperature obtained by querying the physical properties table for refrigerant R134a. For example, when the condensing pressure is 865 kPa, the corresponding saturation temperature is 32.5°C. After subtracting the compressor suction superheat setpoint of 8°C, the lower limit for the real-time evaporation temperature is 4.5°C. The upper limit for the real-time evaporation temperature is always set at 7°C to maintain a constant chilled water supply temperature. Setting the upper limit for the condensing pressure is more complex. The unit's rated pressure is 1200 kPa, but for safety reasons, 85% of this, or 1020 kPa, is used as the base upper limit. To account for the impact of diurnal ambient temperature fluctuations on heat dissipation efficiency, a sinusoidal fluctuation term with an amplitude of 40 kPa and a 24-hour period is introduced. Least squares fitting of seven days of ambient temperature data reveals that the average temperature peak occurs at 3:00 PM. This is used to determine the phase offset of the sine function. The actual upper limit of the condensation pressure measured at 14:00 was 980 kPa, which was 40 kPa lower than the basic upper limit, which was consistent with the actual situation that the heat dissipation conditions were poor in the afternoon.

[0089] Frequency change rate limits are set to prevent mechanical shock to the equipment. The experiment stipulated that the frequency change of the cooling water pump and chiller pump within adjacent 5-minute control cycles must not exceed a preset safety frequency change threshold of 3 Hz. Long-term testing recorded frequency changes over a 24-hour period. The cooling water pump frequency fluctuated between 34 and 45 Hz, and the chiller pump frequency fluctuated between 30 and 42 Hz. The maximum frequency change rate was 2.8 Hz per cycle, which did not exceed the safety threshold. This hard constraint ensures safe equipment operation while improving energy efficiency.

[0090] Example 2: Based on the same inventive concept, Figure 2 As shown, this embodiment also provides a building energy-saving optimization system based on big data analysis, the system comprising: a standardized time series data set generation module, a prediction value generation module, a comprehensive energy efficiency value analysis module, and an optimal frequency instruction sequence analysis module, wherein each module is sequentially connected to each other in communication;

[0091] The standardized time series dataset generation module is used to obtain the cooling water tower outlet water temperature, cooling water flow, cooling water pump operating frequency, and condensing pressure of the cooling water system in real time, as well as the chilled water supply and return water temperature difference, temperature recovery rate, chilled water pump operating frequency, and evaporation temperature of the chilled water system. It then aligns the unified timestamps with the ambient wet-bulb temperature and uses the sliding window linear interpolation method to fill in missing values ​​to generate a standardized time series dataset.

[0092] The prediction value generation module is used to analyze the hysteresis effect of cooling water flow rate changes on chilled water evaporation temperature and the conduction effect of chilled water temperature recovery rate on condensation pressure through the standardized time series data set through the long short-term memory network, and obtain the final evaporation temperature prediction value and the final condensation pressure prediction value;

[0093] The comprehensive energy efficiency value analysis module is used to perform data analysis on the final evaporation temperature prediction value and the final condensation pressure prediction value and the preset hard constraint conditions to obtain an analysis result. If the analysis result shows that the final evaporation temperature prediction value and the final condensation pressure prediction value both meet the preset hard constraint conditions, the comprehensive energy efficiency value is obtained through energy efficiency analysis calculation;

[0094] The optimal frequency instruction sequence analysis module is used to iteratively solve the minimum comprehensive energy efficiency value that meets the hard constraints through the interior point method, output the frequencies of the cooling water pump and the refrigeration pump corresponding to the minimum comprehensive energy efficiency value as the optimal frequency instruction sequence and send it to the frequency conversion controller.

[0095] It should be noted that, regarding the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated on here.

[0096] Finally, it should be noted that although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments, or make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A building energy-saving optimization method based on big data analysis, characterized in that: The method comprises: The cooling water system's cooling tower outlet water temperature, cooling water flow, cooling water pump operating frequency, and condensing pressure are acquired in real time, as well as the chilled water supply and return water temperature difference, temperature recovery rate, chilled water pump operating frequency, and evaporation temperature of the chilled water system. These data are then aligned with the ambient wet-bulb temperature and missing values ​​are filled using a sliding window linear interpolation method to generate a standardized time series dataset. The standardized time series data set is analyzed using a long short-term memory network to analyze the hysteresis effect of cooling water flow changes on chilled water evaporation temperature and the conduction effect of chilled water temperature recovery rate on condensation pressure, and the final evaporation temperature and final condensation pressure prediction values ​​are obtained. The final evaporation temperature prediction value and the final condensation pressure prediction value are analyzed with the preset hard constraint conditions to obtain an analysis result. If the analysis result shows that the final evaporation temperature prediction value and the final condensation pressure prediction value both meet the preset hard constraint conditions, the comprehensive energy efficiency value is obtained through energy efficiency analysis calculation; The minimum comprehensive energy efficiency value that satisfies the hard constraints is solved iteratively through the interior point method, and the frequencies of the cooling water pump and the refrigeration pump corresponding to the minimum comprehensive energy efficiency value are output as the optimal frequency instruction sequence and sent to the frequency conversion controller; The method of analyzing the hysteresis effect of cooling water flow rate change on chilled water evaporation temperature and the conduction effect of chilled water temperature recovery rate on condensation pressure by using a long short-term memory network to analyze the standardized time series data set, and obtaining the final evaporation temperature prediction value and the final condensation pressure prediction value includes: The cooling water flow rate time series data and the chilled water evaporation temperature time series data were obtained and time-lag cross-correlation analysis was performed to obtain the peak lag time of the lag effect of the cooling water flow rate change on the chilled water evaporation temperature. The historical cooling water flow rate data within the lag time window was extracted and input into the long-short-term memory network to obtain the preliminary evaporation temperature prediction value. The sensitivity coefficient of the cooling water pump frequency change to the chilled water evaporation temperature was obtained through automatic differentiation calculation. Obtaining chilled water temperature recovery rate time series data and condensing pressure time series data, and determining the transmission direction of the chilled water temperature recovery rate to the condensing pressure through statistical analysis methods, and obtaining the lag time step of the influence of the chilled water temperature recovery rate on the condensing pressure, constructing the input feature vector of the long short-term memory network based on the lag time step and the ambient wet-bulb temperature, outputting a preliminary condensing pressure prediction value, and calculating the transmission coefficient of the chilled water temperature recovery rate to the condensing pressure; The sensitivity coefficient and the conduction coefficient are subjected to multi-factor collaborative prediction analysis to obtain a final evaporation temperature prediction value and a final condensation pressure prediction value.

2. The building energy-saving optimization method based on big data analysis according to claim 1 is characterized in that: The method of obtaining the final evaporation temperature prediction value and the final condensation pressure prediction value by using the sensitivity coefficient and the conduction coefficient through multi-factor collaborative prediction analysis includes: The cooling water pump frequency change and its sensitivity coefficient are obtained, and combined with the cumulative correction term of the cooling water flow rate within the lag time window, the preliminary evaporation temperature prediction value is corrected to obtain the final evaporation temperature prediction value; The chilled water temperature recovery rate and its conduction coefficient are obtained, and a real-time correction term for the ambient wet-bulb temperature is introduced to dynamically adjust the initial condensing pressure prediction value to obtain the final condensing pressure prediction value.

3. The building energy-saving optimization method based on big data analysis according to claim 1 is characterized in that: The method of obtaining cooling water flow time series data and chilled water evaporation temperature time series data and performing time lag cross-correlation analysis to obtain the peak lag time of the lag effect of cooling water flow change on chilled water evaporation temperature, and extracting historical cooling water flow data within the lag time window and inputting it into the long short-term memory network includes: Normalize the cooling water flow rate and the chilled water evaporation temperature and calculate the correlation coefficient of the preset sliding window; The peak lag time corresponding to the maximum value of the cross-correlation coefficient was obtained, and the statistical significance of the peak lag time was verified by bootstrap resampling, and the lag time window of the influence of cooling water flow rate change on chilled water evaporation temperature was obtained.

4. The building energy-saving optimization method based on big data analysis according to claim 3 is characterized in that: The method of verifying the statistical significance of the peak lag time by resampling using the bootstrap method to obtain the lag time window of the effect of the cooling water flow rate change on the chilled water evaporation temperature includes: The cooling water flow and chilled water evaporation temperature are phase-randomized to generate a preset number of alternative time series and calculate the correlation coefficient. The peak lag time is extracted and the distribution histogram of the peak lag time is obtained through statistical analysis. The distribution histogram is then used to obtain the lag time window according to the preset confidence interval boundaries.

5. The building energy-saving optimization method based on big data analysis according to claim 1 is characterized in that: The method for obtaining the comprehensive energy efficiency value through energy efficiency analysis and calculation includes: The comprehensive energy efficiency value is obtained by calculating the comprehensive energy efficiency function; the comprehensive energy efficiency function is established within a preset rolling time window based on the cooling water pump frequency energy consumption item, the refrigeration pump frequency energy consumption item and the temperature deviation penalty item; the temperature deviation penalty adopts an asymmetric penalty strategy, applying a square penalty to the cooling tower outlet water temperature higher than the set value, and applying a linear penalty to the cooling tower outlet water temperature lower than the set value.

6. The building energy-saving optimization method based on big data analysis according to claim 1 is characterized in that: The method of iteratively solving the minimum comprehensive energy efficiency value that satisfies the hard constraints by the interior point method and outputting the frequencies of the cooling water pump and the refrigeration pump corresponding to the minimum comprehensive energy efficiency value as the optimal frequency instruction sequence includes: Based on the optimal frequency instruction sequence obtained at the previous time, a guessed frequency instruction sequence for the current time is generated through linear extrapolation; the guessed frequency instruction sequence is dynamically iterated in combination with the gradient direction and momentum factor to obtain a guessed solution value; the guessed solution value includes the evaporation temperature, the condensation pressure, and the frequency change rate; the momentum factor dynamically increases with the number of iterations; The guessed solution value is mapped to the feasible domain that satisfies the hard constraints, and the evaporation temperature is corrected with a safety margin. When the residual of the detection iteration is less than the preset residual convergence threshold, it is judged to be converged. At this time, the comprehensive energy efficiency value reaches the minimum value, and the optimal frequency instruction sequence is output. If it is judged to be non-converged, the iteration is continued until it is less than the preset residual convergence threshold.

7. The building energy-saving optimization method based on big data analysis according to claim 6 is characterized in that: The method of determining convergence when the residual of the detection iteration is less than a preset residual convergence threshold includes: The residual includes the original residual and the dual residual. The original residual evaluates the degree to which the guessed solution satisfies the hard constraints, and the dual residual evaluates the degree of gradient convergence. When both the original residual and the dual residual are less than the preset residual convergence threshold, it is judged to be converged.

8. The building energy-saving optimization method based on big data analysis according to claim 1 is characterized in that: The hard constraints are: The hard constraints include a dynamic threshold of evaporation temperature, an upper limit of condensation pressure, and a frequency change rate limit; The real-time lower limit of the evaporation temperature dynamic threshold is obtained by querying the corresponding saturation temperature in the refrigerant physical property table of the current condensing pressure and subtracting the compressor suction superheat setting value; the real-time upper limit of the evaporation temperature is set by the maximum evaporation temperature allowed by safety; The condensing pressure upper limit is dynamically adjusted to 85% of the rated pressure, and a sinusoidal fluctuation term is introduced to match the effect of diurnal ambient temperature changes on heat dissipation efficiency. The phase offset of the sinusoidal function is determined by fitting the peak ambient temperature time of the day using the least squares method. The frequency change rate is limited to ensure that the frequency change amount of the cooling water pump and the refrigeration pump in adjacent control cycles does not exceed a preset safe frequency change threshold.

9. Building energy-saving optimization system based on big data analysis, characterized by: The system includes: a standardized time series data set generation module, a prediction value generation module, a comprehensive energy efficiency value analysis module, and an optimal frequency instruction sequence analysis module, and each module is sequentially connected to each other; The standardized time series dataset generation module is used to obtain the cooling water tower outlet water temperature, cooling water flow, cooling water pump operating frequency, and condensing pressure of the cooling water system in real time, as well as the chilled water supply and return water temperature difference, temperature recovery rate, chilled water pump operating frequency, and evaporation temperature of the chilled water system. It then aligns the unified timestamps with the ambient wet-bulb temperature and uses the sliding window linear interpolation method to fill in missing values ​​to generate a standardized time series dataset. The prediction value generation module is used to analyze the hysteresis effect of cooling water flow rate changes on chilled water evaporation temperature and the conduction effect of chilled water temperature recovery rate on condensation pressure through the standardized time series data set through the long short-term memory network, and obtain the final evaporation temperature prediction value and the final condensation pressure prediction value; The comprehensive energy efficiency value analysis module is used to perform data analysis on the final evaporation temperature prediction value and the final condensation pressure prediction value and the preset hard constraint conditions to obtain an analysis result. If the analysis result shows that the final evaporation temperature prediction value and the final condensation pressure prediction value both meet the preset hard constraint conditions, the comprehensive energy efficiency value is obtained through energy efficiency analysis calculation; The optimal frequency instruction sequence analysis module is used to iteratively solve the minimum comprehensive energy efficiency value that meets the hard constraints through the interior point method, output the cooling water pump and refrigeration pump frequencies corresponding to the minimum comprehensive energy efficiency value as the optimal frequency instruction sequence and send it to the frequency conversion controller; The method of analyzing the hysteresis effect of cooling water flow rate change on chilled water evaporation temperature and the conduction effect of chilled water temperature recovery rate on condensation pressure by using a long short-term memory network to analyze the standardized time series data set, and obtaining the final evaporation temperature prediction value and the final condensation pressure prediction value includes: The cooling water flow rate time series data and the chilled water evaporation temperature time series data were obtained and time-lag cross-correlation analysis was performed to obtain the peak lag time of the lag effect of the cooling water flow rate change on the chilled water evaporation temperature. The historical cooling water flow rate data within the lag time window was extracted and input into the long-short-term memory network to obtain the preliminary evaporation temperature prediction value. The sensitivity coefficient of the cooling water pump frequency change to the chilled water evaporation temperature was obtained through automatic differentiation calculation. Obtaining chilled water temperature recovery rate time series data and condensing pressure time series data, and determining the transmission direction of the chilled water temperature recovery rate to the condensing pressure through statistical analysis methods, and obtaining the lag time step of the influence of the chilled water temperature recovery rate on the condensing pressure, constructing the input feature vector of the long short-term memory network based on the lag time step and the ambient wet-bulb temperature, outputting a preliminary condensing pressure prediction value, and calculating the transmission coefficient of the chilled water temperature recovery rate to the condensing pressure; The sensitivity coefficient and the conduction coefficient are subjected to multi-factor collaborative prediction analysis to obtain a final evaporation temperature prediction value and a final condensation pressure prediction value.

Images