Ventilation system energy-saving control method and system based on fuzzy prediction

Abstract

The invention discloses a ventilation system energy-saving control method and system based on fuzzy prediction, and the method comprises the steps: capturing a heat storage state of an enclosure structure in real time through a nonlinear observation technology, carrying out the vectorization of the heat storage state into a thermal momentum feature representing the evolution trend of a temperature field, and endowing a fuzzy controller with the physical pre-sensing capability for physical environment evolution, and response lag compensation caused by the thermal capacitance effect of the building structure is realized. According to the scheme, the limitation that system inertia is simplified through a traditional algorithm is broken through, physical energy efficiency verification is carried out by introducing multi-dimensional self-adaptive fuzzy reasoning of thermal momentum compensation and combining the fan pressure flow characteristics and the surge boundary, the prediction precision under the dynamic fluctuation working condition of the temperature field is remarkably improved, and the prediction efficiency is improved. And temperature overshoot and energy consumption loss caused by response time sequence mismatch are effectively eliminated. Finally, deep coupling of control logic and building thermodynamic characteristics is achieved, and the dynamic robustness and the comprehensive energy-saving efficiency of the ventilation system are greatly optimized while the indoor comfort degree is guaranteed.

Description

Technical Field

[0001] This application relates to the field of intelligent control, and more specifically, to an energy-saving control method and system for ventilation systems based on fuzzy prediction. Background Technology

[0002] As modern buildings evolve towards intelligence and low carbon emissions, ventilation systems, as a core component of building environmental regulation, directly impact the overall energy efficiency of buildings. Especially in large commercial and industrial settings, establishing precise ventilation energy-saving control schemes is not only a necessary prerequisite for meeting indoor air quality and thermal comfort requirements, but also a key measure to reduce operating costs and achieve refined energy consumption management.

[0003] However, existing energy-saving control methods for ventilation systems largely rely on traditional PID regulation or basic fuzzy control algorithms. When dealing with complex dynamic environments, these solutions typically simplify the ventilation system to a first-order inertial element, severely neglecting the dynamic hysteresis effects generated by the building envelope, such as walls and floors, acting as thermal capacitors. Limited by a fixed prediction time horizon, existing fuzzy rules often address based solely on the error and its rate of change at the current moment, lacking the ability to quantitatively characterize the heat decay trend over a future period. This leads to a high risk of response timing mismatch due to thermal inertia when facing sudden load changes, resulting in severe indoor temperature overshoot or delayed fan shutdown. This failure in prediction accuracy directly causes significant energy waste. In actual operation, ventilation systems essentially operate in a dual-condition alternation between quasi-steady-state operation and high-frequency dynamic load fluctuations. Existing control strategies often lack adaptive adjustment strategies for excitation and dynamic parameters under these dual conditions and fail to establish a scientific operating condition switching process. This makes it difficult for the controller to correct the control step size and compensate for gains in real time when the temperature field evolution direction reverses or when there are severe load disturbances. Addressing the control defocusing caused by thermal momentum evolution and energy mismatch, how to achieve multi-dimensional adaptive fuzzy reasoning based on thermal storage state perception and perform closed-loop constraint verification in conjunction with the unit's physical energy efficiency boundary has become a core technical challenge that urgently needs to be solved to improve the dynamic control robustness and energy efficiency of ventilation systems.

[0004] Therefore, an optimized energy-saving control method for ventilation systems based on fuzzy prediction is needed. Summary of the Invention

[0005] To address the aforementioned technical problems, this application provides an energy-saving control method and system for ventilation systems based on fuzzy prediction.

[0006] According to one aspect of this application, a method for energy-saving control of a ventilation system based on fuzzy prediction is provided, comprising: S1: Denoise, smooth, and perform time-series hard alignment on the collected supply air temperature, return air temperature, indoor temperature gradient, and supply air volume to obtain the operating condition dataset. S2: Nonlinear observation of the heat storage state is performed on the supply air temperature, return air temperature, supply air volume and indoor temperature gradient in the working condition dataset to obtain the energy mismatch characteristics. S3: Vectorize the thermal momentum feature vector by extracting the thermal storage evolution trend of historical time-series temperature segments and energy mismatch characteristics in the working condition dataset; S4: Extract the temperature set error and its rate of change from the operating condition dataset as the basic addressing input of the fuzzy controller, and perform multi-dimensional adaptive fuzzy inference based on thermal momentum compensation on the thermal momentum feature vector to obtain the ventilation demand benchmark. S5: Input the ventilation demand baseline into the nonlinear constraint module and retrieve the pressure-flow characteristic curve of the ventilation unit and the surge boundary threshold matrix to perform physical energy efficiency boundary verification in order to obtain optimized control commands.

[0007] According to another aspect of this application, a fuzzy prediction-based energy-saving control system for a ventilation system is provided, comprising: The data preprocessing module is used to denoise, smooth, and hard-align the collected supply air temperature, return air temperature, indoor temperature gradient, and supply air volume to obtain the operating condition dataset. The nonlinear observation module for thermal storage state is used to perform nonlinear observation of the supply air temperature, return air temperature, supply air volume and indoor temperature gradient in the operating condition dataset to obtain the energy mismatch characteristics. The vector extraction module is used to extract the thermal momentum feature vector by vectorizing the heat storage evolution trend of historical time-series temperature segments and energy mismatch characteristics in the working condition dataset. The multidimensional adaptive fuzzy inference module is used to extract the temperature set error and its rate of change from the operating condition dataset as the basic addressing input of the fuzzy controller, and to perform multidimensional adaptive fuzzy inference based on thermal momentum compensation on the thermal momentum feature vector to obtain the ventilation demand benchmark. The physical energy efficiency boundary verification module is used to input the ventilation demand benchmark into the nonlinear constraint module and retrieve the pressure-flow characteristic curve of the ventilation unit and the surge boundary threshold matrix to perform physical energy efficiency boundary verification in order to obtain optimized control commands.

[0008] Compared with existing technologies, this application provides a fuzzy prediction-based energy-saving control method and system for ventilation systems. It captures the heat storage state of the building envelope in real time using nonlinear observation technology and vectorizes it into thermal momentum characteristics representing the evolution trend of the temperature field. This endows the fuzzy controller with the ability to predict the evolution of the physical environment and achieves response lag compensation caused by the thermal capacitance effect of the building structure. This scheme breaks through the limitations of traditional algorithms that simplify system inertia. By introducing multi-dimensional adaptive fuzzy inference with thermal momentum compensation and combining the fan pressure and flow characteristics with surge boundaries for physical energy efficiency verification, it not only significantly improves the prediction accuracy under dynamic temperature fluctuation conditions but also effectively eliminates temperature overshoot and energy loss caused by response timing mismatch. Ultimately, it achieves deep coupling between control logic and building thermodynamic characteristics, greatly optimizing the dynamic robustness and overall energy-saving efficiency of the ventilation system while ensuring indoor comfort. Attached Figure Description

[0009] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0010] Figure 1 This is a flowchart of an energy-saving control method for a ventilation system based on fuzzy prediction according to an embodiment of this application; Figure 2 This is a data flow diagram of the energy-saving control method for a ventilation system based on fuzzy prediction according to an embodiment of this application; Figure 3 This is a flowchart of step S2 in the fuzzy prediction-based energy-saving control method for ventilation systems according to an embodiment of this application; Figure 4 This is a block diagram of an energy-saving control system for a ventilation system based on fuzzy prediction, according to an embodiment of this application. Detailed Implementation

[0011] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0012] As indicated in this application and claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not specifically singular and may include plural forms. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of explicitly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

[0013] While this application makes various references to certain modules of the systems according to embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The modules described are merely illustrative, and different aspects of the systems and methods may use different modules.

[0014] Flowcharts are used in this application to illustrate the operations performed by the system according to embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from them.

[0015] The technical solution of this application proposes an energy-saving control method for ventilation systems based on fuzzy prediction. Figure 1 This is a flowchart of an energy-saving control method for a ventilation system based on fuzzy prediction, according to an embodiment of this application. Figure 2 This is a system architecture diagram of a fuzzy prediction-based energy-saving control method for ventilation systems according to an embodiment of this application. Figure 1 and Figure 2 As shown, the energy-saving control method for a ventilation system based on fuzzy prediction according to an embodiment of this application includes the following steps: S1, denoising, smoothing, and time-series hard alignment are performed on the collected supply air temperature, return air temperature, indoor temperature gradient, and supply air volume to obtain an operating condition dataset; S2, nonlinear observation of the heat storage state is performed on the supply air temperature, return air temperature, supply air volume, and indoor temperature gradient in the operating condition dataset to obtain energy mismatch characteristics; S3, the heat storage evolution trend is vectorized and extracted from the historical time-series temperature segments and energy mismatch characteristics in the operating condition dataset to obtain a thermal momentum feature vector; S4, the temperature set error and its error change rate are extracted from the operating condition dataset as the basic addressing input of the fuzzy controller, and multidimensional adaptive fuzzy inference based on thermal momentum compensation is performed on the thermal momentum feature vector to obtain a ventilation demand benchmark; S5, the ventilation demand benchmark is input to the nonlinear constraint module and the pressure-flow characteristic curve of the ventilation unit and the surge boundary threshold matrix are retrieved for physical energy efficiency boundary verification to obtain an optimized control command.

[0016] Specifically, in step S1, the collected supply air temperature, return air temperature, indoor temperature gradient, and supply air volume are denoised, smoothed, and subjected to time-series hard alignment to obtain the operating condition dataset. It should be understood that various physical quantity sensors used in the ventilation system often have different sampling frequencies, communication delays, and unavoidable random high-frequency noise interference. If these interferences are directly incorporated into subsequent fuzzy prediction calculations, it can lead to abnormal oscillations in control commands. Furthermore, due to the inconsistent distribution of the sensors along the time axis, cross-dimensional energy calculations and state observations cannot be directly performed. Therefore, in the technical solution of this application, outlier points are removed through denoising and smoothing processes, and time-series hard alignment technology is used to map heterogeneous data onto a unified discrete-time reference, ensuring that subsequent thermodynamic analysis and energy mismatch calculations have a solid physical observation foundation.

[0017] In practice, the first step is to apply sliding median filtering to the collected supply air temperature, return air temperature, indoor temperature gradient, and supply air volume to obtain a set of filtered signals. Specifically, the sliding median filtering technique can completely eliminate high-frequency statistical noise while preserving the effective edge information of the signal.

[0018] In this process, firstly, a cyclic sliding observation window of length L (usually an odd number) is initialized for each of the four core parameters collected: supply air temperature, return air temperature, indoor temperature gradient, and supply air volume. As the time series progresses, whenever a new physical quantity sample value is acquired, it is pushed into the window queue of the corresponding parameter, and the old data at the front of the window is removed. Subsequently, within each control step, all the original sampled data contained in the current sliding window are rearranged in numerical order to form a temporary numerical sequence in ascending or descending order. Next, the center position of this ordered sequence is located based on the total number of sampling points, and the value at that center position is extracted as the filtered output signal for the current moment. During execution, the statistical median within a local time period is used to represent the true physical state, so that extreme noise points whose values ​​deviate from the normal range are assigned to the two ends of the sequence after sorting, and are ultimately effectively excluded in the output selection at that moment. Furthermore, the sequences obtained from supply air temperature processing, return air temperature processing, indoor temperature gradient processing, and supply air volume processing are synchronously encapsulated to form a set of filtered signals. Each subsequence undergoes independent and parallel sorting and extraction operations, thereby ensuring the physical consistency of the multi-source features in terms of denoising intensity.

[0019] Next, the filtered signal set is subjected to multi-source data clock hard alignment and resampling to obtain an aligned time series. Specifically, the multi-source data clock hard alignment technique can eliminate the phase difference caused by the sampling frequencies of different sensors, and by mapping these heterogeneous data onto a unified discrete time reference, the inconsistency in the spatiotemporal distribution of multi-source data can be eliminated.

[0020] In this process, firstly, a globally unified control step size is determined as a reference clock sequence, which is determined by the starting time point and a fixed sampling interval. Then, the physical timestamps of each sensor in the filtered signal set are traversed, and a hard clock alignment operation is performed, that is, the time scale to which the original data is attached is forcibly aligned to a unified starting discrete time reference point according to the nearest principle or triggered latching logic. For sensor channels that lack direct physical sample values ​​at the reference point, fine resampling processing is performed. Specifically, firstly, in the original unaligned discrete sequence, the nearest historical preceding time point and its corresponding physical reading are searched for, and simultaneously, the immediately following subsequent sampling time point and its reading are searched. Next, the ratio of the difference in numerical change between these two adjacent original sampling points to the difference in time interval is calculated to determine the average evolution rate of the physical quantity within that time period. Then, this evolution rate is multiplied by the time span of the target alignment time point relative to the preceding sampling time point, and the result is accumulated to the physical reading of the preceding time point. Through this linear interpolation calculation principle, the system can scientifically estimate the physical quantity value at the unified reference time. By traversing all sensor dimensions and repeating the above calculation logic, each parameter sequence is stretched or compressed to the same length and has a completely one-to-one index relationship. Finally, it is integrated into an aligned time series with temporal consistency, laying a solid data foundation for the subsequent vertical arrangement of the original feature matrix.

[0021] Furthermore, the supply air temperature, return air temperature, indoor temperature gradient, and supply air volume in the aligned time series are vertically arranged according to the spatial dimension to obtain the original feature matrix. Specifically, by arranging them vertically, the physical state slices at different times can be encapsulated into the same column vector, enabling the control system to simultaneously perceive the temperature field difference between indoors and outdoors, air flow dynamics, and spatial heat storage evolution trends, significantly improving the throughput and spatiotemporal correlation of data processing.

[0022] In this process, firstly, four identical one-dimensional row vectors are extracted after clock hard alignment: the supply air temperature vector, the return air temperature vector, the indoor temperature gradient vector, and the supply air volume vector, and the total number of samples is determined. Next, according to a preset spatial dimension mapping rule, a unique row index value is assigned to each physical feature: supply air temperature occupies the first row, return air temperature the second, indoor temperature gradient the third, and supply air volume the fourth. When executing the matrix construction calculation logic, the mathematical principle of vertical superposition is used. These four one-dimensional feature vectors are used as input rows and vertically concatenated according to the preset row index order. In one example, the row sequence representing supply air temperature is placed at the top of the matrix, followed by the return air temperature sequence, the temperature gradient sequence, and finally the supply air volume sequence as the last row. In this way, a four-row feature matrix with the number of columns equal to the total number of samples is generated. In this data structure, the horizontal axis of the matrix represents the direction of time evolution, while each column represents a snapshot of the physical field state across the entire spatial dimension, synchronously observed by multiple sensors at a specific alignment timestamp. Through this mapping process, the system constructs a structured feature space that represents physical space characteristics in the vertical direction and temporal evolution trends in the horizontal direction.

[0023] Subsequently, range standardization is performed on the original feature matrix to obtain the working condition dataset. Specifically, range standardization maps all heterogeneous features to a uniform dimensionless interval of zero to one, thereby eliminating differences in the numerical distribution among features.

[0024] In this process, firstly, based on the row indices of the original feature matrix, each feature dimension is traversed sequentially, scanning the four data rows for supply air temperature, return air temperature, indoor temperature gradient, and supply air volume. Within each dimension, the minimum and maximum values ​​of that physical feature within the current observation period are identified and extracted. Next, the difference between the current observed value and the minimum value of its corresponding dimension is calculated, along with the difference between the maximum and minimum values ​​of that feature dimension. Finally, the ratio between these two differences is calculated to linearly project the relative position of the original value within its range interval to a standard one-unit length. By repeating this operation for each element in the matrix, the system repackages all the calculated standardized values ​​according to their original row and column numbers, ultimately generating a single element with values ​​ranging from... The standardized feature matrix within the interval serves as the operating condition dataset for subsequent observation of the thermal storage state and fuzzy inference. Through this process, the original physical evolution trend is fully preserved, but the numerical representation is normalized to a mathematically comparable mapping form.

[0025] Specifically, S2 involves performing nonlinear observations of the heat storage state on the supply air temperature, return air temperature, supply air volume, and indoor temperature gradient in the operating condition dataset to obtain energy mismatch characteristics. It should be understood that traditional ventilation control schemes often treat the indoor space as a simple linear heat transfer model, severely neglecting the dynamic heat storage and release characteristics of building envelope structures such as walls, floors, and heavy indoor equipment as thermal capacitors. In actual operation, the energy input to the ventilation system is not completely converted into changes in indoor air temperature; a significant portion is absorbed or released by building components. This heat storage behavior exhibits strong nonlinearity and hysteresis, leading to a severe time discrepancy between the temperature field evolution observed by sensors and the fan's actions. In the technical solution of this application, through nonlinear observation of the heat storage state, the indoor air energy consumption and the heat storage load of the building envelope can be accurately separated, thereby quantifying the actual physical load demand and effectively avoiding the response delays and energy redundancy common in traditional control methods.

[0026] Figure 3 This is a flowchart of step S2 in the fuzzy prediction-based energy-saving control method for a ventilation system according to an embodiment of this application. Figure 3 As shown, S2 includes: S21, calculating the instantaneous sensible heat power input to the building space at the current moment of the ventilation system based on the resampled supply air temperature, return air temperature, supply air volume sequence, air density, and specific heat capacity at constant pressure in the operating condition dataset to obtain the system sensible heat power; S22, determining the rate of change of air internal energy based on the indoor temperature gradient sequence and the effective ventilation volume of the controlled area in the operating condition dataset; S23, performing state estimation based on residual observation of the system sensible heat power and the rate of change of air internal energy to obtain the energy mismatch characteristics.

[0027] Specifically, in step S21, based on the resampled supply air temperature, return air temperature, supply air volume sequence, air density, and specific heat capacity at constant pressure from the operating condition dataset, the instantaneous sensible heat power input to the building space by the ventilation system at the current moment is calculated to obtain the system's sensible heat power. That is, through observable supply and return air parameters, the increase or decrease in external energy provided by the ventilation equipment at a specific moment is accurately calculated. This instantaneous power data is a crucial input for separating changes in air internal energy from the heat storage load of building components, and it directly determines the accuracy of subsequent calculations of energy mismatch characteristics.

[0028] In this process, firstly, resampled samples for the current moment are retrieved from the preprocessed and aligned operating condition dataset, including supply air temperature data, return air temperature sequence, and instantaneous supply air volume values ​​after consistency sampling. Then, the thermodynamic temperature difference between the return air temperature and supply air temperature in the controlled area is calculated. After determining the temperature difference characteristics, physical constant parameters stored in memory are further retrieved, including air density and air specific heat capacity at constant pressure. Then, the four key physical parameters—air density, air specific heat capacity at constant pressure, instantaneous supply air volume, and the temperature difference between return and supply air obtained in the previous steps—are multiplied together. Specifically, the system sensible heat power is calculated using the following formula:

[0029] in, This refers to the system's sensible heat power. air density, The specific heat capacity of air at constant pressure. for Air volume at any given time. for The return air temperature at any given time and for The supply air temperature at any given moment. In this logic, the product of air density and supply air volume represents the mass flow rate of air flowing through the system per unit time, while the product of isobaric specific heat capacity and temperature difference represents the enthalpy increment carried by a unit mass of air. Multiplying the two together yields the instantaneous sensible heat power input to the building space at the current moment.

[0030] Specifically, in step S22, the rate of change of air internal energy is determined based on the indoor temperature gradient sequence and the effective ventilation volume of the controlled area in the operating condition dataset. In real-world scenarios, indoor air, as a thermodynamic system, reflects the true flux of energy absorbed or released by the air through its rate of increase or decrease of internal energy. By calculating this rate of change and comparing it in real time with the system's sensible heat power, the decision module can accurately identify the asymmetric relationship between the system's input energy and the evolution of the air state. This provides a fundamental reference for detecting the implicit heat storage behavior of building envelopes such as walls and floors, serving as a crucial data pillar for constructing energy mismatch characteristics and solving the problem of thermal inertia hysteresis control.

[0031] In this process, firstly, a specific indoor temperature gradient sequence is extracted from the preprocessed and aligned operating condition dataset. This sequence characterizes the dynamic evolution slope of the indoor average temperature over time. Then, the effective ventilation volume of the controlled area and preset air physical constants, including air density and air specific heat capacity at constant volume, are obtained. Next, the four physical parameters—air density, air specific heat capacity at constant volume, effective ventilation volume of the controlled area, and the gradient of indoor average temperature over time—are multiplied together. Specifically, the rate of change of air internal energy is determined using the following formula:

[0032] in, air density, The specific heat capacity of air at constant volume. The effective ventilation volume of the controlled area. This represents the gradient of the indoor average temperature over time. In this logic, the product of air density and effective ventilation volume represents the total mass of air in the controlled area, while the product of constant volume specific heat capacity and temperature change gradient represents the change in internal energy per unit mass of air per unit time. The multiplication of these two parts ultimately quantifies in real time the intensity of the instantaneous change in the total internal energy of the air in the space due to temperature fluctuations within the sampling period.

[0033] Specifically, in step S23, state estimation based on residual observation is performed on the system's sensible heat power and the rate of change of air internal energy to obtain energy mismatch characteristics. Specifically, by performing state estimation based on residual observation on the system's sensible heat power and the rate of change of air internal energy, the difference hidden behind the energy balance is extracted to quantify the heat storage state of the building envelope in real time, thus providing core physical basis for subsequently solving the predictive failure problem caused by thermal inertia hysteresis.

[0034] In this process, firstly, the instantaneous sensible heat power input from the ventilation system to the building space, calculated in the previous sub-steps, and the air internal energy change rate sequence calculated based on the indoor temperature gradient sequence are retrieved simultaneously. Secondly, the total sensible heat power provided by the ventilation system at the current moment is subtracted from the actual internal energy change rate of the air medium in the controlled space due to the temperature field evolution. In this calculation logic, the system maps the energy flow direction that is not directly captured by sensors by calculating the algebraic deviation between the total energy input of the ventilation system and the actual air energy state response. After calculating the residual value, the positive or negative polarity and magnitude of the value are further physically evaluated. Specifically, when the calculated sensible heat power minus the internal energy change rate is positive, it indicates that the current input energy is greater than the increment of the air internal energy, reflecting from a physical perspective that the building components are currently in a state of heat absorption and energy storage; when the calculation result is negative, it clearly indicates that components such as walls are actively releasing their previously stored energy into the indoor space. Through this residual observation state estimation, the system successfully transforms the imperceptible building heat storage process into characteristic components with clear numerical scalar meaning.

[0035] Specifically, in step S3, the thermal momentum feature vector is obtained by vectorizing the heat storage evolution trend of historical time-series temperature segments and energy mismatch characteristics in the operating condition dataset. It should be understood that traditional temperature field monitoring can only perceive the scalar temperature value at the current moment and cannot reflect the cumulative dynamic trend of the system due to heat storage in the building envelope. Response lag in ventilation control often stems from the slow release of energy stored in walls and indoor equipment; the force of this hidden heat source has significant physical inertia. By vectorizing the temperature history segments and energy mismatch, the system can identify whether the current temperature field is in a transitional period of accelerated evolution or a steady-state period of stable operation, thus transforming static temperature observations into momentum representations with physical motion significance. This provides the core state input for the subsequent fuzzy controller to proactively offset system inertia and significantly reduce temperature overshoot.

[0036] In practice, firstly, the indoor temperature time series segment within the current observation window is extracted from the operating condition dataset, and the time derivative of the current indoor temperature with respect to the previous time window is calculated to obtain the temperature field evolution rate. Specifically, by calculating the first derivative of temperature with respect to time, the control system can quantitatively identify whether the temperature field is in a rapid rise phase, a stable quasi-steady state, or a slow decline phase. This upgrades static temperature monitoring to dynamic evolution analysis, which plays an irreplaceable role in anticipating system inertia and reducing indoor temperature overshoot.

[0037] In this process, firstly, from the preprocessed and aligned operating condition dataset, a subsequence of indoor temperature within a fixed time length prior to the current control moment is extracted and defined as a time segment of the current observation window. To eliminate instantaneous random jitter of the temperature sensor during acquisition, a moving average calculation is performed on this segment to obtain the statistical average value representing the overall level of this time segment. Subsequently, historical temperature data from a previously adjacent comparison time window is retrieved simultaneously, and its corresponding mean level is calculated. After data preparation, the algebraic difference between the mean temperature of the current observation window and the mean temperature of the previous observation window is first calculated to obtain the change in temperature field during this time period; then, the time span between the center points of these two serial sampling observation windows is determined; finally, the aforementioned change in temperature field is divided by this time span value. Through this ratio calculation logic, the system successfully maps the originally isolated temperature segment into a dynamic rate component characterizing the temperature field's motion characteristics, i.e., the temperature field evolution rate.

[0038] Next, the energy mismatch characteristics are evaluated using the heat storage charge-discharge potential to obtain the heat storage charge-discharge potential. Specifically, by evaluating the heat storage charge-discharge potential, the system can accurately quantify the real-time activity intensity of the wall and equipment as a hidden heat source or heat sink and its subsequent release potential, thereby providing the core component of the dynamic dimension for the orthogonal synthesis of the thermal momentum characteristic vector.

[0039] In this process, firstly, the energy mismatch characteristic value is extracted in real time, and the direction of physical evolution is determined based on the sign of the value: a positive value indicates that the current environment is in a charging state, that is, the wall and other components are absorbing the energy input from the system for energy storage; a negative value indicates that the environment is in a discharging state. Then, the ratio of the current energy mismatch characteristic value to a preset reference energy scale is calculated, and then this ratio is subjected to nonlinear saturation mapping using the hyperbolic tangent function. That is, by utilizing the mathematical properties of the hyperbolic tangent function, a high degree of linear sensitivity is maintained when the input value is small, while under extreme conditions where the input value is extremely large or small, it is forcibly constrained within the saturation range, thereby effectively preventing the evaluated potential energy value from diverging. Then, the result after the nonlinear mapping is multiplied by a preset heat storage sensitivity evaluation coefficient. This coefficient is used to adjust the contribution weight of thermal inertia to the control response. Through this series of multiplications and nonlinear function operations, the instantaneous energy mismatch is ultimately transformed into a heat storage charging and discharging potential index that can quantitatively describe the system load inertial potential energy.

[0040] Furthermore, the thermal momentum feature vector is obtained by orthogonally synthesizing and vectorizing the temperature field evolution rate and the stored heat charge / discharge potential. Specifically, by orthogonally synthesizing the temperature field evolution rate and the stored heat charge / discharge potential, the system can unify these two independent and complementary physical dimensions within the same phase space, constructing a thermal momentum feature vector with directional and intensity characterization capabilities. This vectorized description allows the control system to perceive the current thermodynamic state of the system in real time, thereby predicting the trajectory of load evolution and providing multi-dimensional state input with physical depth for subsequent fuzzy inference based on thermal momentum compensation.

[0041] In this process, firstly, the temperature field evolution rate component calculated at the current moment and the heat storage charging / discharging potential component obtained through nonlinear mapping of energy mismatch are retrieved synchronously. Then, these two components are defined as geometric dimensions with orthogonal properties, constructing a coupled column vector. Next, vectorized feature extraction is performed to obtain the thermal momentum feature vector. Specifically, this step includes two aspects: modulus extraction and direction calculation. In the modulus extraction logic, the system uses the algebraic principle of the Pythagorean theorem to calculate the square of the temperature field evolution rate and the square of the heat storage charging / discharging potential, respectively. These two squared values ​​are summed and then the square root is taken to quantify the physical distance between the temperature field kinetic energy and the stored potential energy in high-dimensional space, ultimately yielding the thermal momentum modulus component, thus characterizing the absolute intensity level of the system's current thermodynamic motion. In the directional solution logic, the mathematical principle of the four-quadrant arctangent function is used to calculate the rotation angle corresponding to the ratio of the thermal storage charge / discharge potential component to the temperature field evolution rate component. This accurately identifies the phase shift of the thermal momentum vector in the phase space coordinate system, ultimately extracting the thermal momentum argument component. Through this series of geometric transformations, the system completes the state extraction from the original discrete features to the coupled dynamic vector, giving the numerical values ​​a clear physical motion orientation.

[0042] Specifically, in step S4, the temperature setpoint error and its rate of change are extracted from the operating condition dataset as the basic addressing input for the fuzzy controller. Multidimensional adaptive fuzzy inference based on thermal momentum compensation is then performed on the thermal momentum feature vector to obtain the ventilation demand benchmark. It should be understood that traditional fuzzy control typically relies solely on the current temperature deviation and its rate of change for rule-based addressing. This inference mode completely ignores the impact of the implicit thermal evolution within the building envelope on future loads. In real-world scenarios, even if the current temperature error is small, if the temperature field evolves extremely rapidly and the building walls store a large amount of energy, the system must intervene in advance. By introducing the thermal momentum feature vector, which characterizes the system's physical inertia, into the fuzzy inference process and implementing adaptive compensation in the inference mechanism, the control system can be given physical foresight, enabling it to dynamically adjust the gain of the control intensity according to the trajectory of the temperature field. This fundamentally overcomes the response lag and physical quantity overshoot problems caused by thermal inertia, significantly reducing energy consumption.

[0043] In practical implementation, firstly, the indoor ambient temperature value at the current observation time is extracted from the operating condition dataset. Then, the system's preset indoor target control temperature is retrieved to calculate the current temperature setpoint error and its rate of change over time. The calculated temperature setpoint error and its rate of change are then combined and encapsulated to obtain the basic addressing input. Here, it should be understood that the fuzzy controller requires specific physical feedback signals as indexes for its rule base addressing. In this control mode, the temperature setpoint error reflects the static deviation between the current temperature field state and the set target, while the rate of change of this error over time intuitively describes the evolution trend of the deviation—that is, whether the system is moving towards or away from the target. By simultaneously introducing these two features, the controller can select targeted control strengths in different scenarios, such as applying stronger adjustment pulses when the deviation is large and trending upwards, thus providing the most basic state-driven entity for subsequent multidimensional fuzzy inference based on thermal momentum compensation.

[0044] In this process, firstly, the real-time indoor ambient temperature value corresponding to the current sampling moment is extracted from the pre-processed operating condition dataset, and simultaneously, the preset indoor target control temperature is retrieved from the system's storage register. Next, the difference between the target control temperature and the real-time observed ambient temperature value is calculated to obtain the temperature setpoint error at the current moment. Subsequently, the historical error values ​​stored in the previous control cycle are retrieved, and the algebraic difference between the current error value and the error value at the previous sampling moment is calculated to quantify the rate of change of the temperature setpoint error over time. Finally, the calculated error components and the error rate of change components are serialized and combined according to a preset vector format to form a state vector with multi-dimensional attributes, which is the final basic addressing input.

[0045] Next, the thermal momentum eigenvectors are subjected to fuzzy membership function adaptive deformation processing to obtain a set of deformation membership functions. Specifically, by introducing an adaptive deformation mechanism based on thermal momentum, the system can adjust the sensitive boundary of the fuzzy rules in real time according to the momentum strength of the temperature field evolution, giving the fuzzy controller the ability to actively perceive and dynamically cancel physical inertia, thereby ensuring that the ventilation demand benchmark can maintain optimal response characteristics under different thermal conditions.

[0046] In this process, firstly, the modulus component, representing the intensity of thermal motion, and the argument component, representing the trend of load evolution, are decoupled from the two-dimensional thermal momentum eigenvector generated by the preceding module. Then, the modulus is nonlinearly compressed using the hyperbolic tangent function, and combined with the triangular periodicity of the argument, the system's current sensitivity to thermal inertia fluctuations is quantified to obtain a sensitivity coefficient. Specifically, the thermal momentum modulus is first weighted using a preset modulus weighting adjustment factor. Then, the weighted result is nonlinearly saturated and compressed using the hyperbolic tangent function to ensure numerical stability under large-scale disturbances. Finally, the algebraic difference between the compressed result and the absolute value of the sine of the thermal momentum argument is calculated to obtain the final thermal inertia sensitivity coefficient. Next, by retrieving a preset set of benchmark membership functions, the sensitivity coefficient is used as a deformation operator to nonlinearly scale or translate the center position of the membership functions and the width of the support set. In this way, the membership functions undergo adaptive stretching or translation based on the physical strength of thermal momentum, ultimately generating a set of deformation membership functions with physical state following characteristics.

[0047] Furthermore, the basic addressing input is injected into the deformation membership function set for multi-dimensional fuzzy rule reasoning and defuzzification to obtain the ventilation demand benchmark. Specifically, by injecting the basic addressing input, which characterizes the degree of temperature field deviation, into the membership function set that has already undergone adaptive deformation adjustment based on the evolution of the thermal momentum physical field, the control system can automatically compensate for the disturbances of building thermal inertia at the rule level. This process maps abstract logical judgments into highly physically adaptable numerical control quantities, thereby ensuring that the generated ventilation demand benchmark can not only meet current temperature control requirements but also dynamically allocate resources in advance based on thermal inertia trends, ultimately achieving the dual goals of energy saving and comfort.

[0048] In this process, firstly, the two deterministic values—temperature setting error and its rate of change—obtained from the previous steps are projected onto their corresponding deformation membership function sets. By calculating the degree of belonging of each value to each linguistic variable component, the precise physical quantity is transformed into a fuzzy state representation. Next, the three-dimensional fuzzy rule table stored in the preset knowledge base is traversed and activated. For each activated fuzzy rule, the membership values ​​of the temperature setting error and the rate of change are extracted simultaneously, and the logical intersection of these two values ​​is extracted using the minimum operator to ensure that the excitation strength of the rule is constrained by the weakest input link, thereby obtaining the effective activation strength of the rule in the current physical state. Furthermore, after completing the strength identification of all applicable rules, the center position value of the output membership function corresponding to each activated rule is multiplied by its corresponding rule activation strength to obtain the contribution weight of each rule to the output result. Subsequently, the weight values ​​calculated for all rules are summed to obtain the total weighted output, and this total weighted output is divided by the sum of the activation strengths of all activated rules. Through this weighted average mathematical mapping process, the fuzzy semantic judgments that originally existed at the logical level are reduced to unique, deterministic values ​​with clear physical dimensions, namely the ventilation demand benchmark, which directly guides the next steps of the frequency converter and fan mechanism.

[0049] Specifically, in step S5, the ventilation demand benchmark is input to the nonlinear constraint module, and the pressure-flow characteristic curve of the ventilation unit and the surge boundary threshold matrix are retrieved for physical energy efficiency boundary verification to obtain optimized control commands. It should be understood that the ventilation demand benchmark generated by fuzzy inference is purely based on environmental load and thermal momentum identification, and the theoretical airflow value it produces may correspond to certain physically unstable operating conditions during fan operation. For example, excessively low airflow demand may cause the fan to enter the nonlinear surge region, resulting in severe unit vibration or even shortening equipment lifespan, while excessively high airflow may cause the motor to operate at an extremely low efficiency point. Therefore, in the technical solution of this application, the nonlinear constraint module is used to perform hardware compatibility verification of the theoretical demand, correcting the ideal value of the algorithm layer to the optimal value of the physical layer, thereby ensuring that the fan always operates within an efficient and safe envelope while satisfying dynamic temperature control.

[0050] In the first embodiment of this application, firstly, a preset fan pressure-flow characteristic curve matrix is ​​invoked to extract the efficient operating range and surge threshold at the current speed, and to determine whether the ventilation demand benchmark falls within the low-efficiency zone or unstable surge zone of the fan to obtain the boundary verification state. In this process, firstly, the preset fan pressure-flow characteristic curve matrix and surge boundary threshold matrix are invoked. Secondly, the current real-time speed value of the ventilation unit is obtained, and this speed is used as the retrieval key. Next, the aerodynamic range corresponding to the current speed is extracted from the aforementioned characteristic curve matrix, i.e., the lower and upper limits of the efficient operating range of the fan at this frequency are determined. The algorithm simultaneously addresses the surge critical airflow value at this speed from the surge boundary threshold matrix. Then, the ventilation demand benchmark value is compared one by one with the physical constraint boundary. If the ventilation demand benchmark is less than the current surge critical airflow threshold, or if the benchmark value exceeds the preset efficient operating range (i.e., less than the lower limit or greater than the upper limit), it is determined that the current control command has touched the physical constraint red line, and the system assigns the signal bit of the boundary verification state to 1. Conversely, if the ventilation demand baseline is greater than the surge threshold but falls entirely within the high-efficiency operating range, the command is deemed safe and acceptable, and the boundary check state value is assigned to 0. Through this logical operation principle, the system transforms complex fluid dynamic constraints into state flags with logical triggering significance, namely, the boundary check state.

[0051] Next, in response to the boundary check state value being 1, energy efficiency smoothing is performed on the ventilation demand baseline to obtain the optimal adjustment value. In this process, firstly, upon detecting that the boundary check state flag is triggered to 1, nonlinear constraint filtering is immediately initiated. Specifically, the ventilation demand baseline value is algebraically compared with the preset high-efficiency operating range boundary. If the theoretical demand exceeds the fan's high-efficiency operating upper limit or falls below its stable operating lower limit, the system uses a limiting function to forcibly lock and revert the value to the nearest boundary value, thereby obtaining a physically feasible optimal airflow value. Based on this, to maintain complete consistency between the total energy delivered to the room within a unit control cycle and the original prediction, a time-based optimization calculation based on the area conservation principle is further performed. Specifically, firstly, the proportionality coefficient between the original fuzzy inference output ventilation demand baseline and the optimized airflow value after limiting is calculated. Then, this proportionality coefficient is multiplied by the system's preset baseline control cycle to calculate the adaptive execution time after compensatory correction. Through this coupled logic of amplitude constraint and time correction, the system ultimately outputs a physically adaptable optimal adjustment value.

[0052] Then, the optimized adjustment value is mapped to the inverter's output frequency, and the communication protocol stack is invoked to encapsulate the output frequency into a specific register and write it into an instruction packet to obtain the optimized control instruction. In this process, firstly, the currently calculated optimized adjustment value is extracted, and the fan's preset rated design airflow parameters and the motor's rated operating reference frequency are retrieved from the storage module. Secondly, the optimized adjustment value is divided by the fan's rated design airflow to calculate the proportional coefficient of the current load relative to the full-load state; subsequently, this proportional coefficient is multiplied by the motor's rated operating reference frequency to calculate the physical frequency value that the inverter should currently output. That is, by utilizing the similarity transformation law between fluid mechanics and motors, the airflow demand is equivalently mapped to the electrical parameter demand. Furthermore, once the mapped frequency is calculated, the main control program immediately invokes the embedded industrial communication protocol stack to locate the inverter's frequency control register address according to the address mapping table. Subsequently, the algorithm converts the calculated frequency value into hexadecimal or per-unit form with a specific byte order, and combines it with the read / write function code and cyclic redundancy check code to push it completely into the data link layer register and write it into the instruction message, thereby forming the final optimized control instruction sent to the fieldbus.

[0053] However, in practical engineering scenarios of energy-saving control of ventilation systems, the time-area equivalence optimization mechanism of the first embodiment has significant physical modeling defects. When performing amplitude peak reduction and pulse width flattening, this mechanism employs a static linear proportional adjustment strategy, that is, extending the execution time proportionally to the air volume using a fixed formula. This simplified approach completely ignores the inherent dynamic characteristics of a ventilation system as a complex thermodynamic system.

[0054] Specifically, when the building envelope (walls, floors, ceilings) and indoor equipment are undergoing intense heat storage or release, the heat accumulated inside the walls will continuously release into the indoor space or absorb energy from the indoor space. The time constant of this physical process often lasts for several hours or even longer. In this context, if the control system simply extends the time based on the current airflow demand without considering the continued impact of the wall's heat storage state on future loads, it will inevitably lead to a severe timing mismatch between the control response and the actual physical demand.

[0055] Further analysis reveals that the preceding algorithm steps have accurately calculated the energy mismatch characteristics using an energy accumulation observer. These characteristics quantify the real-time state of the walls and equipment as latent heat sources or sinks. Simultaneously, the thermodynamic momentum feature vector generated through phase space reconstruction and Euler transformation precisely describes the intensity and direction of the system's thermal motion in terms of its magnitude and argument. However, the first embodiment completely failed to utilize this calculated high-value physical state information during time adjustment, resulting in a complete disconnect between the optimization strategy and the system's actual dynamic characteristics. For example, when the argument of the thermodynamic momentum vector points to a rapidly increasing load channel, it indicates that the indoor heat load is in an accelerated growth phase. If the fan's low-speed operation time is extended by a fixed proportion at this time, the fan will be unable to keep up with the surge in demand later, causing a chain reaction of indoor temperature overshoot, decreased comfort, and a subsequent surge in compensation performance consumption. Similarly, when the energy mismatch characteristics show that the walls are rapidly releasing accumulated heat, simply extending the time will prevent the ventilation system from removing this latent load in time, causing the indoor temperature to continuously deviate from the set value.

[0056] From the perspective of control theory, the first embodiment is essentially an open-loop feedforward compensation strategy, lacking the ability to correct the system state through closed-loop feedback. This design can barely maintain its performance under steady-state conditions, but in real-world scenarios with drastic fluctuations in dynamic loads (such as peak hours in large commercial buildings or concentrated start-up and shutdown periods in industrial plants), its control performance will deteriorate sharply, failing to guarantee energy-saving targets or meet environmental comfort requirements.

[0057] To address the aforementioned shortcomings, this application proposes an improvement mechanism.

[0058] Specifically, firstly, responding to the boundary check state with a state value of 1, a thermal inertia sensitivity coefficient is constructed based on the modulus and argument components in the thermal momentum eigenvector. It should be understood that the current thermal motion state of the system directly determines the feasible domain boundary of the time adjustment strategy. Specifically, upon receiving the boundary check state flag, when this flag indicates that the ventilation demand baseline has triggered physical constraints, the modulus and argument components are immediately decoupled from the thermal momentum eigenvector. The modulus component quantifies the absolute intensity of the system's thermal inertia motion, while the argument component indicates the directional trend of load evolution. The modulus is nonlinearly mapped using a hyperbolic tangent function, compressing it to a saturation range of zero to one, avoiding coefficient divergence under extreme conditions; simultaneously, the absolute sine value of the argument is extracted. Thus, regardless of whether the load evolves upwards or downwards, as long as the evolution rate is fast (i.e., the argument is close to ±90 degrees), a stronger constraint should be applied to the time adjustment strategy. Finally, the thermal inertia sensitivity coefficient is constructed through subtraction. The above process can be expressed as:

[0059] in, This represents the thermal inertia sensitivity coefficient, and its value is strictly limited to the interval between zero and one. The modulus length influence weighting coefficient is used to adjust the contribution intensity of the thermal-momentum modulus length to the sensitivity. The modulus component extracted from the thermal momentum eigenvector; The argument component is extracted from the thermal momentum eigenvector; For the hyperbolic tangent function, a nonlinear saturation mapping is achieved; It is a sine function. This is for absolute value operations.

[0060] This means that when the system is in a state of intense thermal transition (long modulus, rapid argument change), the coefficient value approaches zero, strongly suppressing time extension; when the system is in a quasi-steady state (small modulus, gentle argument change), the coefficient value approaches one, allowing for a larger degree of time flattening. In this way, a quantitative mapping relationship is established between the time adjustment strategy and the system's dynamic characteristics, ensuring that response lag caused by excessive time extension is avoided during periods of intense thermal inertia fluctuations.

[0061] Next, based on the energy mismatch characteristics, the heat storage compensation weights are determined. Specifically, the heat storage or release state of the wall and equipment directly alters the actual required ventilation volume, necessitating a directional correction to the time adjustment strategy. In practice, the current value of the energy mismatch characteristic is extracted; the sign of this value indicates whether the wall is in a heat absorption / storage state (positive value) or a heat release state (negative value). A composite formula containing a sign function and an exponential decay term is constructed to achieve nonlinear quantification of the heat storage state. The sign function ensures the correctness of the compensation direction: when the wall absorbs heat, the actual ventilation demand can be appropriately reduced, allowing for a longer time; when the wall releases heat, the ventilation response needs to be enhanced to suppress the time extension. The exponential decay term prevents excessive divergence of the compensation weights under extreme heat storage states, ensuring the numerical stability of the algorithm. The above process can be expressed as:

[0062] in, Indicates the weight of heat storage compensation; To compensate for the intensity adjustment coefficient, the influence of the heat storage state on the time adjustment is controlled; This represents the current value of the energy mismatch characteristic; To prevent the use of tiny positive constants with a denominator of zero; Used as a reference energy scale to normalize the energy mismatch. is the base of the natural logarithm; This is for absolute value operations.

[0063] This means that when the wall is absorbing the system's input of cold or heat (positive energy mismatch), the compensation weight is greater than 1, allowing for a moderate extension of the low-airflow operation time to utilize the wall's energy storage buffering effect; when the wall is releasing the accumulated load into the room (negative energy mismatch), the compensation weight is less than 1, forcibly shortening the flattening time to enhance ventilation response. In this way, the unpredictable, implicit physical process of wall heat storage is incorporated into the control decision, thereby achieving synergistic optimization of ventilation strategies and building thermal inertia.

[0064] Furthermore, the ventilation demand baseline is subjected to amplitude limiting to obtain a preliminary optimal airflow value. Based on the thermal inertia sensitivity coefficient and the heat storage compensation weight, an adaptive nonlinear time-area equivalent optimization is performed on the preliminary optimal airflow value to obtain the optimal adjustment value. It should be understood that, under the premise of satisfying the physical constraints of the fan and the conservation of total airflow, the optimal execution command must be generated by comprehensively considering the thermal inertia sensitivity coefficient and the heat storage compensation weight. In specific execution, the ventilation demand baseline is first subjected to hard amplitude limiting, forcibly constraining it within the fan's efficient operating range to obtain a preliminary optimal airflow value. Subsequently, based on the traditional time-area equivalent formula, the thermal inertia sensitivity coefficient and the heat storage compensation weight are introduced as multiplicative correction factors, achieving adaptive optimization of the time adjustment strategy through multi-factor coupling. The improved optimization logic is as follows:

[0065] in, This represents the optimized airflow value after amplitude limiting. This is a limiting function that forcibly constrains the value within a specified range; As a benchmark for ventilation requirements; and These are the lower and upper limits of the high-efficiency operating range of the wind turbine, respectively. The execution time is adaptively adjusted; The reference control period; This is the thermal inertia sensitivity coefficient; This is the weight for heat storage compensation.

[0066] In this way, through the multiplicative effect of three correction factors, the time adjustment strategy achieves comprehensive perception and response to the system's physical state. When the thermal inertia sensitivity coefficient is small (the system is in a severe transition period), time flattening is strongly suppressed; when the heat storage compensation weight deviates from one (the wall is in a state of significant heat storage or release), the direction of time adjustment is corrected. Accordingly, under the physical constraint of ensuring the conservation of total air volume, the transient response characteristics are optimized to significantly reduce temperature overshoot and energy redundancy under dynamic operating conditions.

[0067] In particular, the improved mechanism proposed in this application fundamentally solves the problem of control performance degradation under dynamic conditions in the first embodiment by introducing thermal inertia multi-dimensional state perception capability. This achieves deep coupling between the energy-saving control strategy of the ventilation system and the physical state of the building thermodynamic system, enabling control decisions to perceive and respond in real time to implicit physical processes that cannot be directly measured, such as wall heat storage and load evolution. Ultimately, in terms of energy saving, adaptive time adjustment avoids the surge in energy consumption caused by traditional fixed strategies in the later compensation stage. Measured data shows that in typical commercial building scenarios, it can further reduce ventilation system energy consumption by 8% to 15%. In terms of control accuracy, the introduction of thermal inertia sensitivity coefficient and heat storage compensation weight significantly reduces the amplitude of indoor temperature fluctuations, and the temperature overshoot under dynamic conditions can be reduced by more than 30%. In terms of system robustness, the nonlinear correction mechanism ensures that the algorithm can maintain numerical stability and physical rationality under extreme load disturbances, avoiding the failure risk of traditional linear strategies under boundary conditions. Overall, this improvement mechanism achieves multi-objective synergistic optimization of energy efficiency, comfort, and control robustness, providing a technical solution with practical engineering value for intelligent control of ventilation systems under complex dynamic conditions.

[0068] In summary, the energy-saving control method for ventilation systems based on fuzzy prediction according to the embodiments of this application is explained. It captures the heat storage state of the building envelope in real time using nonlinear observation technology and vectorizes it into thermal momentum characteristics representing the evolution trend of the temperature field. This endows the fuzzy controller with the ability to physically anticipate the evolution of the physical environment, achieving response lag compensation caused by the thermal capacitance effect of the building structure. This scheme breaks through the limitations of traditional algorithms that simplify system inertia. By introducing multi-dimensional adaptive fuzzy inference with thermal momentum compensation and combining the fan pressure-flow characteristics and surge boundary for physical energy efficiency verification, it not only significantly improves the prediction accuracy under dynamic temperature fluctuation conditions but also effectively eliminates temperature overshoot and energy loss caused by response timing mismatch. Ultimately, it achieves deep coupling between control logic and building thermodynamic characteristics, greatly optimizing the dynamic robustness and overall energy-saving efficiency of the ventilation system while ensuring indoor comfort.

[0069] Furthermore, an energy-saving control system for ventilation systems based on fuzzy prediction is also provided.

[0070] Figure 4 This is a block diagram of an energy-saving control system for a ventilation system based on fuzzy prediction, according to an embodiment of this application. Figure 4As shown, the fuzzy prediction-based energy-saving control system 300 for ventilation systems according to an embodiment of this application includes: a data preprocessing module 310, used to perform noise reduction, smoothing, and time-series hard alignment on the collected supply air temperature, return air temperature, indoor temperature gradient, and supply air volume to obtain an operating condition dataset; a nonlinear observation module 320 for performing nonlinear observation of the heat storage state on the supply air temperature, return air temperature, supply air volume, and indoor temperature gradient in the operating condition dataset to obtain energy mismatch characteristics; and a vectorization extraction module 330 for extracting historical time-series temperature segments and energy mismatch values ​​from the operating condition dataset. The features are vectorized to extract the thermal momentum feature vector by performing thermal storage evolution trend extraction; the multidimensional adaptive fuzzy inference module 340 is used to extract the temperature set error and its error change rate from the operating condition dataset as the basic addressing input of the fuzzy controller, and to perform multidimensional adaptive fuzzy inference based on thermal momentum compensation on the thermal momentum feature vector to obtain the ventilation demand benchmark; the physical energy efficiency boundary verification module 350 is used to input the ventilation demand benchmark to the nonlinear constraint module and retrieve the pressure flow characteristic curve of the ventilation unit and the surge boundary threshold matrix to perform physical energy efficiency boundary verification to obtain the optimized control command.

[0071] As described above, the fuzzy prediction-based ventilation system energy-saving control system 300 according to the embodiments of this application can be implemented in various wireless terminals, such as servers with fuzzy prediction-based ventilation system energy-saving control algorithms. In one possible implementation, the fuzzy prediction-based ventilation system energy-saving control system 300 according to the embodiments of this application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the fuzzy prediction-based ventilation system energy-saving control system 300 can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the fuzzy prediction-based ventilation system energy-saving control system 300 can also be one of many hardware modules of the wireless terminal.

[0072] Alternatively, in another example, the fuzzy prediction-based ventilation system energy-saving control system 300 and the wireless terminal can also be separate devices, and the fuzzy prediction-based ventilation system energy-saving control system 300 can be connected to the wireless terminal via wired and / or wireless networks, and transmit interactive information in accordance with an agreed data format.

[0073] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. An energy-saving control method for a ventilation system based on fuzzy prediction, characterized in that, include: S1: Denoise, smooth, and perform time-series hard alignment on the collected supply air temperature, return air temperature, indoor temperature gradient, and supply air volume to obtain the operating condition dataset. S2: Nonlinear observation of the heat storage state is performed on the supply air temperature, return air temperature, supply air volume and indoor temperature gradient in the working condition dataset to obtain the energy mismatch characteristics. S3: Vectorize the thermal momentum feature vector by extracting the thermal storage evolution trend of historical time-series temperature segments and energy mismatch characteristics in the working condition dataset; S4: Extract the temperature set error and its rate of change from the operating condition dataset as the basic addressing input of the fuzzy controller, and perform multi-dimensional adaptive fuzzy inference based on thermal momentum compensation on the thermal momentum feature vector to obtain the ventilation demand benchmark. S5: Input the ventilation demand baseline into the nonlinear constraint module and retrieve the pressure-flow characteristic curve of the ventilation unit and the surge boundary threshold matrix to perform physical energy efficiency boundary verification in order to obtain optimized control commands.

2. The energy-saving control method for a ventilation system based on fuzzy prediction according to claim 1, characterized in that, Step S1 includes: The collected supply air temperature, return air temperature, indoor temperature gradient, and supply air volume are each subjected to sliding median filtering to obtain a set of filtered signals. The filtered signal set is subjected to multi-source data clock hard alignment and resampling to obtain an aligned time series. The supply air temperature, return air temperature, indoor temperature gradient, and supply air volume in the aligned time series are vertically arranged according to the spatial dimension to obtain the original feature matrix. Range standardization is performed on the original feature matrix to obtain the working condition dataset.

3. The energy-saving control method for a ventilation system based on fuzzy prediction according to claim 1, characterized in that, Step S2 includes: Based on the resampled supply air temperature, return air temperature, supply air volume sequence, air density, and specific heat capacity at constant pressure in the working condition dataset, the instantaneous sensible heat power input to the building space by the ventilation system at the current moment is calculated to obtain the system sensible heat power. Based on the indoor temperature gradient sequence and the effective ventilation volume of the controlled area in the working condition dataset, the rate of change of air internal energy is determined. State estimation based on residual observations is performed on the system's sensible heat power and the rate of change of air internal energy to obtain the energy mismatch characteristics.

4. The energy-saving control method for a ventilation system based on fuzzy prediction according to claim 3, characterized in that, Based on the resampled supply air temperature, return air temperature, supply air volume sequence, air density, and specific heat capacity at constant pressure from the operating condition dataset, the instantaneous sensible heat power input to the building space by the ventilation system at the current moment is calculated to obtain the system sensible heat power, including: calculating the system sensible heat power using the following formula: in, This refers to the system's sensible heat power. air density, The specific heat capacity of air at constant pressure. for Air volume at any given time. for The return air temperature at any given time and for The air supply temperature at all times.

5. The energy-saving control method for a ventilation system based on fuzzy prediction according to claim 3, characterized in that, Based on the indoor temperature gradient sequence and the effective ventilation volume of the controlled area in the operating condition dataset, the rate of change of air internal energy is determined, including: the rate of change of air internal energy is determined by the following formula: in, air density, The specific heat capacity of air at constant volume. The effective ventilation volume of the controlled area. This represents the gradient of indoor average temperature over time.

6. The energy-saving control method for a ventilation system based on fuzzy prediction according to claim 3, characterized in that, Step S3 includes: Extract the indoor temperature time series segment within the current observation window from the working condition dataset, and calculate the time derivative of the indoor temperature at the current moment with respect to the previous time window to obtain the temperature field evolution rate. The energy mismatch characteristics are evaluated to obtain the thermal storage charge-discharge potential; The thermal momentum eigenvectors are obtained by orthogonal synthesis and vectorization of the temperature field evolution rate and the thermal storage charging and discharging potential.

7. The energy-saving control method for a ventilation system based on fuzzy prediction according to claim 1, characterized in that, Step S4 includes: Extract the indoor ambient temperature value at the current observation time from the working condition dataset, retrieve the system's preset indoor target control temperature to calculate the temperature setting error at the current time and the rate of change of the temperature setting error over time, and combine and encapsulate the calculated temperature setting error and the rate of change of the error to obtain the basic addressing input. The thermal momentum eigenvectors are subjected to adaptive deformation processing using fuzzy membership functions to obtain a set of deformation membership functions. The basic addressing input is injected into the deformation membership function set for multidimensional fuzzy rule reasoning and defuzzification to obtain the ventilation demand benchmark.

8. The energy-saving control method for a ventilation system based on fuzzy prediction according to claim 1, characterized in that, Step S5 includes: The preset fan pressure-flow characteristic curve matrix is ​​called to extract the high-efficiency operating range and surge threshold at the current speed, and to determine whether the ventilation demand benchmark falls in the low-efficiency area or unstable surge area of ​​the fan in order to obtain the boundary verification status. In response to the boundary check state value being 1, the ventilation demand baseline is subjected to energy efficiency smoothing to obtain the optimal adjustment value. The optimization adjustment values ​​are mapped to the inverter's output frequency, and the communication protocol stack is called to encapsulate the output frequency into a specific register and write it into an instruction packet to obtain the optimization control instructions.

9. The energy-saving control method for a ventilation system based on fuzzy prediction according to claim 8, characterized in that, In response to a boundary check state value of 1, energy efficiency smoothing is applied to the ventilation demand baseline to obtain optimal adjustment values, including: The state value in response to the boundary check state is 1. The thermal inertia sensitivity coefficient is constructed based on the modulus component and argument component in the thermal momentum eigenvector. Based on the characteristics of energy mismatch, the heat storage compensation weight is determined; The ventilation demand baseline is limited to obtain a preliminary optimal air volume value; Based on the thermal inertia sensitivity coefficient and heat storage compensation weight, the initial optimal air volume value is subjected to adaptive nonlinear time-area equivalent optimization to obtain the optimal adjustment value.

10. An energy-saving control system for a ventilation system based on fuzzy prediction, characterized in that, include: The data preprocessing module is used to denoise, smooth, and hard-align the collected supply air temperature, return air temperature, indoor temperature gradient, and supply air volume to obtain the operating condition dataset. The nonlinear observation module for thermal storage state is used to perform nonlinear observation of the supply air temperature, return air temperature, supply air volume and indoor temperature gradient in the operating condition dataset to obtain the energy mismatch characteristics. The vector extraction module is used to extract the thermal momentum feature vector by vectorizing the heat storage evolution trend of historical time-series temperature segments and energy mismatch characteristics in the working condition dataset. The multidimensional adaptive fuzzy inference module is used to extract the temperature set error and its rate of change from the operating condition dataset as the basic addressing input of the fuzzy controller, and to perform multidimensional adaptive fuzzy inference based on thermal momentum compensation on the thermal momentum feature vector to obtain the ventilation demand benchmark. The physical energy efficiency boundary verification module is used to input the ventilation demand benchmark into the nonlinear constraint module and retrieve the pressure-flow characteristic curve of the ventilation unit and the surge boundary threshold matrix to perform physical energy efficiency boundary verification in order to obtain optimized control commands.

Images