Adaptive finite-time neural network fan coil outlet air temperature control method and verification method

Abstract

The present invention belongs to the field of electric power, and discloses an adaptive finite-time neural network fan coil outlet air temperature control method and verification method, including: step 1, constructing a mathematical model of a flow control valve, a fan coil and a temperature sensor; step 2, using a neural network approximation to design a finite-time controller; step 3, controlling the fan coil outlet air temperature through a finite-time controller. Compared with asymptotically stable control, the present invention systematically proposes a finite-time neural controller, so that the temperature regulation error converges to a small area around the origin within a finite time; the present invention introduces a correction element to compensate the first-order damping element in the thermistor to capture the temperature in real time.

Description

Technical Field

[0001] The present invention relates to the electric power field, and in particular to an adaptive finite-time neural network fan coil outlet air temperature control method and a verification method. Background Art

[0002] With the development of demand response technology, the widespread use of time-of-use electricity prices and even real-time electricity prices has become an inevitable trend. On the other hand, the continuous improvement of smart distribution network technology and the widespread application of intelligent management and control equipment have provided conditions for the two-way interaction between the power supply side and the demand side, making it possible to intelligently control power equipment on the user side to actively respond to grid demand. Users can obtain the reference control target of the load under the real-time electricity price according to the load optimization index, intelligently control the power load, actively respond to grid demand, and reduce electricity costs at the same time. Considering the time-varying nature of electricity costs under real-time electricity prices, in order to reduce electricity costs while taking into account user needs, the reference index of load control is often time-varying. In this context, traditional asymptotically stable control methods are difficult to quickly track load control targets and cannot reduce electricity costs while taking into account user needs. Unlike asymptotically stable control, finite-time control can make the control target approach the equilibrium point within a limited time, so that load control can respond to time-varying electricity prices while meeting user needs, and achieve a simultaneous reduction in electricity costs. As one of the important demand-side loads in the power system, air conditioning load accounts for more than 30% during the peak period in summer, and it shows an increasing trend year by year. Research on the optimization control of air conditioning load can not only effectively cope with the peak load challenges faced by the power system and alleviate the imbalance between supply and demand, but also improve the operating efficiency and stability of the power system. Therefore, the present invention studies the control problem of power load under demand response and solves the problem of adaptive finite time control of power load under real-time electricity price.

[0003] In the air conditioning load optimization control, the air source heat pump air conditioning (ASHPAC) system uses the solar radiation energy contained in the air as a cooling or heating source. It follows the reversible Carnot cycle when working, and drives the compressor to achieve energy transfer through a small amount of electrical energy. Compared with the general HVAC system, its performance coefficient is higher. As the core component of the indoor loop of the ASHPAC system, the performance of the fan coil unit determines the comfort of the indoor environment. The existing ASHPAC system temperature control methods are mostly PID, robust control, MPC control, etc., which can only guarantee the asymptotic convergence of the temperature adjustment error, that is, the convergence time of the temperature adjustment error is infinite. In order to create a constant and comfortable indoor environment, it is of great significance to design a flexible, reliable and stable algorithm to improve the control accuracy and response speed of the fan coil outlet temperature. Summary of the invention

[0004] The purpose of the present invention is to provide an adaptive finite-time neural network fan coil outlet air temperature control method and verification method to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] An adaptive finite-time neural network fan coil outlet air temperature control method, comprising:

[0007] Step 1, constructing mathematical models of flow control valve, fan coil unit and temperature sensor;

[0008] Step 2, using neural network approximation to design a finite time controller;

[0009] Step 3: Control the outlet air temperature of the fan coil unit through a finite time controller.

[0010] Furthermore, the mathematical model of the flow control valve constructed in step 1 includes:

[0011] definition To adjust the ratio, the mathematical model of the equal percentage flow characteristic of the flow control valve is set as:

[0012] ,

[0013] in, is the maximum valve core travel, is the valve core stroke, is the chilled water flow rate, is the maximum flow rate of the flow control valve, It is the minimum flow rate of the flow control valve.

[0014] Furthermore, constructing the mathematical model of the fan coil unit in step 1 includes: setting the single-capacity first-order differential equation of the fan coil unit to:

[0015] ,

[0016] in, , is the output air temperature of the fan coil unit, t is the time, C V is the specific heat at constant volume of air, is the volume of air circulated to the ventilator, is the air density, is the air flow rate of the fan coil unit, is the specific heat of air, is the chilled water temperature difference, is the specific heat of cooling water, is the inlet air temperature of the fan coil unit, is the total power of the fan coil.

[0017] Furthermore, the mathematical model of the temperature sensor constructed in step 1 includes: the temperature sensor adopts a thermistor, and the sensing temperature of the thermistor is set With output temperature The relationship between them is:

[0018] ,

[0019] in, is the time constant of the thermistor;

[0020] The existence of the time constant makes the thermal potential change of the thermistor always lag behind the perceived temperature change. The response characteristics of the thermistor can be regarded as a first-order inertial element. Description, that is:

[0021] ,

[0022] in, is a complex frequency domain variable in Laplace transform. To improve the measurement accuracy of the thermistor, a correction element is connected in series in the first-order inertial element, namely:

[0023] ,

[0024] in, is the time constant of the correction element, and are the transfer functions of the correction element and the feedback loop after correction, respectively.

[0025] Furthermore, the step 2 comprises:

[0026] Step 2.1, the response characteristics of the thermistor can be regarded as unit feedback. According to the characteristics of the flow control valve, we get:

[0027] ,

[0028] in, ;

[0029] According to formula (2) and formula (6), we can get:

[0030] ,

[0031] in, yes The derivative with respect to time, , and is uncertain in controller design, let ,as well as , then the fan coil control system becomes:

[0032] ,

[0033] Step 2.2, define tracking error for:

[0034] ,

[0035] in, yes The expected value to be tracked is derived from equation (9) to obtain:

[0036] ,

[0037] in for The derivative of is the total uncertainty;

[0038] Step 2.3, in equation (10), the adaptive neural network is used to approximate the total uncertainty function g, which is expressed as:

[0039] ,

[0040] in, is the neural network weight value The transposed matrix of Is The approximation error is limited. is a small positive constant, is a neural network basis function with the following sigmoid form:

[0041] ,

[0042] in, is the exponential function operator, , , and is a custom positive constant;

[0043] Step 2.4, use the neural network approximation to transform the finite time controller Designed for:

[0044] ,

[0045] in, , and is a positive number and satisfies is the symbolic function, yes The estimated value, weight Adaptive update law for:

[0046] ,

[0047] in is a normal number.

[0048] The present invention also provides a verification method for the above-mentioned adaptive finite time neural network fan coil outlet air temperature control method, comprising:

[0049] Step 1: In order to analyze the finite-time stability of the fan coil control system, the conditions required for the general nonlinear system to meet the actual finite-time stability are given, and the corresponding convergence time boundary is given;

[0050] Step 2: Based on the fan coil control system, design the Lyapunov function, prove that the system adjustment error can converge to a small area near the origin within a finite time, and give the convergence time range.

[0051] Furthermore, the step 1 specifically includes: In order to analyze the finite time stability of the fan coil control system, the following lemma is given:

[0052] Lemma 1: For a system , if there exists a continuous function , and there are multiple positive constants Eligible for the following conditions: and and meet ,in Expressed as The derivative of Expressed as of The system The trajectory of is actually finite-time stable, The residual set of the solution is ,in Representative time Convergence time hour The limit value of Indicates the minimum value, a normal number satisfy , convergence time The boundaries are:

[0053] (15)

[0054] in, Indicates taking the maximum value, is the initial time;

[0055] Lemma 2: For all positive numbers There exists a positive constant , so that the following inequality holds:

[0056] .

[0057] Furthermore, the step 2 specifically includes:

[0058] Design the following theorem to verify the designed finite time controller:

[0059] Theorem 1: Considering the fan coil control system expression (8), using the proposed finite time controller expression (13) and adaptive update law expression (14), it is proved that the regulation error It can converge to a small region near the origin in finite time:

[0060] Proof: Choose the Lyapunov function for:

[0061] ,

[0062] in ;

[0063] Taking the derivative of formula (17), we get:

[0064] ,

[0065] in, yes The derivative with respect to time, yes Regarding the time derivative, considering equations (10), (11), (13), (14) and (18), as well as Young's inequality, we obtain:

[0066] ,

[0067] in ;

[0068] According to Lemma 2, we get:

[0069] ,

[0070] in , , represents the minimum value, and ;

[0071] According to Lemma 1, the fan coil control system expression (8) satisfies the actual finite time stability characteristic, and the adjustment error It can converge to the following area ,Right now:

[0072] ,

[0073] in Representative time Convergence time hour In addition,

[0074] ,

[0075] And the constant satisfy ,constant satisfy , Indicates the maximum value.

[0076] Compared with the prior art, the present invention has the following beneficial effects:

[0077] 1) Compared with asymptotically stable control, the present invention systematically proposes a finite-time neural controller, which makes the temperature regulation error converge to a small area around the origin within a finite time;

[0078] 2) The present invention introduces a correction element to compensate the first-order damping element in the thermistor to capture the temperature in real time. In addition, a neural network-based adaptive law is designed to compensate for the unknown and nonlinear functions of the ASHPAC system without the need to obtain the complete model knowledge of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 This is a flow chart of an adaptive finite-time neural network fan coil outlet air temperature control method of the present invention.

[0080] Figure 2 This is a schematic diagram of the ASHPAC system.

[0081] Figure 3 This is a schematic diagram of the fan coil control structure.

[0082] Figure 4 Schematic diagram of the tracking performance of fan coil output temperature without neural network compensation.

[0083] Figure 5 Schematic diagram of the tracking performance of fan coil output temperature using neural network compensation.

[0084] Figure 6 Schematic diagram of the tracking error of the fan coil output temperature without neural network compensation.

[0085] Figure 7 Schematic diagram of the tracking error of the fan coil output temperature using neural network compensation.

[0086] Figure 8 Schematic diagram of the control input of the fan coil outlet temperature without neural network compensation.

[0087] Fig. 9 Schematic diagram of the control input of the fan coil outlet temperature using neural network compensation. DETAILED DESCRIPTION

[0088] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 creative work are within the scope of protection of the present invention.

[0089] See also Figure 1-Figure 3 , an adaptive finite time neural network fan coil outlet temperature control method, such as Figure 2 As shown, the ASHPAC system includes an air source side loop, a heat pump loop and an indoor loop. Figure 3 The control structure of the fan coil unit is composed of a flow control valve, a fan coil unit and a temperature sensor. Figure 3 It can be seen that and are the valve core travel and chilled water flow at a certain opening, is the output air temperature of the fan coil unit, is the output value of the temperature sensor, is the tracking target. The control method includes the following steps:

[0090] Step 1: Provide mathematical models of flow control valve, fan coil unit and temperature sensor, including:

[0091] Step 1.1, flow control valves generally have four ideal flow characteristics, namely linear characteristics, parabolic characteristics, quick opening characteristics and equal percentage characteristics. The percentage flow increases exponentially with the valve core stroke, and the same incremental valve stroke produces the same percentage change in the existing flow coefficient. Definition To adjust the ratio, the mathematical model of the equal percentage flow characteristic of the flow control valve is set as:

[0092] ,

[0093] in, is the maximum valve core travel, is the valve core stroke, is the chilled water flow rate, is the maximum flow rate of the flow control valve, It is the minimum flow rate of the flow control valve.

[0094] Step 1.2, the fan coil is the terminal equipment of the heat pump air conditioning system. It is a surface heat exchanger used to lower or raise the room temperature and set an indoor air environment suitable for human comfort. According to the law of conservation of energy, ignoring the heat storage effect of the fan coil, the mathematical model of the fan coil is simplified to a single-capacity first-order differential equation:

[0095] ,

[0096] in, , is the output air temperature of the fan coil unit, t is the time, C V is the specific heat at constant volume of air, is the volume of air circulated to the ventilator, is the air density, is the air flow rate of the fan coil unit, is the specific heat of air, is the chilled water temperature difference, is the specific heat of cooling water, is the inlet air temperature of the fan coil unit, is the total power of the fan coil.

[0097] Step 1.3, the temperature sensor uses a thermistor. Thermistors are often used as temperature sensors in air conditioning closed-loop control systems. Set the sensing temperature of the thermistor. With output temperature The relationship between them is:

[0098] ,

[0099] in, is the time constant of the thermistor.

[0100] The existence of the time constant makes the thermal potential change of the thermistor always lag behind the perceived temperature change. The response characteristics of the thermistor can be regarded as a first-order inertial element. Description, that is:

[0101] ,

[0102] in, is the complex frequency domain variable in Laplace transform. Fan coil temperature control requires fast and accurate The influence of is not negligible. In order to improve the measurement accuracy of thermistor temperature sensor, a correction element is connected in series with the first-order inertial element, that is:

[0103] ,

[0104] in, is the time constant of the correction element, and are the transfer functions of the correction element and the feedback loop after correction, respectively.

[0105] Step 2, using neural network approximation, design a finite time controller, specifically including:

[0106] Step 2.1, introduce the correction element, and eliminate the hysteresis of the thermistor through the correction element, so that the characteristics of the thermistor can be regarded as unit feedback. According to the characteristics of the flow control valve, for the convenience of expression, we get:

[0107] ,

[0108] in, .

[0109] According to formula (2) and formula (6), we can get:

[0110] ,

[0111] in, yes The derivative with respect to time, , and is uncertain in controller design, let ,as well as , then the fan coil control system becomes:

[0112] .

[0113] Step 2.2, define tracking error for:

[0114] ,

[0115] in, yes The expected value to be tracked. Derivative of equation (9) yields:

[0116] ,

[0117] in for The derivative of is the total uncertainty.

[0118] Step 2.3, in equation (10), the adaptive neural network is used to approximate the total uncertainty function g, which is expressed as:

[0119] ,

[0120] in, is the neural network weight value The transposed matrix of Is The approximation error is limited. is a small positive constant, and is a neural network (NN) basis function with the following sigmoid form:

[0121] ,

[0122] in, is the exponential function operator, , , and is a custom positive constant.

[0123] Step 2.4, use the neural network approximation to transform the finite time controller Designed for:

[0124] ,

[0125] in, represents the activation function of the neural network, , and is a positive number and satisfies is the symbolic function, yes The estimated value of for:

[0126] ,

[0127] in is a normal number.

[0128] Step 3: Control the outlet air temperature of the fan coil unit through a finite time controller.

[0129] The present invention also provides a verification method for the above-mentioned adaptive finite-time neural network fan coil outlet air temperature control method, comprising: designing a suitable Lyapunov function to verify the stability of the fan coil control system, specifically including:

[0130] Step 1, to analyze the finite-time stability of the fan coil control system, the following lemma is given:

[0131] Lemma 1: For a system , if there exists a continuous function , multiple custom scalars satisfy Eligible for the following conditions: and and meet ,in Expressed as The derivative of Expressed as of The system The trajectory of is actually finite-time stable, The residual set of the solution is ,in, Representative time Convergence time hour The limit value of Indicates taking the minimum value, satisfy , convergence time The boundaries are:

[0132] (15)

[0133] Lemma 2: For all positive numbers There is a constant , so that the following inequality holds:

[0134] ,

[0135] Step 2, design the following theorem to verify the designed finite time controller:

[0136] Theorem 1: Considering the fan coil control system expression (8), using the proposed finite time controller expression (13) and adaptive update law expression (14), it is proved that the regulation error It can converge to a small region near the origin in finite time:

[0137] Proof: Choose the Lyapunov function for:

[0138] ,

[0139] in .

[0140] Taking the derivative of formula (17), we get:

[0141] ,

[0142] in, yes The derivative with respect to time, yes Derivative with respect to time. Considering equations (10), (11), (13), (14) and (18), as well as Young's inequality, we obtain:

[0143] ,

[0144] in .

[0145] According to Lemma 2, we get:

[0146] ,

[0147] in , , represents the minimum value, and .

[0148] According to Lemma 1, the fan coil control system expression (8) satisfies the actual finite time stability characteristic, and the adjustment error It can converge to the following area ,Right now:

[0149] ,

[0150] in Representative time Convergence time hour In addition,

[0151] ,

[0152] And the constant satisfy ,constant satisfy , Indicates the maximum value.

[0153] In order to verify the effectiveness of the proposed adaptive finite time neural network fan coil outlet air temperature control scheme for the ASHPAC system, the present invention takes the "SF05RC" fan coil produced by PHNIX and the electric small flow regulating valve produced by VTON as the object, calculates the system parameters in Table 1, and gives the measurement units of these items.

[0154] Table 1 System parameters (all parameters calculated at 27°C)

[0155]

[0156] The mathematical models of the flow control valve and fan coil are:

[0157]

[0158] The target is selected to track the control step signal shown in Table 2, where r(t) is the step signal. In order to verify the effectiveness of the proposed scheme, two comparative cases with and without neural network compensation are given. For fair comparison, the parameters of the two schemes are the same. The corresponding simulation results are shown in Figure 4-Figure 9 The tracking performance without neural network compensation is shown in Figure 4 As shown. Figure 4 It can be seen that the output air temperature of the fan coil unit cannot reach the first target (24°C) and the second target (26°C) in a short time. Figure 5 The tracking performance after NN compensation is described. Figure 5 It can be seen that the tracking performance is superior. Figure 4 and Figure 5 It can be seen that the finite time controller can achieve fast and accurate temperature regulation with the help of neural network compensation. The temperature tracking errors e without neural network compensation and with neural network compensation are respectively as follows: Figure 6 and Figure 7 As shown. Figure 6 It can be seen that the temperature tracking error e is very large. Figure 7 It can be seen that when the temperature changes rapidly, the temperature tracking error e becomes larger, but the temperature tracking error e is generally small. In addition, the control input u without NN compensation and with NN compensation are as follows: Figure 8 and Fig. 9 As shown. Figure 8 and Fig. 9 It can be seen that NN compensation requires larger control input.

[0159] Table 2 Step signal tracking target

[0160]

[0161] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An adaptive finite time neural network fan coil outlet air temperature control method, characterized in that: include: Step 1, constructing mathematical models of flow control valve, fan coil unit and temperature sensor; The mathematical model of the flow control valve includes: definition To adjust the ratio, the mathematical model of the equal percentage flow characteristic of the flow control valve is set as: ， in, is the maximum valve core travel, is the valve core stroke, is the chilled water flow rate, is the maximum flow rate of the flow control valve, is the minimum flow rate of the flow control valve; The mathematical model of the fan coil unit is constructed by setting the single-capacity first-order differential equation of the fan coil unit as: ， in, , is the output air temperature of the fan coil unit, t is the time, C V is the specific heat at constant volume of air, is the volume of air circulated to the ventilator, is the air density, is the air flow rate of the fan coil unit, is the specific heat of air, is the chilled water temperature difference, is the specific heat of cooling water, is the inlet air temperature of the fan coil unit, is the total power of the fan coil; Step 2, using neural network approximation, design a finite time controller, including: Step 2.1, the response characteristics of the thermistor can be regarded as unit feedback. According to the characteristics of the flow control valve, we get: ， in, ; According to formula (2) and formula (6), we can get: ， in, yes The derivative with respect to time, , and is uncertain in controller design, let ,as well as , then the fan coil control system becomes: ， Step 2.2, define tracking error for: ， in, yes The expected value to be tracked is derived from equation (9) to obtain: ， in for The derivative of is the total uncertainty function; Step 2.3, in equation (10), the adaptive neural network is used to approximate the total uncertainty function g, which is expressed as: ， in, is the neural network weight value The transposed matrix of Is The approximation error is limited. is a small positive constant, is a neural network basis function with the following sigmoid form: ， in, is the exponential function operator, , , and is a custom positive constant; Step 2.4, use the neural network approximation to transform the finite time controller Designed for: ， in, , and is a positive number and satisfies is the symbolic function, yes The estimated value, weight Adaptive update law for: ， in is a positive constant; Step 3: Control the outlet air temperature of the fan coil unit through a finite time controller.

2. The adaptive finite time neural network fan coil outlet air temperature control method according to claim 1 is characterized in that: The mathematical model of the temperature sensor constructed in step 1 includes: the temperature sensor adopts a thermistor, and the sensing temperature of the thermistor is set. With output temperature The relationship between them is: ， in, is the time constant of the thermistor; The existence of the time constant makes the thermal potential change of the thermistor always lag behind the perceived temperature change. The response characteristics of the thermistor can be regarded as a first-order inertial element. Description, that is: ， in, is a complex frequency domain variable in Laplace transform. To improve the measurement accuracy of the thermistor, a correction element is connected in series in the first-order inertial element, namely: ， in, , T' is the time constant of the correction element, and are the transfer functions of the correction element and the feedback loop after correction, respectively.

3. A verification method for the adaptive finite time neural network fan coil outlet air temperature control method as claimed in claim 1 or 2, characterized in that: include: Step 1: In order to analyze the finite-time stability of the fan coil control system, the conditions required for the general nonlinear system to meet the actual finite-time stability are given, and the corresponding convergence time boundary is given, including: In order to analyze the finite-time stability of fan coil control systems, the following lemma is given: Lemma 1: For a system , if there exists a continuous function , and there are multiple positive constants Eligible for the following conditions: and and meet ,in Expressed as The derivative of Expressed as of The system The trajectory of is actually finite-time stable, The residual set of the solution is ,in Representative time Convergence time hour The limit value of Indicates the minimum value, a normal number satisfy , convergence time The boundaries are: (15)， in, Indicates taking the maximum value, is the initial time; Lemma 2: For all positive numbers There exists a positive constant , so that the following inequality holds: ； Step 2: Based on the fan coil control system, design the Lyapunov function to prove that the system adjustment error can converge to a small area near the origin within a finite time, and give the convergence time range, including: Design the following theorem to verify the designed finite time controller: Theorem 1: Considering the fan coil control system expression (8), using the proposed finite time controller expression (13) and adaptive update law expression (14), it is proved that the regulation error It can converge to a small region near the origin in finite time: Proof: Choose the Lyapunov function for: ， in ; Taking the derivative of formula (17), we get: ， in, yes The derivative with respect to time, yes Regarding the time derivative, considering equations (10), (11), (13), (14) and (18), as well as Young's inequality, we obtain: ， in ; According to Lemma 2, we get: ， in , , represents the minimum value, and ; According to Lemma 1, the fan coil control system expression (8) satisfies the actual finite time stability characteristic, and the adjustment error It can converge to the following area ,Right now: ， in Representative time Convergence time hour In addition, ， And the constant satisfy ,constant satisfy , Indicates the maximum value.

Images