Building cold and heat balance temperature control system and method based on neural network control algorithm

Abstract

The invention discloses a building cold and heat balance temperature control system and method based on a neural network control algorithm. The system comprises a meteorological monitoring module, a temperature control system, a temperature sensor, a chilled water circulation system and a heat source complementation system. The method comprises the steps that external environment parameters are monitored in real time, the internal temperature condition of a building is fed back to a temperature control system through a temperature sensor, and coordinated control is conducted on the temperature sensor and a fan coil; the temperature control system adjusts the cold and hot water flow or the cold and heat quantity provided by the fan coil to accurately control the indoor temperature; and constructing a temperature balance model, predicting and optimizing the cold and hot demands of the building through a neural network control algorithm, and providing a regulation and control strategy for a chilled water circulation system and a heat source complementation system. According to the invention, the cold and hot demands of the building are accurately predicted and optimized, the problems of response lag and insufficient control precision of a traditional temperature control system are reduced, and the consistency and stability of temperature regulation and control are improved.

Description

Technical Field

[0001] The present invention relates to the fields of artificial intelligence and automation control, building environment management, and energy efficiency optimization, and particularly to a building cooling and heating balance temperature control system and method based on a neural network control algorithm. Background Art

[0002] The energy consumption of the construction industry accounts for a relatively large proportion of the global total energy consumption, and the energy consumption of the heating, ventilation, and air conditioning (HVAC) system accounts for a relatively high proportion. As a key technology for building energy conservation and environmental comfort guarantee, the building cooling and heating balance temperature control system has important research and application values. Traditional building cooling and heating balance temperature control systems mainly rely on pre-set temperature thresholds and simple control algorithms. Indoor temperature data is collected through temperature sensors, and after comparison with the set value, equipment such as air-conditioning hosts and fan coil units are controlled. However, this control method has many limitations. On the one hand, the traditional system can only adjust based on the current temperature data, lacking forward-looking prediction of meteorological condition changes, unable to adjust the temperature control strategy in advance, resulting in a lag in response during sudden weather changes, large fluctuations in indoor temperature, and affecting user comfort. On the other hand, the control algorithm of the traditional system is relatively single, difficult to comprehensively consider the coupled effects of multiple factors such as the thermal inertia of the building, human activities, and equipment heat dissipation, with limited control accuracy and low energy utilization efficiency. In this context, it has become an urgent task to develop a new type of building cooling and heating balance temperature control system that can respond in real time, accurately regulate, and be energy-efficient. Summary of the Invention

[0003] Object of the Invention: The object of the present invention is to provide a building cooling and heating balance temperature control system and method based on a neural network control algorithm.

[0004] Technical Solution: The building cooling and heating balance temperature control system based on a neural network control algorithm described in the present invention includes a meteorological monitoring module, a temperature control system, temperature sensors, a chilled water circulation system, and a heat source complementary system; the meteorological monitoring module monitors the temperature, humidity, wind speed, and solar radiation parameters of the external environment in real time, providing external environment data support for the temperature control system, the temperature sensors are deployed in various areas inside the building to monitor the indoor temperature changes in real time, the temperature control system coordinates the control of the temperature sensors and the fan coil units, the data of the temperature sensors and the fan coil units are transmitted to the host of the temperature control system, and after the host of the temperature control system receives the data, it is transmitted to the temperature balance model, and the neural network control algorithm is used to predict and optimize the building cooling and heating demands, and provide control strategies for the chilled water circulation system and the heat source complementary system.

[0005] Further, the chilled water circulation system generates a cold source through a chiller unit, collaborates with a cooling tower for heat dissipation, and transports the chilled water to the fan coil unit to absorb indoor heat, realizing accurate regulation of the building cooling demand.

[0006] Furthermore, the heat source complementary system integrates multiple heat sources such as plate heat exchangers, boilers, and solar collectors, and flexibly allocates them according to actual conditions under heating demands.

[0007] The building cooling and heating balance temperature control method based on the neural network control algorithm of the present invention includes the following steps:

[0008] (1) Real-time monitor the temperature, humidity, wind speed, and solar radiation parameters of the external environment, and feedback the temperature condition inside the building to the temperature control system through temperature sensors to form a closed-loop control architecture;

[0009] (2) Coordinate the control of temperature sensors and fan coils to maintain the temperature inside the building within a set comfortable range;

[0010] (3) The temperature sensor real-time monitors the indoor temperature change, the fan coil adjusts the indoor temperature, and the temperature control system adjusts the cold and hot water flow rate or the cooling and heating capacity provided by the fan coil to accurately control the indoor temperature;

[0011] (4) Construct a temperature balance model, predict and optimize the cooling and heating demands of the building through the neural network control algorithm, and provide control strategies for the chilled water circulation system and the heat source complementary system.

[0012] Furthermore, the internal temperature condition feedback process in step (1) is as follows:

[0013] (11) Data encoding and quantization:

[0014] Quantize and encode the temperature value T collected by the sensor within the range of the measuring range:

[0015]

[0016] In the formula, D represents the temperature data after quantization encoding, T min represents the minimum measuring range of the temperature sensor, ΔT represents the quantization step, which determines the resolution of the temperature data, and round represents the rounding operation;

[0017] (12) Data verification and transmission:

[0018] Adopt a check code mechanism, and its calculation formula is:

[0019] In the formula, C represents the check code, D i represents the i-th temperature data after quantization encoding, n represents the number of temperature data points collected within a data transmission period, and mod256 represents the modulo operation to ensure that the check code is between 0 and 255.

[0020] Furthermore, in step (2), the working state of the fan coil is adjusted according to the control signal u(t), and the specific formula is as follows:

[0021] The output power P of the fan coil unit out (t) and the relationship of the control signal:

[0022]

[0023] In the formula, P max is the maximum output power of the fan coil unit, U max is the maximum value of the control signal;

[0024] The air velocity V of the fan connection plate fan (t) and the relationship of the control signal:

[0025]

[0026] In the formula, V max is the maximum air velocity of the fan coil unit;

[0027] The water valve opening A of the fan coil unit valve (t) and the relationship of the control signal:

[0028]

[0029] In the formula, A max is the maximum opening of the water valve.

[0030] Furthermore, the monitoring calculation formula of the temperature sensor in the step (3) is as follows:

[0031] Monitoring formula of a single temperature sensor:

[0032]

[0033] In the formula, T i (t) is the temperature detected by the temperature sensor at position i, t is the time, f(t) is the true change function of the temperature over time, is the sensor measurement error;

[0034] Monitoring formula for the fusion of multiple temperature sensors:

[0035] When multiple temperature sensors are deployed inside the building, in order to more accurately reflect the overall temperature condition inside the building, the method of sensor fusion can be adopted; assume there are N temperature sensors in total, and their monitoring data are T1(t), T2(t),..., T N (t), the monitoring calculation formula for the fused temperature is as follows:

[0036]

[0037] In the formula, w i is the weight coefficient of the i-th sensor.

[0038] Further, the construction process of the temperature balance model in step (4) is as follows:

[0039] (41) Construct a temperature balance model. The formula for the rate of change of indoor temperature over time is as follows:

[0040]

[0041] In the formula, Q supply (t) is the heat or cold supplied by the temperature control system, and Q demand (t) is the heating and cooling demand of the building at time t;

[0042] (42) Establish a dynamic balance differential equation for indoor temperature:

[0043]

[0044] In the formula, C is the equivalent heat capacity of the building, T ext is the external environmental temperature, and Φ is the set of environmental parameters;

[0045] (43) The objective function J(t) is defined as the sum of the squares of the errors between the temperature setpoint T set (t) and the actual indoor temperature T room (t). Considering the dynamic response characteristics of the system, the calculation formula is as follows:

[0046]

[0047] In the formula, e(t) = T set (t) - T room (t) is the error between the temperature setpoint and the actual value. σ and β are the weight coefficients of the differential term and the integral term respectively, used to adjust the response speed and steady-state error of the system;

[0048] (44) Construct a neural network prediction model:

[0049]

[0050] In the formula, T in (t - nΔt) is the indoor temperature at t - nΔt, T ext (t - nΔt) is the outdoor temperature at t - nΔt, Φ(t - nΔt) is the solar radiation intensity at t - nΔt, Δt is the time step, n is the length of the time window, H j is the output of the j-th neuron in the hidden layer, and σ is the activation function. The activation function is mainly used to limit the output of the neuron so that its output remains within a specific range. is the weight from the i-th neuron in the input layer to the j-th neuron in the hidden layer. X iis the output of the i-th neuron in the input layer, is the bias term of the j-th neuron in the hidden layer, is the weight from the i-th neuron in the hidden layer to the output layer, H i is the output of the i-th neuron in the hidden layer, b (2) is the bias term of the output layer;

[0051] (45) Define the loss function and parameter update rule:

[0052]

[0053] In the formula, L is the loss function of the neural network, which is used to measure the difference between the predicted value and the actual value;

[0054] is the mean square error, which is used to measure the predicted value and the actual value the difference between them; N is the number of training samples; λ||W|| 2 is the L2 regularization term, which is used to prevent overfitting; λ is the L2 regularization coefficient, which is used to control the strength of regularization;

[0055] (46) Control strategy generation:

[0056]

[0057] In the formula, is the heat or cold supplied by the optimized temperature control system, is to find the minimum heat or cold supplied by the temperature control system, β1 is the temperature deviation weight, β2 is the energy consumption weight, T in is the actual indoor temperature, △T is the prediction time domain, is used to measure the comprehensive influence of temperature deviation and energy consumption within the prediction time domain △T.

[0058] Furthermore, the optimization process of the neural network prediction model is as follows:

[0059] Introduce a dynamic sparse training strategy to optimize the neural network control algorithm, including the following process:

[0060] Neuron retention probability:

[0061]

[0062] In the formula, is the retention probability of neuron j at time t, is the weight of neuron j at time t, is the maximum absolute value of all neuron weights at time t, τ (t) is the threshold at time t;

[0063] Update threshold:

[0064] τ (t) = τ0·(1 - e -ηt )

[0065] where τ0 is the initial threshold, η is the attenuation coefficient with a default value of 0.05, which is used to control the attenuation speed of the threshold over time;

[0066] Gradient compensation:

[0067]

[0068] where is the gradient of the loss function L with respect to the weight w j , is the gradient of the loss function with respect to the pruned weight , and γ is the gradient suppression coefficient with a default value of 0.1, which is used to control the gradient attenuation speed.

[0069] Furthermore, the calculation formula of the heat source complementary system in step (4) is as follows:

[0070] Heat transfer area A of the plate heat exchanger:

[0071]

[0072] where Q is the heat load, K is the heat transfer coefficient, and △t m is the logarithmic mean temperature difference;

[0073] Logarithmic mean temperature difference △t m :

[0074]

[0075] where T1 and T2 are the inlet and outlet temperatures of the hot fluid, and t1 and t2 are the inlet and outlet temperatures of the cold fluid respectively;

[0076] Boiler thermal efficiency η1:

[0077]

[0078] where Q1 is the effective heat of the boiler, Q ar，net is the lower calorific value of the fuel as received, and B is the fuel consumption;

[0079] Solar heat collection efficiency η solar :

[0080]

[0081] where Q solar is the heat collected by solar energy, Q incidentis the incident solar heat;

[0082] The solar heat collection power P solar :

[0083] P solar = η solar ·I·A solar

[0084] In the formula, I is the solar irradiance intensity, and A solar is the solar heat collection area;

[0085] Total system output:

[0086] Q total = Q chiller + Q boiler + Q solar

[0087] In the formula, Q total is the total heat output of the system.

[0088] Advantages: Compared with the prior art, the present invention has the following remarkable advantages: By integrating an advanced meteorological monitoring module and a building feedback system, the present invention realizes the automation of temperature control inside the building. The system can monitor the external environmental parameters and internal temperature changes in real time, forming a closed-loop control architecture, reducing the need for manual intervention, lowering the operation complexity, and at the same time improving the continuity and stability of the temperature control operation; By adopting a neural network control algorithm to construct a temperature balance model, this system can accurately predict the changes in the heating and cooling demands of the building and generate optimized control strategies. This intelligent control method significantly improves the accuracy of temperature control, controls the indoor temperature fluctuation within a small range, effectively improves the user comfort. At the same time, by dynamically adjusting the operating states of the chilled water circulation system and the heat source complementary system, the system can adapt to complex and changeable environmental conditions to ensure the dynamic balance of the temperature inside the building; By intelligently allocating multiple heat sources and optimizing the chilled water circulation, the energy consumption is significantly reduced and the energy utilization efficiency is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0089] Figure 1 is a schematic structural diagram of the present invention;

[0090] Figure 2 is a flow chart of the neural network control algorithm of the present invention;

[0091] Figure 3 is an optimized flow chart of the neural network control algorithm of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0092] The technical solution of the present invention will be further described below with reference to the drawings.

[0093] AsFigure 1 As shown in Figure 1 , the building cooling and heating balance temperature control system based on the neural network control algorithm of the present invention includes a meteorological monitoring module, a temperature control system, temperature sensors, a chilled water circulation system, and a heat source complementary system; the meteorological monitoring module monitors the temperature, humidity, wind speed, and solar radiation parameters of the external environment in real time, providing external environment data support for the temperature control system, the temperature sensors are deployed in various areas inside the building to monitor the indoor temperature changes in real time, the temperature control system coordinates the control of the temperature sensors and the fan coils, the data of the temperature sensors and the fan coils are transmitted to the host of the temperature control system, and after the host of the temperature control system receives the data, it is transmitted to the temperature balance model, and the neural network control algorithm is used to predict and optimize the cooling and heating demands of the building, and provide control strategies for the chilled water circulation system and the heat source complementary system.

[0094] The building cooling and heating balance temperature control method based on the neural network control algorithm of the present invention improves the accuracy of temperature control inside the building and the energy utilization efficiency, and significantly improves the user comfort and the economy of system operation. The system realizes the dynamic balance of the indoor temperature by comprehensively processing the meteorological monitoring data and the building feedback information, and mainly includes the following steps:

[0095] 1) The meteorological monitoring module monitors the parameters such as temperature, humidity, wind speed, and solar radiation of the external environment in real time, providing external environment data support for the temperature control system, enabling it to respond in advance according to weather changes, optimize energy utilization and automatic temperature adjustment. As the application object of the temperature control system, the indoor temperature condition of the building is fed back to the temperature control system by the temperature sensors, forming a closed-loop control architecture, so as to dynamically adjust according to actual needs and achieve the accurate temperature control effect.

[0096] The temperature calculation formula is:

[0097]

[0098] In the formula, T avg represents the average temperature, T max represents the highest temperature, T min represents the lowest temperature.

[0099] The humidity calculation formula is:

[0100]

[0101] In the formula, RH is the relative humidity, e a is the actual vapor pressure, e s is the saturated vapor pressure.

[0102] The wind speed calculation formula is:

[0103]

[0104] Wherein, u represents the wind speed, and Δd represents the distance that the air flows within the time interval Δt.

[0105] The solar radiation calculation formula is:

[0106] R n = R ns - R nl

[0107] Wherein, R n represents the net radiation, R ns represents the net shortwave radiation, and R nl represents the net longwave radiation.

[0108] The temperature sensor converts the physical quantity of the temperature inside the building into an electrical signal, and its mathematical calculation formula is:

[0109] U = U0 + k(T - T0)

[0110] U is the voltage value output by the sensor, U0 is the output voltage of the sensor at the reference temperature T0, k is the sensitivity of the sensor, reflecting the linear relationship between the temperature change and the output voltage change, and T is the actually measured temperature value.

[0111] The process of feedback of the internal temperature condition is as follows:

[0112] 11) Data encoding and quantization:

[0113] Quantize and encode the temperature value T collected by the sensor within the range of the measurement range:

[0114]

[0115] Wherein, D represents the temperature data after quantization encoding, T min represents the minimum measurement range of the temperature sensor, ΔT represents the quantization step, which determines the resolution of the temperature data, and round represents the rounding operation.

[0116] 12) Data verification and transmission:

[0117] In order to ensure the reliability of data transmission, a check code mechanism is adopted, and its calculation formula is:

[0118] Wherein, C represents the check code, D i represents the i-th temperature data after quantization encoding, n represents the number of temperature data points collected within a data transmission period, and mod256 represents the modulo operation to ensure that the check code is between 0 and 255.

[0119] 2) As the control core of the entire building's heating and cooling balance temperature control system, the temperature control system coordinates the control of the temperature sensor and the fan coil unit to ensure that the temperature inside the building is maintained within the set comfortable range.

[0120] The feedback control formula of the temperature sensor is as follows:

[0121] e(t) = T set -T actual (t)

[0122] In the formula, T actual is the actual temperature measured by the temperature sensor, T set is the set temperature, and e(t) is the temperature error.

[0123] The control signal generation formula of the fan coil unit is as follows:

[0124]

[0125] In the formula, K p is the proportional gain, K i is the integral gain, K d is the derivative gain, which is used to eliminate long-term errors, is the rate of change of the error, which is used to predict future trends and make early adjustments, and u(t) is the control signal, which is used to adjust parameters such as the output power, wind speed, or water valve opening of the fan coil unit.

[0126] According to the control signal u(t), the working state of the fan coil unit is adjusted, and the specific formula is as follows:

[0127] 1. The relationship between the output power P out (t) of the fan coil unit and the control signal:

[0128]

[0129] In the formula, P max is the maximum output power of the fan coil unit, U max is the maximum value of the control signal.

[0130] 2. The relationship between the wind speed V fan (t) of the fan coil unit and the control signal:

[0131]

[0132] In the formula, V max is the maximum wind speed of the fan coil unit.

[0133] 3. The relationship between the water valve opening A valve (t) of the fan coil unit and the control signal:

[0134]

[0135] In the formula, A max is the maximum opening of the water valve.

[0136] 3) Temperature sensors are deployed in various areas inside the building to monitor the indoor temperature changes in real time. Fan coil units are installed in each room of the building and are responsible for adjusting the indoor temperature. By adjusting the cold and hot water flow rates or the cooling and heating capacities provided by the fan coil units, precise control of the indoor temperature is achieved.

[0137] The monitoring calculation formula of the temperature sensor is as follows:

[0138] Monitoring formula for a single temperature sensor:

[0139]

[0140] In the formula, T i (t) is the temperature measured by the temperature sensor at position i, t is the time, f(t) is the true change function of temperature over time, is the sensor measurement error.

[0141] Monitoring formula for the fusion of multiple temperature sensors:

[0142] When multiple temperature sensors are deployed inside the building, in order to more accurately reflect the overall temperature situation inside the building, a sensor fusion method can be adopted. Suppose there are a total of N temperature sensors, and their monitoring data are T1(t), T2(t),..., T N (t). The calculation formula for the fused temperature monitoring is as follows:

[0143]

[0144] In the formula, w i is the weight coefficient of the i-th sensor.

[0145] The calculation formula for the adjustment parameters of the fan coil unit is as follows:

[0146] Indoor temperature adjustment formula:

[0147]

[0148] In the formula, T room (t) is the indoor temperature, Q heat (t) is the heat provided by the fan coil unit, Q cool (t) is the cooling capacity provided by the fan coil unit, C room is the heat capacity of the room.

[0149] Cold and hot water flow rate adjustment formula:

[0150]

[0151] In the formula, Q water (t) is the cold and hot water flow rate at the fan connection, Q max is the maximum cold and hot water flow rate of the fan coil unit, Umax is the maximum value of the control signal.

[0152] The formula for the cooling and heating capacity provided by the fan coil unit:

[0153]

[0154] In the formula, is the mass flow rate of the chilled and hot water, c water is the specific heat capacity of water, T water，in and T water，out are the inlet and outlet temperatures of the chilled and hot water respectively.

[0155] 31) Data is transmitted to the main unit of the temperature control system

[0156] The data collected by the temperature sensor and the fan coil unit is sent to the main unit of the temperature control system through the data transmission link. The data transmission calculation formula is as follows:

[0157] D host (t) = {T1(t), T2(t),..., T N (t), Q water (t), V fan (t)}

[0158] In the formula, D host (t) represents the set of data received by the main unit of the temperature control system at time t.

[0159] 32) Data is transmitted to the temperature balance model

[0160] The main unit of the temperature control system further transmits the received data to the temperature balance model. The data transmission calculation formula is as follows:

[0161] D model (t) = F(D host (t))

[0162] In the formula, F is the data transmission function, and D model (t) is the data received by the temperature balance model at time t.

[0163] 4) Transmit the data of the temperature sensor and the fan coil unit to the main unit of the temperature control system. After the main unit of the temperature control system receives the data, it is transmitted to the temperature balance model. Through the neural network control algorithm, predict and optimize the heating and cooling demands of the building, and provide scientific control strategies for the chilled water circulation system and the heat source complementary system to achieve efficient utilization of energy and precise balance of temperature.

[0164] The construction process of the temperature balance model is as follows:

[0165] 41) Construct the temperature balance model. The formula for the rate of change of the indoor temperature over time is as follows:

[0166]

[0167] Wherein, Q supply (t) is the heat or cold supplied by the temperature control system, and Q demand (t) is the heating and cooling demand of the building at time t.

[0168] 42) Establish the differential equation of indoor temperature dynamic balance:

[0169]

[0170] Wherein, C is the equivalent heat capacity of the building, T ext is the external environmental temperature, and Φ is the set of environmental parameters.

[0171] 43) The objective function J(t) is defined as the sum of the squares of the errors between the temperature setpoint T set (t) and the actual indoor temperature T room (t), while considering the dynamic response characteristics of the system. The calculation formula is as follows:

[0172]

[0173] Wherein, e(t) = T set (t) - T room (t) is the error between the temperature setpoint and the actual value, and α and β are the weight coefficients of the differential term and the integral term respectively, which are used to adjust the response speed and steady-state error of the system.

[0174] Since the heating and cooling balance of the building is a dynamic change, traditional control algorithms are difficult to adjust the control strategy in real time to adapt to these changes, while the neural network control algorithm can dynamically adjust the control strategy according to real-time data to ensure that the output of the temperature control system is always optimal. The specific steps of the neural network control algorithm are as follows:

[0175] 44) As Figure 2 shown, construct the neural network prediction model:

[0176]

[0177] Wherein, T in (t - nΔt) is the indoor temperature at t - nΔt, T ext (t - nΔt) is the outdoor temperature at t - nΔt, Φ(t - nΔt) is the solar radiation intensity at t - nΔt, Δt is the time step, n is the length of the time window, H j is the output of the j-th neuron in the hidden layer, and σ is the activation function. The activation function is mainly used to limit the output of the neuron to keep its output within a specific range. is the weight from the i-th neuron in the input layer to the j-th neuron in the hidden layer. X i is the output of the i-th neuron in the input layer, is the bias term of the j-th neuron in the hidden layer, is the weight from the i-th neuron in the hidden layer to the output layer, H i is the output of the i-th neuron in the hidden layer, b (2) is the bias term of the output layer.

[0178] 45) Define the loss function and parameter update rules:

[0179]

[0180] In the formula, L is the loss function of the neural network, which is used to measure the difference between the predicted value and the actual value; is the mean square error, which is used to measure the predicted value and the actual value The difference between them; N is the number of training samples; λ||W|| 2 is the L2 regularization term, which is used to prevent overfitting; λ is the L2 regularization coefficient, which is used to control the strength of regularization.

[0181] 46) Control strategy generation:

[0182]

[0183] In the formula, is the heat or cold supplied by the optimized temperature control system, is to find the minimum heat or cold supplied by the temperature control system, β1 is the temperature deviation weight, β2 is the energy consumption weight, T in is the actual indoor temperature, ΔT is the prediction time domain, is used to measure the comprehensive influence of temperature deviation and energy consumption within the prediction time domain ΔT.

[0184] 47) Although the neural network control algorithm performs well on the training data, it may lead to performance degradation due to overfitting on the test temperature control data. Therefore, through the dynamic sparse training strategy, the neuron pruning mechanism can automatically remove unimportant neurons to optimize the neural network control algorithm, thereby reducing the risk of overfitting. As Figure 3 shown, the process of introducing the dynamic sparse training strategy is as follows:

[0185] 47.1) Neuron retention probability:

[0186]

[0187] In the formula, is the retention probability of neuron j at time t, The weight of neuron j at time t is the maximum absolute value of the weights of all neurons at time t, τ (t) is the threshold at time t.

[0188] 47.2) Update the threshold:

[0189] τ (t) = τ0·(1 - e -ηt )

[0190] In the formula, τ0 is the initial threshold, η is the attenuation coefficient, and the default value is 0.05, which is used to control the attenuation speed of the threshold over time.

[0191] 47.3) Gradient compensation:

[0192]

[0193] In the formula, is the gradient of the loss function L with respect to the weight w j , is the gradient of the loss function with respect to the pruned weight , and γ is the gradient suppression coefficient, with a default value of 0.1, which is used to control the gradient attenuation speed.

[0194] The generalization ability of the improved neural network control algorithm on the test temperature data is significantly improved, and it can better adapt to the complex environment in practical applications and obtain the minimum error e(t) min .

[0195] 5) The chilled water circulation system generates a cold source through the chiller, dissipates heat in coordination with the cooling tower, and delivers the chilled water to the fan coil unit to absorb indoor heat, realizing precise control of the building's cooling demand.

[0196] The refrigerating capacity Q of the chiller chiller is calculated as follows:

[0197] Q chiller = m water · c water · (T e - T l )

[0198] In the formula, m water is the mass flow rate of the chilled water, T e is the temperature of the chilled water entering the chiller, and T l is the temperature of the chilled water leaving the chiller.

[0199] The heat dissipation efficiency η of the cooling tower coolingtower is calculated as follows:

[0200]

[0201] In the formula, Q rejected is the heat dissipated by the cooling tower.

[0202] 6) The heat source complementary system integrates multiple heat sources such as plate heat exchangers, boilers, and solar collectors. Under heating demand, it is flexibly allocated according to the actual situation to ensure the stability and economy of heating.

[0203] The calculation formula for the heat source complementary system is as follows:

[0204] 61) Heat transfer area A of the plate heat exchanger:

[0205]

[0206] In the formula, Q is the heat load, K is the heat transfer coefficient, and Δt m is the logarithmic mean temperature difference.

[0207] 62) Logarithmic mean temperature difference Δt m :

[0208]

[0209] In the formula, T1 and T2 are the inlet and outlet temperatures of the hot fluid, and t1 and t2 are the inlet and outlet temperatures of the cold fluid respectively.

[0210] 63) Boiler thermal efficiency η1:

[0211]

[0212] In the formula, Q1 is the effective heat of the boiler, Q ar，net is the lower calorific value of the fuel as received, and B is the fuel consumption.

[0213] 64) Solar heat collection efficiency η solar :

[0214]

[0215] In the formula, Q solar is the heat collected by the solar energy, and Q incident is the incident solar heat.

[0216] 65) Solar heat collection power P solar :

[0217] P solar = η solar ·I·A solar

[0218] In the formula, I is the solar irradiance intensity, and A solar is the solar heat collection area.

[0219] 66) Total system output:

[0220] Q total = Q chiller + Q boiler + Q solar

[0221] Wherein, Q total is the total heat output of the system.

Claims

1. A building heat and cold balance temperature control system based on a neural network control algorithm, characterized in that: It includes a meteorological monitoring module, a temperature control system, a temperature sensor, a chilled water circulation system and a heat source complementary system; the meteorological monitoring module monitors the temperature, humidity, wind speed and solar radiation parameters of the external environment in real time, and provides external environment data support for the temperature control system; the temperature sensor is deployed in various areas inside the building to monitor indoor temperature changes in real time; the temperature control system coordinates and controls the temperature sensor and the fan coil unit; the data of the temperature sensor and the fan coil unit are transmitted to the temperature control system host; after receiving the data, the temperature control system host transmits the data to the temperature balance model; the building's cooling and heating needs are predicted and optimized through a neural network control algorithm, and a regulation strategy is provided for the chilled water circulation system and the heat source complementary system.

2. The building heat and cold balance temperature control system based on the neural network control algorithm according to claim 1 is characterized in that: The chilled water circulation system generates a cold source through refrigeration by the refrigeration unit, cooperates with the cooling tower to dissipate heat, and transports chilled water to the fan coil to absorb indoor heat, thereby achieving precise control of the building's cooling needs.

3. The building heat and cold balance temperature control system based on the neural network control algorithm according to claim 1 is characterized in that: The heat source complementary system integrates multiple heat sources such as plate heat exchangers, boilers and solar collectors, and is flexibly allocated according to actual heating needs.

4. A building heat and cold balance temperature control method based on a neural network control algorithm, characterized in that: The steps include: (1) Real-time monitoring of the temperature, humidity, wind speed and solar radiation parameters of the external environment, and feedback of the internal temperature conditions of the building to the temperature control system through the temperature sensor, forming a closed-loop control architecture; (2) Coordinate the control of temperature sensors and fan coil units to maintain the internal temperature of the building within the set comfort range; (3) The temperature sensor monitors indoor temperature changes in real time, the fan coil adjusts the indoor temperature, and the temperature control system adjusts the flow of hot and cold water or the amount of cold and heat provided by the fan coil to accurately control the indoor temperature; (4) Construct a temperature balance model, predict and optimize the building's heating and cooling demands through a neural network control algorithm, and provide control strategies for the chilled water circulation system and heat source complementary system.

5. The building heat and cold balance temperature control method based on the neural network control algorithm according to claim 4 is characterized in that: The internal temperature condition feedback process in step (1) is as follows: (11) Data coding and quantization: The temperature value T collected by the sensor is quantized and encoded within the range: Where D represents the quantized temperature data, T mm Indicates the minimum range of the temperature sensor, ΔT indicates the quantization step size, which determines the resolution of the temperature data, and round indicates the rounding operation; (12) Data verification and transmission: The verification code mechanism is adopted, and its calculation formula is: In the formula, C represents the check code, D i represents the temperature data after quantization encoding of the i-th value, n represents the number of temperature data points collected in one data transmission cycle, and mod256 represents the modulo operation to ensure that the check code is between 0 and 255.

6. The building heat and cold balance temperature control method based on the neural network control algorithm according to claim 4 is characterized in that: The step (2) adjusts the working state of the fan coil unit according to the control signal u(t), and the specific formula is as follows: Output power of fan coil unit P out (t) Relationship with control signal: Where P max is the maximum output power of the fan coil unit, U max is the maximum value of the control signal; Wind speed V of fan receiving plate fan (t) Relationship with control signal: Where V max is the maximum wind speed of the fan coil unit; Fan coil water valve opening A valve (t) Relationship with control signal: In the formula, A max is the maximum opening of the water valve.

7. The building heat and cold balance temperature control method based on the neural network control algorithm according to claim 4 is characterized in that: The temperature sensor monitoring calculation formula in step (3) is as follows: Single temperature sensor monitoring formula: Where, T i (t) is the temperature detected by the temperature sensor at position i, t is time, and f(t) is the actual change function of temperature over time. is the sensor measurement error; Multiple temperature sensor fusion monitoring formula: When multiple temperature sensors are deployed inside a building, in order to more accurately reflect the overall temperature conditions inside the building, a sensor fusion method can be used. Suppose there are N temperature sensors in total, and their monitoring data are T1(t), T2(t), ..., T N (t), the temperature monitoring calculation formula after fusion is as follows: In the formula, w i is the weight coefficient of the i-th sensor.

8. The building heat and cold balance temperature control method based on the neural network control algorithm according to claim 4 is characterized in that: The construction process of the temperature balance model in step (4) is as follows: (41) The temperature balance model is constructed, and the formula for the rate of change of indoor temperature over time is as follows: In the formula, Q supply (t) is the heat or cold supplied by the temperature control system, Q demand (t) is the heating and cooling demand of the building at time t; (42) Establish the dynamic equilibrium differential equation of indoor temperature: Where C is the equivalent heat capacity of the building, T ext is the external environment temperature, Φ is the set of environmental parameters; (43) The objective function J(t) is defined as the temperature setting value T set (t) and the actual indoor temperature T room (t), taking into account the dynamic response characteristics of the system, the calculation formula is as follows: Where, e(t) = T set (t)-T room (t) is the error between the temperature setting value and the actual value, α and β are the weight coefficients of the differential term and the integral term, respectively, which are used to adjust the response speed and steady-state error of the system; (44) Constructing a neural network prediction model: Where, T in (t-nΔt) is the indoor temperature at t-nΔt, T ext (t-nΔt) is the outdoor temperature at t-nΔt, Φ(t-nΔt) is the solar radiation intensity at t-nΔt, Δt is the time step, n is the time window length, H j is the output of the jth neuron in the hidden layer, σ is the activation function, and the activation function is mainly used to limit the output of the neuron so that its output remains within a specific range. is the weight from the i-th neuron in the input layer to the j-th neuron in the hidden layer. i is the output of the ith neuron in the input layer, is the bias term of the jth neuron in the hidden layer, is the weight from the i-th neuron in the hidden layer to the output layer, H i is the output of the i-th neuron in the hidden layer, b (2) is the bias term of the output layer; (45) Define the loss function and parameter update rules: Where L is the loss function of the neural network, which is used to measure the difference between the predicted value and the actual value; is the mean square error, which is used to measure the predicted value With actual value The difference between; N is the number of training samples; λ||W|| 2 is the L2 regularization term, used to prevent overfitting; λ is the L2 regularization coefficient, used to control the strength of regularization; (46) Control strategy generation: In the formula, The heat or cold supplied to the optimized temperature control system, To find the minimum heat or cold supplied by the temperature control system, β1 is the temperature deviation weight, β2 is the energy consumption weight, T in is the actual indoor temperature, ΔT is the prediction time domain, It is used to measure the combined impact of temperature deviation and energy consumption within the prediction time domain ΔT.

9. The building heat and cold balance temperature control method based on the neural network control algorithm according to claim 8 is characterized in that: The optimization process of the neural network prediction model is as follows: The dynamic sparse training strategy is introduced to optimize the neural network control algorithm, including the following process: Probability of retaining neurons: In the formula, is the retention probability of neuron j at time t, is the weight of neuron j at time t, is the absolute maximum value of all neuron weights at time t, τ (t) is the threshold value at time t; Update threshold: t (t) =τ0·(1-e ηt ) In the formula, τ0 is the initial threshold, η is the attenuation coefficient, the default value is 0.05, which is used to control the attenuation speed of the threshold over time; Gradient compensation: In the formula, is the loss function L for weight w j The gradient of The loss function is the weight after pruning The gradient of γ is the gradient suppression coefficient, the default value is 0.1, which is used to control the speed of gradient decay.

10. The building heat and cold balance temperature control method based on the neural network control algorithm according to claim 1 is characterized in that: The calculation formula of the heat source complementary system in step (4) is as follows: Heat transfer area of ​​plate heat exchanger A: Where Q is the heat load, K is the heat transfer coefficient, △t m is the logarithmic mean temperature difference; Logarithmic mean temperature difference △t m : Where T1 and T2 are the inlet and outlet temperatures of the hot fluid, and T1 and t2 are the inlet and outlet temperatures of the cold fluid; Boiler thermal efficiency η1: In the formula, Q1 is the effective heat of the boiler, Q ar，net is the basic low calorific value of the fuel received, and B is the fuel consumption; Solar thermal efficiency η solar : In the formula, Q solar The heat collected by solar energy, Q incident is the incident solar heat; Solar thermal power P solar : P solar =the solar ·I·A solar Where I is the solar radiation intensity, A solar is the solar thermal collection area; Total system output: Q total =Q chiller +Q boiler +Q solar In the formula, Q total is the total heat output of the system.

Images