Dynamic regulation and control method and system for photovoltaic capacity of air conditioning system considering flexible load

Abstract

The invention discloses an air conditioning system photovoltaic capacity dynamic regulation and control method and system considering a flexible load, and the method comprises the steps: collecting system operation data, building an energy consumption mathematical model of main energy consumption equipment of a heating ventilation air conditioning system, developing supply and demand double-side prediction model algorithm software, and building a photovoltaic capacity dynamic tracking model prediction control cost function; according to the method, the flexible load-considered air conditioning system photovoltaic capacity dynamic tracking model prediction control dynamic regulation and control system is obtained, the chilled water supply temperature, the chilled water flow and the cooling water return temperature are subjected to optimization calculation, feedback correction is carried out on the system output cooling capacity, and automatic regulation and control are achieved. According to the method, the demand side air conditioner load consumption curve dynamically tracks the change of the photovoltaic power generation curve, the system energy consumption cost is minimized, power grid fluctuation caused by current large-scale use of solar energy is considered, and the method is of great significance to improvement of power grid stability and realization of building demand response.

Description

Technical Field

[0001] The present invention belongs to the technical field of photovoltaic building demand response, and relates to a method and system for dynamically regulating photovoltaic production capacity of an air-conditioning system taking flexible load into consideration. Background Art

[0002] In recent years, the degree of grid connection of renewable energy has increased significantly. Due to the strong volatility, intermittency and randomness of renewable energy, building photovoltaic grid-connected power generation will lead to a series of problems such as power imbalance and voltage fluctuation. The integration of building power systems with the power grid to achieve power balance and ensure the stable operation of the power grid has brought new challenges. Since the HVAC system consumes a large proportion of electricity, it serves as a flexible load of the building and provides multiple possibilities for demand response. Therefore, considering the inherent uncertainty of renewable energy, it is crucial to formulate a dynamic control strategy for photovoltaic tracking, combine it with flexible load management, and optimize the operating parameters of the HVAC system.

[0003] Model predictive control (MPC) is based on the dynamic model of the system. It can accurately predict the future behavior of the system and make control decisions based on these predictions. This method effectively improves the response speed and stability of the system. At the same time, the MPC strategy has been fully applied in power systems and HVAC systems, and has achieved satisfactory control effects. However, there are few studies on MPC regulation of flexible loads in HVAC systems. At the same time, the objective function in MPC rarely considers adjusting electricity consumption according to the output of renewable energy. Therefore, it has become a technical problem that needs to be solved urgently to establish an MPC regulation model based on flexible load allocation at the HVAC system level by tracking renewable energy output. It has important research value for improving the balance between supply and demand and the reliability of the power grid. Summary of the invention

[0004] The purpose of the present invention is to solve the problem that there are few studies on MPC control for flexible loads of HVAC systems in the prior art, and at the same time, the objective function in MPC rarely considers adjusting electricity consumption according to the output of renewable energy, and to provide a method and system for dynamic control of photovoltaic power generation of air-conditioning systems taking flexible loads into account.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A method for dynamically controlling photovoltaic power generation of an air-conditioning system taking flexible load into consideration comprises:

[0007] Collect historical data of photovoltaic power generation system and HVAC system, and build HVAC system cooling load database and photovoltaic power generation system database;

[0008] Based on the operating parameters in the collected HVAC historical data, a mathematical model between the energy consumption of energy-consuming equipment and historical data is constructed to obtain the HVAC power consumption;

[0009] On the supply side, based on the photovoltaic power generation system database, the Grey Wolf Optimization Algorithm GWO is used to optimize the Long Short-Term Memory LSTM neural network algorithm, and a solar radiation prediction model based on GWO-LSTM is established. The predicted solar radiation value is used as the input parameter to establish a photovoltaic power generation prediction model based on the back propagation neural network to obtain the power generation of the photovoltaic power generation system.

[0010] On the demand side, based on the HVAC system cooling load database, a building cooling load prediction model based on generalized regression neural network GRNN is established, and the indoor temperature prediction model is obtained based on the linear transformation of the cooling load prediction model;

[0011] Based on the prediction models of HVAC power consumption, photovoltaic power generation system power generation and indoor temperature, a photovoltaic power generation dynamic tracking model predicts the control cost function;

[0012] Based on the genetic algorithm, the cost function is optimized globally online multiple times to minimize the difference between the power consumption curve of the HVAC system and the photovoltaic power generation curve, and to optimize the room temperature penalty caused by the difference to obtain the optimal control variables in the future limited time domain. Then, the chilled water supply temperature, chilled water flow rate and cooling water return temperature are optimized and calculated, and the system cooling capacity is feedback corrected to complete the dynamic regulation of the photovoltaic capacity of the air-conditioning system.

[0013] A further improvement of the present invention is:

[0014] Furthermore, the historical data of the photovoltaic power generation system and the HVAC system include: equipment operation parameters, indoor temperature parameters, outdoor meteorological parameters, solar photovoltaic power generation and building cooling load, and a HVAC system cooling load database and a photovoltaic power generation system database are established respectively.

[0015] Furthermore, a mathematical model between the energy consumption of the energy-consuming equipment and the equipment operating parameters is constructed, specifically: based on theoretical or empirical formulas, the mathematical model of the energy-consuming equipment is obtained through the particle swarm optimization algorithm PSO.

[0016] Furthermore, the Grey Wolf Optimization Algorithm (GWO) is used to optimize the Long Short-Term Memory (LSTM) neural network algorithm, specifically:

[0017] Collect data sets and preprocess them;

[0018] Divide the preprocessed data set into training set and test set;

[0019] Initialize the hyperparameters of the LSTM neural network and the gray wolf optimization algorithm respectively; input the training set into the gray wolf optimization algorithm, and output the optimal solution α wolf position as the GWO-LSTM prediction model in the hyperparameter input;

[0020] Verify based on the test set, output the results, and evaluate the prediction performance of the GWO-LSTM model.

[0021] Furthermore, the prediction performance of the GWO-LSTM model was evaluated, specifically: the prediction performance of the GWO-LSTM model was evaluated by the correlation coefficient R, the relative mean absolute error rMAE, and the relative root mean square error rRMSE;

[0022] Among them, the correlation coefficient R is:

[0023]

[0024] The relative mean absolute error rMAE is:

[0025]

[0026] The relative root mean square error rRMSE is:

[0027]

[0028] Among them, G Pi is the predicted value of solar radiation intensity; is the average value of the solar radiation intensity prediction value series; G Mi is the actual measured value of solar radiation intensity; is the average value of the actual measurement value sequence of solar radiation intensity; N is the number of samples.

[0029] Furthermore, the solar radiation prediction model based on GWO-LSTM is a black box model; the solar radiation prediction model based on GWO-LSTM is established by selecting the current horizontal solar total radiation value G(t), solar azimuth A(t), ambient relative humidity RH(t), and outdoor dry-bulb temperature T at the prediction time in the photovoltaic power generation system database. w (t+i), outdoor relative humidity RH(t+i) and solar azimuth A(t+i) are used as input parameters of the GWO-LSTM solar radiation prediction model, and the solar radiation prediction model outputs the horizontal total solar radiation value G(t+i) at the prediction moment.

[0030] Furthermore, the photovoltaic power generation prediction model based on the back propagation neural network is a black box model; the solar radiation prediction value is used as an input parameter to establish a photovoltaic power generation prediction model based on the back propagation neural network to obtain the power generation of the photovoltaic power generation system, specifically: based on the photovoltaic power generation E(t), the horizontal solar total radiation value G(t), and the ambient relative humidity RH(t) at the current moment, the horizontal solar total radiation value G(t+i) and the ambient relative humidity RH(t+i) at the prediction moment are input as input parameters into the BP neural network photovoltaic power generation prediction model to obtain the photovoltaic power generation E(t+i) at the prediction moment;

[0031] The BP neural network photovoltaic power generation prediction model structure has 5 input layer nodes, 13 hidden layer neurons, and 1 output layer node.

[0032] Furthermore, the building cooling load prediction model is a black box model; the indoor temperature prediction model is obtained based on the linear transformation of the building cooling load prediction model, specifically: the current cooling load Q(t) and indoor temperature T in the HVAC system cooling load database are selected. n (t), outdoor temperature T at the predicted time w (t+1), total solar radiation E(t+1), ambient relative humidity RH(t+1) and room temperature setting value T set As the input parameter of the cooling load prediction model, the cooling load prediction model predicts and outputs the cooling load prediction value Q(t+1);

[0033] The relationship between the predicted room temperature and the current room temperature and cooling load is obtained through linear regression, which simplifies the algorithm design and optimization time. The relationship is as follows:

[0034] T n (t+1)=0.244+0.993×T n (t)-0.003×Q(t+1)+0.003×Q(t).

[0035] Furthermore, a photovoltaic capacity dynamic tracking model prediction control cost function is constructed, specifically:

[0036]

[0037] Among them, P is the power consumption of the HVAC system; E is the power generation of the photovoltaic power generation system; λ and γ are weight coefficients, T set,i is the room temperature setting value at the i-th moment, ℃; T is the optimal indoor setting temperature, ℃; E(k+i-1) is the photovoltaic power generation at the current moment; E(k+i) is the photovoltaic power generation at the predicted moment; P(k+i-1) is the power consumption of the HVAC system at the current moment; P(k+i) is the power consumption of the HVAC system at the future moment.

[0038] A photovoltaic power generation dynamic control system for an air conditioning system taking flexible load into account comprises:

[0039] A collection module, wherein the collection module collects historical data of the photovoltaic power generation system and the HVAC system to build a HVAC system cold load database and a photovoltaic power generation system database;

[0040] A first construction module, wherein the first construction module constructs a mathematical model between the energy consumption of the energy-consuming equipment and the historical data based on the operating parameters in the collected HVAC historical data to obtain the HVAC power consumption;

[0041] An acquisition module, on the supply side, based on a photovoltaic power generation system database, uses a Grey Wolf Optimization Algorithm (GWO) to optimize a Long Short-Term Memory (LSTM) neural network algorithm, establishes a solar radiation prediction model based on GWO-LSTM, uses the solar radiation prediction value as an input parameter, establishes a photovoltaic power generation prediction model based on a back propagation neural network, and obtains the power generation of the photovoltaic power generation system;

[0042] A linear transformation module, which establishes a building cooling load prediction model based on a generalized regression neural network GRNN based on a cooling load database of a HVAC system on the demand side, and obtains an indoor temperature prediction model based on a linear transformation of the cooling load prediction model;

[0043] A second building module, wherein the second building module builds a photovoltaic power generation dynamic tracking model prediction control cost function based on the HVAC power consumption, photovoltaic power generation system power generation and indoor temperature prediction model;

[0044] The optimization module performs multiple global online optimizations on the cost function based on a genetic algorithm, minimizes the difference between the HVAC system power consumption curve and the photovoltaic power generation curve, and optimizes the room temperature penalty caused by the difference to obtain the optimal control variables in the future limited time domain, and then optimizes the calculation of the chilled water supply temperature, chilled water flow rate and cooling water return temperature, and feedback corrects the system cooling capacity to complete the dynamic regulation of the photovoltaic capacity of the air-conditioning system.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] The present invention collects system operation data, establishes a mathematical model of energy consumption of the main energy-consuming equipment of the HVAC system, develops supply and demand double-side prediction model algorithm software, constructs a photovoltaic capacity dynamic tracking model prediction control cost function, and obtains a photovoltaic capacity dynamic tracking model prediction control dynamic control system for the air-conditioning system taking into account flexible loads, optimizes and calculates the chilled water supply temperature, chilled water flow rate and cooling water return temperature, and performs feedback correction on the system output cooling capacity to achieve automatic control. The present invention enables the demand-side air-conditioning load consumption curve to dynamically track the changes in the photovoltaic power generation curve, which not only minimizes the system energy consumption cost, but also takes into account the power grid fluctuations caused by the current large-scale use of solar energy, which is of great significance to improving the stability of the power grid and achieving building demand response. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0048] Figure 1 It is a schematic flow chart of a method for dynamically controlling photovoltaic power generation of an air-conditioning system taking flexible load into consideration according to the present invention;

[0049] Figure 2 Establish a flow chart for the GWO-LSTM prediction model;

[0050] Figure 3 Comparison chart between the predicted values ​​and actual values ​​of LSTM and GWO-LSTM models;

[0051] Figure 4 It is the topological diagram of BP neural network photovoltaic power generation prediction model;

[0052] Figure 5 This is a comparison chart between the predicted value and the measured value of the BP neural network photovoltaic power generation prediction model;

[0053] Figure 6 It is a structural schematic diagram of the photovoltaic power generation dynamic control system of the air-conditioning system taking into account the flexible load of the present invention;

[0054] Figure 7 Another schematic diagram of the process of the method for dynamically controlling photovoltaic power generation of an air-conditioning system taking flexible load into consideration according to the present invention;

[0055] Figure 8 Schematic diagram of the cost function conception;

[0056] Fig. 9 This is a control flow diagram of the controller. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0058] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. 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.

[0059] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0060] In the description of the embodiments of the present invention, it should be noted that if the terms "upper", "lower", "horizontal", "inner", etc. indicate an orientation or positional relationship based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship in which the product of the invention is usually placed when in use, it is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. In addition, the terms "first", "second", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0061] In addition, if the term "horizontal" appears, it does not mean that the component must be absolutely horizontal, but can be slightly tilted. For example, "horizontal" only means that its direction is more horizontal than "vertical", which does not mean that the structure must be completely horizontal, but can be slightly tilted.

[0062] In the description of the embodiments of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "set", "install", "connect", and "connect" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal connection of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0063] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0064] See also Figure 1 The present invention discloses a method for dynamically controlling photovoltaic power generation of an air-conditioning system taking flexible load into consideration, comprising:

[0065] S101, collecting historical data of the photovoltaic power generation system and the HVAC system, and building a HVAC system cooling load database and a photovoltaic power generation system database;

[0066] The historical data of the photovoltaic power generation system and the HVAC system include: equipment operating parameters, indoor temperature parameters, outdoor meteorological parameters, solar photovoltaic power generation and building cooling load. The HVAC system cooling load database and the photovoltaic power generation system database are established respectively.

[0067] S102, based on the operating parameters in the collected HVAC historical data, construct a mathematical model between the energy consumption of the energy-consuming equipment and the historical data to obtain the HVAC power consumption;

[0068] Based on theoretical or empirical formulas, the mathematical model of energy-consuming equipment is obtained through the particle swarm optimization algorithm PSO.

[0069] S103, on the supply side, based on the photovoltaic power generation system database, the Grey Wolf Optimization Algorithm GWO is used to optimize the Long Short-Term Memory LSTM neural network algorithm, a solar radiation prediction model based on GWO-LSTM is established, and the solar radiation prediction value is used as an input parameter to establish a photovoltaic power generation prediction model based on the back propagation neural network to obtain the power generation of the photovoltaic power generation system;

[0070] See also Figure 2 , the Grey Wolf Optimization Algorithm GWO is used to optimize the long short-term memory LSTM neural network algorithm, specifically:

[0071] Collect data sets and preprocess them;

[0072] Divide the preprocessed data set into training set and test set;

[0073] Initialize the hyperparameters of the LSTM neural network and the gray wolf optimization algorithm respectively; input the training set into the gray wolf optimization algorithm, and output the optimal solution α wolf position as the GWO-LSTM prediction model in the hyperparameter input;

[0074] Verify based on the test set, output the results, and evaluate the prediction performance of the GWO-LSTM model.

[0075] In the established GWO-LSTM model, the number of wolves in the gray wolf optimization algorithm is 6, the maximum number of iterations is 500, the optimization parameters are the number of hidden layer neurons and the learning rate in the LSTM neural network, and the upper limit ub of the two parameters is set to [150, 0.5], and the lower limit lb is set to [80, 10 -10 ], the maximum number of iterations of the maximum LSTM neural network is 2000. The final optimized LSTM neural network has a learning rate of 0.0359 and a number of hidden layer neurons of 91. At the same time, the relevant network parameters with the smallest error are selected by trial and error, and the established LSTM neural network is used as a comparison. The number of hidden neurons in the LSTM neural network is set to 100, the initial learning rate is 0.5, and the learning rate decreases by 0.9 every 100 generations.

[0076] The comparison between the predicted values ​​and actual values ​​of LSTM and GWO-LSTM models is shown in the figure below. Figure 3 As shown in the figure, it can be seen that the GWO-LSTM model has a higher fit with the true value than the LSTM model, especially in the peak-to-valley fluctuations.

[0077] Evaluate the prediction performance of the GWO-LSTM model, specifically: evaluate the prediction performance of the GWO-LSTM model by the correlation coefficient R, relative mean absolute error rMAE and relative root mean square error rRMSE;

[0078] Among them, the correlation coefficient R is:

[0079]

[0080] The relative mean absolute error rMAE is:

[0081]

[0082] The relative root mean square error rRMSE is:

[0083]

[0084] Among them, G Pi is the predicted value of solar radiation intensity; is the average value of the solar radiation intensity prediction value series; G Mi is the actual measured value of solar radiation intensity; is the average value of the actual measurement value sequence of solar radiation intensity; N is the number of samples.

[0085] The solar radiation prediction model based on GWO-LSTM is a black box model. The solar radiation prediction model based on GWO-LSTM is established by selecting the current horizontal solar total radiation value G(t), solar azimuth A(t), ambient relative humidity RH(t), and outdoor dry-bulb temperature T at the prediction time in the photovoltaic power generation system database. w (t+i), outdoor relative humidity RH(t+i) and solar azimuth A(t+i) are used as input parameters of the GWO-LSTM solar radiation prediction model, and the solar radiation prediction model outputs the horizontal total solar radiation value G(t+i) at the prediction moment.

[0086] The photovoltaic power generation prediction model based on the back propagation neural network is a black box model; the solar radiation prediction value is used as an input parameter to establish a photovoltaic power generation prediction model based on the back propagation neural network to obtain the power generation of the photovoltaic power generation system, specifically: based on the photovoltaic power generation E(t), the horizontal solar total radiation value G(t), and the ambient relative humidity RH(t) at the current moment, the horizontal solar total radiation value G(t+i) and the ambient relative humidity RH(t+i) at the prediction moment are input as input parameters into the BP neural network photovoltaic power generation prediction model to obtain the photovoltaic power generation E(t+i) at the prediction moment;

[0087] See also Figure 4 The structure of the BP neural network photovoltaic power generation prediction model is that the number of input layer nodes is 5, the number of hidden layer neurons is 13, and the number of output layer nodes is 1. The comparison between the predicted value and the measured value of the BP neural network photovoltaic power generation prediction model is as follows: Figure 5 shown.

[0088] S104, on the demand side, based on the HVAC system cooling load database, a building cooling load prediction model based on the generalized regression neural network GRNN is established, and an indoor temperature prediction model is obtained based on the linear transformation of the cooling load prediction model;

[0089] The building cooling load prediction model is a black box model. The indoor temperature prediction model is obtained based on the linear transformation of the building cooling load prediction model. Specifically, the current cooling load Q(t) and indoor temperature T in the cooling load database of the HVAC system are selected. n (t), outdoor temperature T at the predicted time w (t+1), total solar radiation E(t+1), ambient relative humidity RH(t+1) and room temperature setting value T set As the input parameter of the cooling load prediction model, the cooling load prediction model predicts and outputs the cooling load prediction value Q(t+1);

[0090] The relationship between the predicted room temperature and the current room temperature and cooling load is obtained through linear regression, which simplifies the algorithm design and optimization time. The relationship is as follows:

[0091] T n (t+1)=0.244+0.993×T n (t)-0.003×Q(t+1)+0.003×Q(t)

[0092] Among them, in the GRNN cold load prediction model structure, the number of input layer neurons is determined by the sample dimension, and the number of input layer neurons in the present invention is 6. The number of pattern layer neurons is determined according to the number of training samples, and the number of pattern layer neurons in the present invention is 1863. In addition, the number of neurons in the summation layer is set to 2, and the number of neurons in the output layer is set to 1.

[0093] S105, constructing a photovoltaic power generation dynamic tracking model prediction control cost function based on the HVAC power consumption, photovoltaic power generation system power generation and indoor temperature prediction model;

[0094] The cost function takes into account two factors: the first factor aims to minimize the difference between the straight line formed by the photovoltaic power generation at the current moment and the predicted moment and the straight line change trend formed by the HVAC system power consumption at the current moment and the future moment. The second factor is related to the indoor temperature. The cost function of the optimized control system is shown below.

[0095]

[0096] Among them, P is the power consumption of the HVAC system; E is the power generation of the photovoltaic power generation system; λ and γ are weight coefficients, T set,i is the room temperature setting value at the i-th moment, ℃; T is the optimal indoor setting temperature, ℃; E(k+i-1) is the photovoltaic power generation at the current moment; E(k+i) is the photovoltaic power generation at the predicted moment; P(k+i-1) is the power consumption of the HVAC system at the current moment; P(k+i) is the power consumption of the HVAC system at the future moment.

[0097] S106, based on the genetic algorithm, the cost function is globally optimized online multiple times to minimize the difference between the HVAC system power consumption curve and the photovoltaic power generation curve and optimize the room temperature penalty caused by the difference to obtain the optimal control variables in the future limited time domain, and then optimize the calculation of the chilled water supply temperature, chilled water flow rate and cooling water return temperature, and feedback correction is performed on the system cooling capacity to complete the dynamic regulation of the photovoltaic power generation capacity of the air-conditioning system.

[0098] To optimize the variable T setAs the input of the cooling load prediction model, the cooling load prediction model predicts the output cooling capacity prediction value. The rolling optimization uses the genetic algorithm GA to repeatedly optimize the cost function J(t) globally online, and obtains the optimal control variable in the future limited time domain by minimizing the difference between the HVAC system power consumption curve and the photovoltaic power generation curve and the room temperature penalty caused by optimizing the difference. The first optimal control sequence u is used as the control parameter output. The optimized chilled water supply temperature T eo , Chilled water flow M e and cooling water return temperature T ci Acting on the controlled system, the controlled system includes the HVAC system and the indoor thermal environment. When the chilled water supply temperature T eo , Chilled water flow M e and cooling water return temperature T ci When input into the HVAC system and indoor thermal environment, the system outputs the final actual cooling capacity, which is compared with the predicted value of the prediction model to achieve feedback correction.

[0099] See also Figure 6 The present invention discloses a photovoltaic power generation dynamic control system for an air-conditioning system taking flexible load into account, comprising:

[0100] A collection module, wherein the collection module collects historical data of the photovoltaic power generation system and the HVAC system to build a HVAC system cold load database and a photovoltaic power generation system database;

[0101] A first construction module, wherein the first construction module constructs a mathematical model between the energy consumption of the energy-consuming equipment and the historical data based on the operating parameters in the collected HVAC historical data to obtain the HVAC power consumption;

[0102] An acquisition module, on the supply side, based on a photovoltaic power generation system database, uses a Grey Wolf Optimization Algorithm (GWO) to optimize a Long Short-Term Memory (LSTM) neural network algorithm, establishes a solar radiation prediction model based on GWO-LSTM, uses the solar radiation prediction value as an input parameter, establishes a photovoltaic power generation prediction model based on a back propagation neural network, and obtains the power generation of the photovoltaic power generation system;

[0103] A linear transformation module, which establishes a building cooling load prediction model based on a generalized regression neural network GRNN based on a cooling load database of a HVAC system on the demand side, and obtains an indoor temperature prediction model based on a linear transformation of the cooling load prediction model;

[0104] A second building module, wherein the second building module builds a photovoltaic power generation dynamic tracking model prediction control cost function based on the HVAC power consumption, photovoltaic power generation system power generation and indoor temperature prediction model;

[0105] The optimization module performs multiple global online optimizations on the cost function based on a genetic algorithm, minimizes the difference between the HVAC system power consumption curve and the photovoltaic power generation curve, and optimizes the room temperature penalty caused by the difference to obtain the optimal control variables in the future limited time domain, and then optimizes the calculation of the chilled water supply temperature, chilled water flow rate and cooling water return temperature, and feedback corrects the system cooling capacity to complete the dynamic regulation of the photovoltaic capacity of the air-conditioning system.

[0106] Example:

[0107] like Figure 7 As shown, the present invention discloses a method for dynamically controlling photovoltaic power generation of an air-conditioning system taking flexible load into consideration, comprising:

[0108] S1. Install monitoring equipment and collect the required operating parameters of the photovoltaic power generation system and HVAC system;

[0109] S2. Establish a mathematical model of energy consumption and operating parameters of the main energy-consuming equipment in the building HVAC system;

[0110] S3. Develop supply and demand forecasting model algorithm software;

[0111] S4. Constructing a photovoltaic capacity dynamic tracking model to predict the control cost function;

[0112] S5. Develop a controller for the flexible load dynamic control system of the air-conditioning system.

[0113] Preferably, S1 specifically includes: monitoring through existing monitoring equipment and newly added monitoring equipment, collecting historical data such as equipment operating parameters, indoor temperature parameters, outdoor meteorological parameters, solar photovoltaic power generation, building cooling load, etc., and establishing equipment mathematical model database, HVAC system cooling load database and photovoltaic power generation system database respectively.

[0114] Preferably, S2 is specifically: obtaining the energy efficiency model of the ground source heat pump unit and the energy consumption model of the circulating pump in this embodiment through a particle swarm optimization algorithm (PSO) based on theoretical or empirical formulas from the equipment operating parameters in the equipment mathematical model database.

[0115] Preferably, the algorithm for developing the supply and demand double-side prediction model in S3 is specifically as follows: on the supply side, based on the photovoltaic power generation system database, the Grey Wolf (GWO) optimization algorithm is used to optimize the long short-term memory (LSTM) neural network algorithm, and a solar radiation prediction model based on GWO-LSTM is established. Then, the solar radiation prediction value is used as an input parameter to establish a photovoltaic power generation prediction model based on the back propagation (BP) neural network; on the demand side, based on the HVAC system cooling load database, a building cooling load prediction model based on the generalized regression neural network (GRNN) is established, and then the indoor temperature prediction model is obtained based on the linear transformation of the cooling load prediction model. The relationship between the predicted room temperature and the current room temperature and cooling load is shown below.

[0116] T n (t+1)=0.244+0.993×T n (t)-0.003×Q(t+1)+0.003×Q(t)

[0117] like Figure 8 As shown, Figure 8 This is the idea of ​​the cost function of the present invention. Preferably, S4 is specifically: taking the photovoltaic power generation curve as a reference trajectory, making the change trend of the HVAC system power consumption curve as close to it as possible. This is achieved by ensuring that the straight line formed by the photovoltaic power generation E(k+i-1) at the current moment and the photovoltaic power generation E(k+i) at the predicted moment is as close as possible to the straight line formed by the power consumption of the HVAC system P(k+i-1) at the current moment and the power consumption of the HVAC system P(k+i) at the future moment. The cost function of the optimized control strategy is shown below.

[0118]

[0119] like Fig. 9 As shown, Fig. 9 is a control flow chart of the controller of the present invention. Preferably, S5 is specifically as follows: the present invention optimizes the variable T set As the input of the cooling load prediction model, the cooling load prediction model predicts the output cooling capacity prediction value. The rolling optimization uses the genetic algorithm GA to repeatedly optimize the cost function J(t) globally online, and obtains the optimal control variable in the future limited time domain by minimizing the difference between the HVAC system power consumption curve and the photovoltaic power generation curve and the room temperature penalty caused by optimizing the difference. The first optimal control sequence u is used as the control parameter output. The optimized chilled water supply temperature T eo , Chilled water flow M e and cooling water return temperature T ci Acting on the controlled system, the controlled system includes the HVAC system and the indoor thermal environment. When the chilled water supply temperature T eo , Chilled water flow M eand cooling water return temperature T ci When input into the HVAC system and indoor thermal environment, the system outputs the final actual cooling capacity, which is compared with the predicted value of the prediction model to achieve feedback correction.

[0120] Among them, T set is the room temperature set value, which is one of the input parameters of the cooling load prediction model and also a part of the cost function. set It is also the optimization parameter.

[0121] The present invention uses the HVAC system as the control research object. The HVAC flexible load can come from the floating room temperature set value, and the floating room temperature value will cause changes in the indoor thermal environment. The HVAC system is at the equipment level, and the mathematical model is used to establish the system's equipment model. The indoor thermal environment is at the demand level, and the fusion of the HVAC system and the indoor thermal environment uses a black box model, namely a load prediction model.

[0122] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for dynamically controlling photovoltaic power generation of an air-conditioning system taking flexible load into account, characterized in that: include: Collect historical data of photovoltaic power generation system and HVAC system, and build HVAC system cooling load database and photovoltaic power generation system database; Based on the operating parameters in the collected HVAC historical data, a mathematical model between the energy consumption of energy-consuming equipment and historical data is constructed to obtain the HVAC power consumption; On the supply side, based on the photovoltaic power generation system database, the Grey Wolf Optimization Algorithm GWO is used to optimize the Long Short-Term Memory LSTM neural network algorithm, and a solar radiation prediction model based on GWO-LSTM is established. The predicted solar radiation value is used as the input parameter to establish a photovoltaic power generation prediction model based on the back propagation neural network to obtain the power generation of the photovoltaic power generation system. On the demand side, based on the HVAC system cooling load database, a building cooling load prediction model based on generalized regression neural network GRNN is established, and the indoor temperature prediction model is obtained based on the linear transformation of the cooling load prediction model; Based on the prediction models of HVAC power consumption, photovoltaic power generation system power generation and indoor temperature, a photovoltaic power generation dynamic tracking model predicts the control cost function; Based on the genetic algorithm, the cost function is optimized globally online multiple times to minimize the difference between the power consumption curve of the HVAC system and the photovoltaic power generation curve, and to optimize the room temperature penalty caused by the difference to obtain the optimal control variables in the future limited time domain. Then, the chilled water supply temperature, chilled water flow rate and cooling water return temperature are optimized and calculated, and the system cooling capacity is feedback corrected to complete the dynamic regulation of the photovoltaic capacity of the air-conditioning system.

2. The method for dynamically controlling photovoltaic power generation of an air-conditioning system taking flexible load into account according to claim 1 is characterized in that: The historical data of the photovoltaic power generation system and the HVAC system include: equipment operation parameters, indoor temperature parameters, outdoor meteorological parameters, solar photovoltaic power generation and building cooling load, and a HVAC system cooling load database and a photovoltaic power generation system database are established respectively.

3. The method for dynamically controlling photovoltaic power generation of an air-conditioning system taking flexible load into account according to claim 2 is characterized in that: The mathematical model between the energy consumption of the energy-consuming equipment and the equipment operating parameters is constructed in detail by obtaining the mathematical model of the energy-consuming equipment through a particle swarm optimization algorithm PSO based on a theoretical or empirical formula.

4. The method for dynamically controlling photovoltaic power generation of an air-conditioning system taking flexible load into account according to claim 3 is characterized in that: The Grey Wolf Optimization Algorithm GWO is used to optimize the Long Short-Term Memory LSTM neural network algorithm, specifically: Collect data sets and preprocess them; Divide the preprocessed data set into training set and test set; Initialize the hyperparameters of the LSTM neural network and the gray wolf optimization algorithm respectively; input the training set into the gray wolf optimization algorithm, and output the optimal solution α wolf position as the GWO-LSTM prediction model in the hyperparameter input; Verify based on the test set, output the results, and evaluate the prediction performance of the GWO-LSTM model.

5. The method for dynamically controlling photovoltaic power generation of an air-conditioning system taking flexible load into account according to claim 4 is characterized in that: The evaluation of the prediction performance of the GWO-LSTM model is specifically as follows: the prediction performance of the GWO-LSTM model is evaluated by the correlation coefficient R, the relative mean absolute error rMAE and the relative root mean square error rRMSE respectively; Among them, the correlation coefficient R is: The relative mean absolute error rMAE is: The relative root mean square error rRMSE is: in, is the predicted value of solar radiation intensity; is the average value of the solar radiation intensity prediction value series; G Mi is the actual measured value of solar radiation intensity; is the average value of the actual measurement value sequence of solar radiation intensity; N is the number of samples.

6. The method for dynamically controlling photovoltaic power generation of an air-conditioning system taking flexible load into consideration according to claim 5 is characterized in that: The solar radiation prediction model based on GWO-LSTM is a black box model; the solar radiation prediction model based on GWO-LSTM is established by selecting the current horizontal solar total radiation value G(t), solar azimuth A(t), ambient relative humidity RH(t), and outdoor dry-bulb temperature T at the prediction time in the photovoltaic power generation system database. w (t+i), outdoor relative humidity RH(t+i) and solar azimuth A(t+i) are used as input parameters of the GWO-LSTM solar radiation prediction model, and the solar radiation prediction model outputs the horizontal total solar radiation value G(t+i) at the prediction moment.

7. The method for dynamically controlling photovoltaic power generation of an air-conditioning system taking flexible load into consideration according to claim 6, characterized in that: The photovoltaic power generation prediction model based on the back propagation neural network is a black box model; the photovoltaic power generation prediction model based on the back propagation neural network is established by taking the solar radiation prediction value as an input parameter to obtain the power generation of the photovoltaic power generation system, specifically: based on the photovoltaic power generation E(t), the horizontal solar total radiation value G(t), and the ambient relative humidity RH(t) at the current moment, the horizontal solar total radiation value G(t+i) and the ambient relative humidity RH(t+i) at the prediction moment are input as input parameters into the BP neural network photovoltaic power generation prediction model to obtain the photovoltaic power generation E(t+i) at the prediction moment; The BP neural network photovoltaic power generation prediction model structure has 5 input layer nodes, 13 hidden layer neurons, and 1 output layer node.

8. The method for dynamically controlling photovoltaic power generation of an air-conditioning system taking flexible load into account according to claim 7 is characterized in that: The building cooling load prediction model is a black box model; the indoor temperature prediction model is obtained based on the linear transformation of the building cooling load prediction model, specifically: the current cooling load Q(t) and indoor temperature T in the cooling load database of the HVAC system are selected. n (t), outdoor temperature T at the predicted time w (t+1), total solar radiation E(t+1), ambient relative humidity RH(t+1) and room temperature setting value T set As the input parameter of the cooling load prediction model, the cooling load prediction model predicts and outputs the cooling load prediction value Q(t+1); The relationship between the predicted room temperature and the current room temperature and cooling load is obtained through linear regression, which simplifies the algorithm design and optimization time. The relationship is as follows: T n (t+1)=0.244+0.993×T n (t)-0.003×Q(t+1)+0.003×Q(t)。 9. The method for dynamically controlling photovoltaic power generation of an air-conditioning system taking flexible load into account according to claim 8, characterized in that: The photovoltaic capacity dynamic tracking model predicts the control cost function, specifically: Among them, P is the power consumption of the HVAC system; E is the power generation of the photovoltaic power generation system; λ and γ are weight coefficients, T set,i is the room temperature setting value at the i-th moment, ℃; T is the optimal indoor setting temperature, ℃; E(k+i-1) is the photovoltaic power generation at the current moment; E(k+i) is the photovoltaic power generation at the predicted moment; P(k+i-1) is the power consumption of the HVAC system at the current moment; P(k+i) is the power consumption of the HVAC system at the future moment.

10. A photovoltaic power generation dynamic control system for an air conditioning system taking into account flexible loads, characterized in that: include: A collection module, wherein the collection module collects historical data of the photovoltaic power generation system and the HVAC system to build a HVAC system cold load database and a photovoltaic power generation system database; A first construction module, wherein the first construction module constructs a mathematical model between the energy consumption of the energy-consuming equipment and the historical data based on the operating parameters in the collected HVAC historical data to obtain the HVAC power consumption; An acquisition module, on the supply side, based on a photovoltaic power generation system database, uses a Grey Wolf Optimization Algorithm (GWO) to optimize a Long Short-Term Memory (LSTM) neural network algorithm, establishes a solar radiation prediction model based on GWO-LSTM, uses the solar radiation prediction value as an input parameter, establishes a photovoltaic power generation prediction model based on a back propagation neural network, and obtains the power generation of the photovoltaic power generation system; A linear transformation module, which establishes a building cooling load prediction model based on a generalized regression neural network GRNN based on a cooling load database of a HVAC system on the demand side, and obtains an indoor temperature prediction model based on a linear transformation of the cooling load prediction model; A second building module, wherein the second building module builds a photovoltaic power generation dynamic tracking model prediction control cost function based on the HVAC power consumption, photovoltaic power generation system power generation and indoor temperature prediction model; The optimization module performs multiple global online optimizations on the cost function based on a genetic algorithm, minimizes the difference between the HVAC system power consumption curve and the photovoltaic power generation curve, and optimizes the room temperature penalty caused by the difference to obtain the optimal control variables in the future limited time domain, and then optimizes the calculation of the chilled water supply temperature, chilled water flow rate and cooling water return temperature, and feedback corrects the system cooling capacity to complete the dynamic regulation of the photovoltaic capacity of the air-conditioning system.

Images