CN115630331A - Air-conditioning refrigeration station global energy-saving optimization and regulation and control method based on mathematical physical model - Google Patents
Air-conditioning refrigeration station global energy-saving optimization and regulation and control method based on mathematical physical model Download PDFInfo
- Publication number
- CN115630331A CN115630331A CN202211129220.5A CN202211129220A CN115630331A CN 115630331 A CN115630331 A CN 115630331A CN 202211129220 A CN202211129220 A CN 202211129220A CN 115630331 A CN115630331 A CN 115630331A
- Authority
- CN
- China
- Prior art keywords
- optimization
- prediction
- temperature
- water pump
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F24—HEATING; RANGES; VENTILATING
- F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
- F24F11/00—Control or safety arrangements
- F24F11/62—Control or safety arrangements characterised by the type of control or by internal processing, e.g. using fuzzy logic, adaptive control or estimation of values
- F24F11/63—Electronic processing
- F24F11/64—Electronic processing using pre-stored data
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F24—HEATING; RANGES; VENTILATING
- F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
- F24F11/00—Control or safety arrangements
- F24F11/88—Electrical aspects, e.g. circuits
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/086—Learning methods using evolutionary algorithms, e.g. genetic algorithms or genetic programming
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Signal Processing (AREA)
- Mechanical Engineering (AREA)
- Combustion & Propulsion (AREA)
- Chemical & Material Sciences (AREA)
- Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Data Mining & Analysis (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Physiology (AREA)
- Evolutionary Biology (AREA)
- Fuzzy Systems (AREA)
- Air Conditioning Control Device (AREA)
Abstract
The invention belongs to the technical field of energy-saving group control of air-conditioning refrigeration stations, and provides a global energy-saving optimization and regulation method of an air-conditioning refrigeration station based on a mathematical physical model, which comprises the steps of preprocessing running data of the refrigeration station, modeling a prediction variable, establishing a model of each device of the refrigeration station based on the mathematical physical model, and performing global optimization based on a genetic algorithm; and finally, establishing an optimized database. The invention can realize the online energy-saving group control of the refrigerating station system in the control period by optimizing the set values of different equipment and the distribution amount of the similar equipment on the basis of realizing the prediction of outdoor meteorological parameters and indoor dynamic loads. Compared with the energy consumption set value before optimization, the energy efficiency after optimization is greatly improved. Meanwhile, compared with the control period, the global optimization time consumption is greatly reduced, and the advanced optimization and setting of the next working condition can be realized in the control period.
Description
Technical Field
The invention relates to the technical field of energy-saving group control of air-conditioning refrigeration stations, in particular to a global energy-saving optimization and regulation method of an air-conditioning refrigeration station based on a mathematical physical model.
Background
Along with the development of Chinese economy and the sustainable development of China, the attention of people on large public buildings with energy conservation, emission reduction and consumption reduction is increasing. According to the report of the Chinese building energy-saving annual development research report 2022 published by the building energy-saving research center of Qinghua university in the first half of 2022, the common building area of China is about 140 hundred million m in 2020 2 The total energy consumption of public buildings is 3.46 hundred million tce, which accounts for 33 percent of the total energy consumption of the buildings.
The central air-conditioning system is an indispensable part of a public building, the energy consumption of the central air-conditioning system accounts for more than 40% of the total energy consumption of the public building, and the energy consumption of the refrigeration station system accounts for 60% -90% of the energy consumption of the air-conditioning system as the core of the central air-conditioning system. In addition, researches show that the comprehensive energy efficiency ratio EER of the refrigerating station can only reach the average level of 2.5-3.0 generally, and the average value of the central air-conditioning refrigerating station system EER of a part of developed countries reaches 4.0. Compared with developed countries, the central air-conditioning refrigeration station system in China has great energy-saving potential.
The refrigeration station system generally comprises a water chilling unit, a refrigeration water pump, a cooling tower and other equipment, and in actual engineering, different equipment does not operate independently, but has a certain nonlinear coupling relation with each other. Therefore, when considering the energy-saving operation of the refrigerating station, the operation cannot be performed only for a single device, and comprehensive consideration is required. Aiming at the global optimization of the refrigeration station, the overall optimization of the refrigeration station is taken as a target, and the realization of the combined control among equipment is the key for improving the comprehensive performance of the refrigeration station system. In view of the overall optimization control process of the refrigeration station system, firstly, the operation parameters of the equipment are reasonably set on the premise of ensuring the operation requirements of the system and the operation safety of the equipment, so as to achieve the aim of energy-saving operation. After the device setting value is determined, the state parameters of the similar devices need to be optimized, so as to achieve the goal of reducing the overall energy consumption of the similar device group to the maximum extent. However, most of the existing large-scale public building refrigerating station systems are based on simple logic control, control parameters of a plurality of devices are partially adjusted by manual experience, and the whole effective operation of the whole refrigerating station cannot be guaranteed.
Therefore, the invention combines the physical modeling theory and the genetic algorithm, provides a global energy-saving optimization and regulation method of the air-conditioning refrigeration station based on a mathematical physical model, and aims at global optimization energy-saving group control of the air-conditioning refrigeration station. On the basis of realizing the prediction of outdoor meteorological parameters and indoor dynamic loads, the online energy-saving group control of the refrigeration station system is realized in a control period by optimizing the set values of different equipment and the distribution amount of similar equipment.
Disclosure of Invention
The invention aims to provide a global energy-saving optimization and regulation and control method for an air-conditioning refrigeration station based on a mathematical physical model, in particular to a control system and an optimization method for various devices of the air-conditioning refrigeration station, and a set value optimization method and a supply distribution method capable of effectively reducing the energy consumption of an air-conditioning system of a large public building.
The technical scheme of the invention is as follows:
a global energy-saving optimization and regulation and control method of an air-conditioning refrigeration station based on a mathematical physical model is characterized in that on the basis of realizing prediction of outdoor meteorological parameters and indoor dynamic loads, online energy-saving group control of a refrigeration station system is realized in a control period by optimizing setting values of different equipment in the refrigeration station system and distribution amounts of similar equipment; the method comprises the following specific steps;
s1, preprocessing running data of a refrigerating station
S1.1, information acquisition and summarization
Collecting, summarizing and storing the parameter data; the energy consumption of the air-conditioning refrigeration station is related to various parameters, and the parameter data comprises characteristic parameters of each device, indoor and outdoor environment parameters, the temperatures of the freezing side inlet and the cooling side outlet, the evaporation temperature and the condensation temperature of the refrigerator, the load rate of the refrigerator and historical operation data of each device; the characteristic parameters of each device such as lift, power, refrigerating capacity, starting and stopping states and the like, indoor and outdoor environmental parameters and the like.
S1.2, deleting outlier singular points
After the required information is collected and summarized, the obtained data needs to be cleaned by deleting outliers, so that preparation is made for providing subsequent modeling precision. Here it is processed using an isolated forest algorithm.
The isolated forest algorithm is an unsupervised anomaly detection method suitable for continuous data, and the basic idea is to calculate the difference between objects, calculate the score of an anomaly point by detecting the distance between two points or the density of a certain region point, and the higher the score value is, the more likely the object is to be the anomaly point.
The abnormal detection step of the isolated forest algorithm is summarized into two steps:
1. training: sampling from the training set, and constructing an iTree;
2. and (3) testing: each iTree tree in the iForest forest is tested, the path length (path length h (x)) is recorded, and then anomalyscore (abnormal score s) of each piece of test data is calculated according to an abnormal score calculation formula.
Based on an isolated forest algorithm, carrying out data cleaning on the acquired parameter data by deleting outliers;
the formulas (1) - (3) are the calculation process of the isolated forest algorithm;
H(i)＝ln(i)+0.5772156649 (3)
wherein E (h (x)) is the mean of the path lengths of the sample x at t itrees, and h (x) is the path length; c (Ψ) is the average path length for constructing a BST binary search tree by Ψ samples; Ψ is the number of samples; h (i) is used to calculate H (Ψ -1), i representing a variable; s is an abnormal score, and the s value range is (0, 1); n is the number of nodes;
s1.3 selection of modeling input variables
Sorting the importance of the parameters processed in the step S1.2 through Relieff analysis, and selecting the independent parameters with large influence degree on energy consumption as modeling input variables, including outdoor wet bulb temperature T wb Total refrigerating capacity Q of system c Inlet temperature T of chilled water chws Temperature difference T between inlet and outlet of chilled water chws -T chwr Cooling water outlet temperature T ctws Temperature difference T between cooling water supply and return water ctws -T ctwr ；
The ReliefF analysis algorithm is a feature weighting algorithm, different weights are given to features according to the relevance of each feature and category, and features with weights less than a certain threshold value are removed. The implementation of the Relieff analysis algorithm comprises the following specific steps:
Step2 (1) randomly selecting a sample R i Sample book R i The k neighbor samples of the same class are marked as H i Sample R i The k neighbor samples in the heterogeneous samples are marked as M j (C) Repeatedly executing the specified iteration times;
(2) And updating the weight of each feature according to the weight formula (4).
And Step3, repeating the Step2 process for N times to output the feature weight W, sequencing the feature weights from high to low, and extracting the first d features for classification.
Formulas (4) and (5) are the calculation process of the Relieff analysis;
wherein W (A) is weight value, diff (A, R) 1 ,R 2 ) Represents a sample R 1 And sample R 2 Difference in characteristic A, R 1 | A | is a sample R 1 Value at characteristic A, R 2 | A | is a sample R 2 The value at feature a; m j (C) Represents the jth nearest neighbor sample in class C; m represents the number of sample samplings; p (C) represents the probability of the occurrence of the class C; r is i For a randomly drawn sample, sample R i The k similar neighbor samples are marked as H i ，H j The same process is carried out; k is the number of nearest neighbor samples; max (A), min (A) are the upper and lower limits of characteristic A respectively; class (R) is the category to which sample R belongs;
s1.4, determining the predictor variables
Based on the calculation result of the Relieff analysis and the basic theory, the indoor and outdoor parameters determine the overall cooling condition, and the parameters capable of representing the indoor and outdoor conditions are selected as the prediction variables T wb 、Q c ；
S1.5, determining optimization variables
Selecting 2 optimization variables at the freezing side and the cooling side respectively according to an optimization target; the freezing side and the cooling side respectively select any one of water supply temperature and return water temperature as an optimization variable, and two optimization variables are determined; determining the remaining 2 variables as optimization variables according to the six variables of the energy consumption correlation degree obtained in the step S1.3 and two selected from the six variables in the step S1.4 as prediction variables; determining the optimized variable as the inlet temperature T of the chilled water chws Temperature difference T between inlet and outlet of chilled water chws -T chwr Cooling water inlet temperature T ctws Temperature difference T between cooling water supply and return water ctws -T ctwr ；
S2: predictive variable modeling
Determining a predictor variable T wb 、Q c Then modeling is carried out on the model; the GRU model is adopted to carry out the operation of the load of the machine room and the outdoor meteorological parametersPredicting; the GRU model is one kind of recurrent neural network, and is proposed to solve the problems of long-term memory, gradient in back propagation and the like. When capturing long sequence semantic association, the GRU model can effectively inhibit gradient disappearance or explosion, the effect is superior to that of the traditional RNN, and the calculation complexity is small. Each GRU model memory unit has 2 control gates, namely an update gate Wr and a reset gate W z 。
S2.1, prediction calculation and flow of GRU model
The formulas (6) to (10) are the calculation process of the whole GRU model;
r t ＝σ(W r ·[h t-1 ,x t ]) (6)
z t ＝σ(W z ·[h t-1 ,x t ]) (7)
h t ＝(I-z t )×h t-1 +z t ×h t (9)
y t ＝σ(W o ·h t ) (10)
wherein x is t ,h t-1 ,h t ,r t ,z t ,
y t Respectively an input vector, a memory state at the previous moment, a state memory variable at the current moment, a state of an update gate, a state of a reset gate, a state of a current candidate set and an output vector at the current moment; w is a group of r ，W z ，W h ，W o Respectively, update gate, reset gate, candidate set, output vector and x t 、h t-1 The weight parameters multiplied by the formed connection matrix; i represents an identity matrix; []Representing a vector join; represents a matrix dot product; x represents the matrix product; σ denotes a sigmoid activation function for scaling a value to [0, 1%]To (c) to (d); tanh is an activation function for scaling a value to [ -1,1]To (c) to (d);
GRU modelAnd (3) a prediction process: the time sequence value of a prediction variable is used as the input of a GRU model, and the prediction variable T wb 、Q c And as output, the characteristics of the variables are fully utilized to construct a real-time prediction model. Firstly, normalization processing is carried out on the prediction variable data, and the influence on a GRU network model due to large magnitude difference among input data is avoided. Normalization is carried out by adopting a MinMax method, and the normalized data value domain is converted to [0,1 ]]. Then, initializing GRU model parameters randomly, then enabling an input layer to pass through a GRU recurrent neural network layer, and enabling a GRU unit to selectively reserve or forget information and continuously update the information in iteration;
s2.2 selection of prediction step size
According to the control period, the system quality parameter takes temperature or temperature difference as a representative, and the quantity parameter takes pressure or flow as a representative lag response time to preliminarily select a prediction step length; selecting an optimal prediction step length according to the prediction accuracy of the GRU model and whether the actual engineering requirements are met;
s2.3, parameter prediction
Predicting the prediction variable determined in S1.4 according to the optimal prediction step length selected in S2.2;
s3, establishing of each equipment model of refrigerating station based on mathematical physical model
S3.1 model of refrigerator
Wherein PLR is the partial load rate of the cooler, T e And T c Respectively the rated evaporation temperature and the rated condensation temperature of the refrigerator; q ch For refrigerating capacity of refrigerator, Q ch,r The rated refrigerating capacity of the refrigerator is set; t is a unit of er And T cr Respectively is the rated evaporating temperature and the rated condensing temperature of the cooler; a is 1 -a 10 Is a fitting coefficient;
s3.2 frozen water pump model
Wherein, P ch kW is the power of a refrigeration water pump; m is chw Flow rate of water pump, m chw,j The flow rate of the jth chilled water pump,
flow rate measured by sensor for jth chilled water pump, H chw Is the lift of a water pump H chw,j The delivery lift of the jth chilled water pump; g c Is a constant when m has the unit m 3 H, when the unit of H is m.H 2O, then g c 367.3; when m is in kg/s and H is in kPa, g is c Is 134.9; eta chw Efficiency of chilled water pumps; n is a radical of chwp The total number of chilled water pumps, N d,chwp,j The running state of the jth chilled water pump is shown; s. the 0 -S 3 Is a fitting coefficient; the partial load rate of the refrigerator can be indirectly measured by the temperature sensor and the flow sensor which are arranged at the position of the chilled water supply and return pipe of the refrigerator, and the controller can calculate the current actual running COP of the refrigerator according to the three obtained variables.
S3.3 Cooling Water Pump model
Wherein, P cwp The power of a freezing water pump is kW; m is a unit of cw,m Is the actual flow of the mth cooling water pump, m cw，nom,m The rated flow of the mth cooling water pump is set; PLR cwp,m The part load rate of the mth cooling water pump; p cwp,nom,m Rated power of the mth cooling water pump; c 0,m -C 3,m Is a fitting coefficient; n is a radical of hydrogen cwp For the total number of cooling water pumps, N d,cwp,m The start-stop state of the mth cooling water pump is a discrete variable consisting of 0 and 1, wherein 0 represents the shutdown of the equipment, and 1 represents the startup of the equipment;
s3.4 Cooling Tower model
Wherein Q is rej kW is the heat dissipation capacity of the cooling tower; g w For cooling water flow, m 3 /h；G a Is the air inlet volume G of the cooling tower fan a,r Air intake m of cooling tower fan in rated state 3 /h；T wi For the inlet water temperature, T, of the cooling tower wo The temperature of the outlet water of the cooling tower is DEG C; t is wb The temperature of outdoor air wet bulb is DEG C; c 1 ,C 2 ,C 3 Are all fitting coefficients; c. C p,w Specific heat of water, J/(kg. DEG C.); rho is the density of water, kg/m 3 (ii) a Pr is the power of the cooling tower fan in a rated state, and P is the power of the cooling tower fan in any state, KW; regulating the air intake of a fan according to the water outlet temperature and calculating the operation energy consumption of the fan of the cooling tower by using formulas (18), (19) and (20);
s4, global optimization based on genetic algorithm
Aiming at the global optimization problem of the central air-conditioning refrigeration station, a genetic algorithm is selected for solving. Genetic algorithm belongs to one of evolutionary algorithms, and the optimal solution is found by simulating the mechanism of selection and inheritance in nature. The method is mainly characterized in that the method directly operates the structural object without the limitation of derivation and function continuity; the method has the advantages of inherent hidden parallelism and better global optimization capability; by adopting a probabilistic optimization method, the optimized search space can be automatically acquired and guided without a determined rule, and the search direction can be adaptively adjusted. The method is suitable for solving the complex nonlinear problems of global optimization of the air conditioning system.
The genetic algorithm flow comprises seven steps:
step1: generating an initial population;
step2: calculating initial population fitness;
step3: selecting a parent population by adopting a roulette method;
step4: the parent population is crossed with each other to generate a new population;
step5: random variation of new individuals;
step6: calculating the population fitness of the offspring;
step7: whether a termination condition is reached.
The summary is as follows: firstly, before genetic algorithm calculation, setting population scale M, termination evolution algebra T of genetic algorithm and cross probability P c Probability of mutation P m (ii) a Then randomly generating an initial population, adaptively calculating the initial population, and if the initial population does not meet the stop condition, generating a child population by cross selection until the stop condition is met;
s4.2, determining optimization variables of set values
Determining an optimization variable of a set value according to S1.5; optimizing the optimization variables by a genetic algorithm under the same machine room load and outdoor meteorological parameters to realize the minimum sum of energy consumption of all equipment;
s4.3, determining optimization variables of supply quantity distribution
The supply amount allocated to each equipment is in a constraint range, and the formula (21) is a constraint condition of the supply amount of the equipment;
S min≤ S≤S max ,S∈{Q ch,i ,m chw,j ,m cw,m ,G a,n } (21)
S min and S max Respectively setting an upper limit and a lower limit of the supply quantity which can be borne by each device; q ch,i The refrigerating capacity of the ith water chilling unit is obtained; m is chw.j The flow rate of the jth chilled water pump is determined; m is cw,m The flow of the mth cooling water pump; g a,n The air quantity of the nth cooling tower fan is set;
s4.4, global optimization result
After the constrained global optimization problem, namely the formula (21), is converted into an unconstrained optimization problem, a heuristic algorithm is used for optimization.
Establishing an optimized database, wherein the specific steps of the optimized database are as follows;
step one, establishing a machine room load and outdoor meteorological parameter database meeting a prediction model according to historical data;
step two, on the basis of training the database established in the step one, establishing a prediction model by using a GRU method to predict future machine room load and future meteorological parameters;
thirdly, similarity comparison is carried out on the working condition to be optimized and historical data, set values and supply quantity parameters are selected according to the most similar principle, and set value optimization and supply quantity distribution are determined through a database without optimization each time;
and step four, communication and feedback of the global optimization result and the automatic control system.
The invention has the beneficial effects that: compared with the energy consumption set value before optimization, the energy consumption set value optimized by the air-conditioning refrigeration station global energy-saving optimization and regulation method based on the mathematical physical model is greatly improved. Meanwhile, compared with the control period, the global optimization time consumption is greatly reduced, and the advance optimization and setting of the next working condition can be realized in the control period. The invention can realize the online energy-saving group control of the refrigerating station system in the control period by optimizing the set values of different devices and the distribution quantity of the similar devices on the basis of realizing the prediction of outdoor meteorological parameters and indoor dynamic load.
Drawings
FIG. 1 is a flow chart of a global energy-saving optimization and regulation method of an air-conditioning refrigeration station based on a mathematical physical model;
FIG. 2 is a schematic diagram of a GRU model;
FIG. 3 is a genetic algorithm solving process;
fig. 4 is a general schematic diagram of an air-conditioning refrigeration station global energy-saving optimization and regulation method based on a mathematical physical model.
Detailed Description
The following detailed description of the invention refers to the accompanying drawings that illustrate specific embodiments of the invention. The following examples are intended to illustrate the invention, but are not intended to limit the scope of the invention.
Referring to fig. 1, the invention relates to a global energy-saving optimization and regulation method of an air-conditioning refrigeration station based on a mathematical physical model, taking a central air-conditioning refrigeration station of a public building in the large company city as an example, the method comprises the following specific steps:
s1, preprocessing of data center operation data
S1.1, information acquisition and summarization
Firstly, parameter data are required to be collected, summarized and stored, the energy consumption of the air-conditioning refrigerating station is related to various parameters, including the temperatures of the refrigerating and cooling side inlets and outlets, the evaporating temperature and the condensing temperature of a refrigerator, the load rate of the refrigerator, historical operation data of each device, characteristic parameters of each device, such as lift, power, refrigerating capacity, start-stop state and the like, indoor and outdoor environmental parameters and the like.
S1.2, deleting outlier singular points
After the required information is collected and summarized, the obtained data needs to be cleaned by deleting outliers, so that preparation is made for providing subsequent modeling precision. Here it is processed using an isolated forest algorithm.
The isolated forest algorithm is an unsupervised anomaly detection method suitable for continuous data, and the basic idea is to calculate the difference between objects, calculate the score of an anomaly point by detecting the distance between two points or the density of a certain region point, and the higher the score value is, the more likely the object is to be the anomaly point.
The abnormal detection step of the isolated forest algorithm is summarized into two steps:
1. training: sampling from the training set, and constructing an iTree;
2. and (3) testing: and testing each iTree tree in the iForest forest, recording the path length as h (x), and then calculating the abnormal score anomalyscore of each piece of test data as s according to an abnormal score calculation formula.
H(i)＝ln(i)+0.5772156649 (3)
Where E (h (x)) is the mean of the path lengths of the samples over t iTrees; c (Ψ) is the average path length for constructing a BST binary search tree using Ψ samples.
S1.3, selection of modeling input variables
Through Relieff analysis, the importance of the parameters is ranked, and independent parameters which obviously affect energy consumption are selected as modeling input variables. Respectively the temperature T of the outer wet bulb wb Total refrigerating capacity Q of system c Inlet temperature T of chilled water chws Temperature difference T between inlet and outlet of chilled water chws -T chwr Cooling water outlet temperature T ctws Temperature difference T for supplying and returning cooling water ctws -T ctwr 。
The ReliefF analysis algorithm is a feature weight algorithm, different weights are given to features according to the relevance of each feature and category, and features with weights smaller than a certain threshold value are removed. The implementation of the Relieff analysis algorithm comprises the following specific steps:
Step2 (1) randomly selecting a sample R i Sample book R i The k similar neighbor samples are marked as H i Sample R i The k neighbor samples in the heterogeneous samples are marked as M j (C) Repeatedly executing the specified iteration times;
(2) And updating the weight of each feature according to the weight formula (4).
And Step3, repeating the Step2 process for N times to output the feature weight W, sequencing the feature weights from high to low, and extracting the first d features for classification.
Wherein W (A) is weight value, diff (A, R) 1 ,R 2 ) Represents a sample R 1 And sample R 2 Difference in feature a; m is a group of j (C) Represents the jth nearest neighbor sample in class C; m represents the number of sample samplings; p (C) represents the probability of the occurrence of the class C.
S1.4, determining a predictive variable
According to the calculation result of the Relieff analysis and the basic theory (the indoor and outdoor parameters determine the overall cooling condition), the parameters representing the indoor and outdoor conditions are selected as the prediction variables. Determining T wb 、Q c Two parameters are used as prediction variables. Outdoor wet bulb temperature T wb Can represent outdoor working condition and total refrigerating capacity Q of system c Can represent the indoor demand condition.
S1.5, determining optimization variables
Selecting 2 optimization variables at the freezing side and the cooling side respectively according to an optimization target; the freezing side and the cooling side respectively select any one of water supply temperature and return water temperature as an optimization variable, and two optimization variables are determined; determining the remaining 2 variables as optimization variables according to the six variables of the energy consumption correlation degree obtained in the step S1.3 and two selected from the six variables in the step S1.4 as prediction variables; determining the optimized variable as the inlet temperature T of the chilled water chws Temperature difference T between inlet and outlet of chilled water chws -T chwr Cooling water inlet temperature T ctws Temperature difference T for supplying and returning cooling water ctws -T ctwr 。
S2: predictive variable modeling
Determining a predictor variable T wb 、Q c Then, it is modeled. A GRU model is adopted to predict the load of a machine room and outdoor meteorological parameters, is one of the recurrent neural networks, and is proposed for solving the problems of long-term memory, gradient in backward propagation and the like. When capturing long sequence semantic association, the GRU model can effectively inhibit gradient disappearance or explosion, the effect is superior to that of the traditional RNN, and the calculation complexity is small.
S2.1, prediction calculation and flow of GRU model
After the predictor variables are determined, they need to be modeled. And predicting the load of the machine room and the outdoor meteorological parameters by adopting a GRU model. Each GRU model memory unit has 2 control gates, namely an update gate Wr and a reset gate W z . Schematic diagram of GRU model referring to fig. 2, equations (6) - (10) are the calculation process of the entire GRU model.
r t ＝σ(W r ·[h t-1 ,x t ]) (6)
z t ＝σ(W z ·[h t-1 ,x t ]) (7)
h t ＝(I-z t )×h t-1 +z t ×h t (9)
y t ＝σ(W o ·h t ) (10)
Wherein x is t ,h t-1 ,h t ,r t ,z t ,
y t Respectively an input vector, a memory state at the last moment, a state memory variable at the current moment, a state of an update gate, a state of a reset gate, a current waitThe state of the selection set and the output vector at the current moment; w r ，W z ，W h ，W o Respectively, update gate, reset gate, candidate set, output vector and x t 、h t-1 The weight parameters multiplied by the formed connection matrix; i represents an identity matrix; []Representing a vector join; represents a matrix dot product; x represents the matrix product; σ denotes a sigmoid activation function for scaling a value to [0, 1%]In the middle of; tanh is an activation function for scaling a value to [ -1,1]In the meantime.
GRU model prediction process: the time sequence value of a prediction variable is used as the input of a GRU model, and the prediction variable T wb 、Q c And as output, the characteristics of the variables are fully utilized to construct a real-time prediction model. Firstly, the data is normalized, and the influence on a GRU network model due to large magnitude difference between input data is avoided. Normalization is carried out by adopting a MinMax method, and the normalized data value domain is converted into [0,1]. Then, model parameters are initialized randomly, then an input layer passes through a GRU recurrent neural network layer, and the GRU unit selectively reserves or forgets information and updates the information continuously in iteration.
S2.2, selection of parameter step size
According to the control period, the system quality parameter takes temperature or temperature difference as a representative, and the quantity parameter takes pressure or flow as a representative lag response time to preliminarily select a prediction step length; and selecting the optimal prediction step length according to the prediction accuracy of the GRU model and whether the actual engineering requirements are met.
S2.3, parameter prediction
And predicting the prediction variable determined in the S1.4 according to the optimal prediction step length selected in the S2.2.
S3, building of each equipment model of refrigerating station based on mathematical physical model
S3.1 model of refrigerator
Wherein PLR is the partial load rate of the refrigerator, T e And T c Respectively the rated evaporation temperature and the rated condensation temperature of the refrigerator; q ch For refrigerating capacity of refrigerator, Q ch,r The rated refrigerating capacity of the refrigerator is set; t is a unit of er And T cr Respectively is the rated evaporating temperature and the rated condensing temperature of the cooler; a is 1 -a 10 Is a fitting coefficient;
fitting coefficient:
the partial load rate of the refrigerator can be indirectly measured through a temperature sensor and a flow sensor which are arranged at a water supply and return pipe of chilled water of the refrigerator, and the controller can calculate the current actual running COP of the refrigerator according to the obtained three variables.
S3.2 model of chilled water pump
Wherein, P ch The power of a freezing water pump is kW; m is chw Flow rate of water pump, m chw,j The flow rate of the jth chilled water pump,
flow rate measured by sensor for jth chilled water pump, H chw Is the lift of a water pump H chw,j The head of the jth chilled water pump; g c Is a constant when m has the unit m 3 H, when the unit of H is m.H 2O, then g c 367.3; when m is in kg/s and H is in kPa, g is c Is 134.9; eta chw Efficiency of chilled water pump; n is a radical of chwp The total number of chilled water pumps, N d,chwp,j The running state of the jth chilled water pump is shown; s 0 -S 3 Is a fitting coefficient;
s3.3 Cooling Water Pump model
Wherein, P cwp The power of a freezing water pump is kW; m is cw,m Is the actual flow of the mth cooling water pump, m cw，nom,m The rated flow of the mth cooling water pump; PLR cwp,m The part load rate of the mth cooling water pump; p cwp,nom,m Rated power of the mth cooling water pump; c 0,m -C 3,m Is a fitting coefficient; n is a radical of cwp For the total number of cooling water pumps, N d,cwp,m The start-stop state of the mth cooling water pump is a discrete variable consisting of 0 and 1, wherein 0 represents the shutdown of the equipment, and 1 represents the startup of the equipment.
Fitting coefficient:
s3.4 Cooling Tower model
Wherein Q is rej kW is the heat dissipation capacity of the cooling tower; g w For cooling water flow, m 3 /h；G a Is the air inlet volume G of the cooling tower fan a,r Air intake m of cooling tower fan in rated state 3 /h；T wi For the inlet water temperature, T, of the cooling tower wo The temperature of the outlet water of the cooling tower is DEG C; t is a unit of wb The temperature of outdoor air wet bulb is DEG C; c 1 ,C 2 ,C 3 Are all fitting coefficients; c. C p,w Specific heat of water, J/(kg. DEG C.); rho is the density of water, kg/m 3 (ii) a Pr is the power of the cooling tower fan in a rated state, and P is the power of the cooling tower fan in any state, KW.
Fitting coefficient:
and (4) regulating the air inlet amount of the fan according to the outlet water temperature and calculating the operation energy consumption of the fan of the cooling tower by using formulas (18), (19) and (20).
S4, global optimization based on genetic algorithm
Aiming at the global optimization problem of the central air-conditioning refrigeration station, a genetic algorithm is selected for solving. Genetic algorithm belongs to one of evolutionary algorithms, and the optimal solution is found by simulating the mechanism of selection and inheritance in nature. The method is mainly characterized in that the method directly operates the structural object without derivation and function continuity limitation; the method has the advantages of inherent hidden parallelism and better global optimization capability; by adopting a probabilistic optimization method, the optimized search space can be automatically acquired and guided without a determined rule, and the search direction can be adaptively adjusted. The method is suitable for solving the complex nonlinear problems of global optimization of the air conditioning system.
S4.1, calculation flow of genetic algorithm
The genetic algorithm flow is shown in figure 3 and comprises seven steps:
step1: generating an initial population;
step2: calculating initial population fitness;
step3: selecting a parent population by adopting a roulette method;
step4: the parent population is mutually crossed to generate a new population;
step5: random variation of new individuals;
step6: calculating the population fitness of the offspring;
step7: whether a termination condition is reached.
Before the calculation of the genetic algorithm, the population scale M, the termination evolution algebra T of the genetic algorithm and the cross probability P are required to be calculated c Probability of mutation P m And carrying out reasonable setting.
S4.2, determining optimization variables of set values
And determining an optimization variable of the set value according to the S1.5. Under the same load of the machine room and outdoor meteorological parameters, optimization variables are optimized through a genetic algorithm, so that the energy consumption of all equipment is reduced to the maximum extent.
S4.3, determining optimization variables of supply quantity distribution
The supply amount allocated to each equipment is required to be within an acceptable range, and equation (21) is a constraint condition of the supply amount of the equipment.
S min≤ S≤S max ,S∈{Q ch,i ,m chw,j ,m cw,m ,G a,n } (21)
S min And S max Respectively setting an upper limit and a lower limit of the supply quantity which can be borne by each device; q ch,i The refrigerating capacity of the ith water chilling unit is obtained; m is chw.j The flow rate of the jth chilled water pump is determined; m is cw,m The flow of the mth cooling water pump is measured; g a,n The air quantity of the nth cooling tower fan is set;
s4.4, global optimization result
In order to improve the searching efficiency and accuracy of the algorithm, the constrained global optimization problem is converted into an unconstrained optimization problem. Then, a heuristic algorithm is used for optimization, and the energy efficiency of the optimized energy consumption set value is greatly improved compared with that before optimization. Meanwhile, compared with the control period, the global optimization time consumption is greatly reduced, and the advanced optimization and setting of the next working condition can be realized in the control period.
S5: optimizing database establishment
And S5.1, establishing a machine room load and outdoor meteorological parameter database meeting the prediction model according to the historical data.
And S5.2, on the basis of training of the database established in the step S5.1, establishing a prediction model by using a GRU (generalized regression Unit) method, and predicting future machine room load and future meteorological parameters.
And S5.3, comparing the similarity of the working condition to be optimized and the historical data, selecting the set value and the supply quantity parameter according to the most similar principle, and determining the set value optimization and the supply quantity distribution through the database without optimizing every time.
And S5.4, communication and feedback of the global optimization result and the automatic control system, and based on the method, the optimization efficiency and the response speed of the energy-saving group control can be effectively reduced.
Claims (2)
1. A global energy-saving optimization and regulation and control method of an air-conditioning refrigeration station based on a mathematical physical model is characterized in that the global energy-saving optimization and regulation and control method of the air-conditioning refrigeration station based on the mathematical physical model is used for realizing online energy-saving group control of a refrigeration station system in a control period by optimizing set values of different devices in the refrigeration station system and distribution amounts of similar devices on the basis of realizing prediction of outdoor meteorological parameters and indoor dynamic loads; the method comprises the following specific steps;
s1, preprocessing running data of a refrigerating station
S1.1, information acquisition and summarization
Collecting, summarizing and storing the parameter data; the parameter data comprises characteristic parameters of each device, indoor and outdoor environment parameters, the temperatures of the freezing and cooling sides and outlets, the evaporation temperature and the condensation temperature of the refrigerator, the load factor of the refrigerator and historical operation data of each device;
s1.2, deleting outlier singular points
Based on an isolated forest algorithm, carrying out data cleaning on the acquired parameter data by deleting outlier singular points;
the formulas (1) - (3) are the calculation process of the isolated forest algorithm;
H(i)＝ln(i)+0.5772156649 (3)
wherein E (h (x)) is the mean of the path lengths of the sample x at t itrees, and h (x) is the path length; c (Ψ) is the average path length for constructing a BST binary search tree using Ψ samples; Ψ is the number of samples; h (i) is used to calculate H (Ψ -1), i represents a variable; s is the anomaly score, and the s-value range is (0, 1); n is the number of nodes;
s1.3 selection of modeling input variables
Sorting the importance of the parameters processed in the step S1.2 through Relieff analysis, and selecting the independent parameters with large influence degree on energy consumption as modeling input variables, including outdoor wet bulb temperature T wb And total refrigerating capacity Q of system c Chilled water inlet temperature T chws Temperature difference T between inlet and outlet of chilled water chws -T chwr Cooling water outlet temperature T ctws Temperature difference T for supplying and returning cooling water ctws -T ctwr ；
Formulas (4) and (5) are the calculation process of the Relieff analysis;
wherein W (A) is weight value, diff (A, R) 1 ,R 2 ) Represents a sample R 1 And sample R 2 Difference in characteristic A, R 1 | A | is a sample R 1 Value at feature A, R 2 | A | is a sample R 2 The value at feature a; m j (C) Represents the jth nearest neighbor sample in class C; m represents the number of sample samplings; p (C) represents the probability of the occurrence of the class C; r i For a sample drawn at random, sample R i The k similar neighbor samples are marked as H i ，H j In the same way; k is the number of nearest neighbor samples; max (A), min (A) are the upper and lower limits of characteristic A respectively; class (R) is the category to which the sample R belongs;
s1.4, determining the predictor variables
Based on the calculation result and the basic theory of Relieff analysis, parameters capable of representing indoor and outdoor conditions are selected as a prediction variable T wb 、Q c ；
S1.5, determining optimization variables
Selecting 2 optimization variables from the freezing side and the cooling side respectively according to an optimization target; the freezing side and the cooling side respectively select any one of water supply temperature and return water temperature as an optimization variable, and two optimization variables are determined; determining the remaining 2 variables as optimization variables according to the six variables of the energy consumption correlation degree obtained by the S1.3 and two selected from the six variables of the S1.4 as prediction variables; determining the optimized variable as the inlet temperature T of the chilled water chws Temperature difference T between inlet and outlet of chilled water chws -T chwr Cooling water inlet temperature T ctws Temperature difference T between cooling water supply and return water ctws -T ctwr ；
S2: predictive variable modeling
Determining a predictor variable T wb 、Q c Then modeling is carried out on the model; predicting the load of the machine room and outdoor meteorological parameters by adopting a GRU model;
s2.1, prediction calculation and flow of GRU model
The formulas (6) to (10) are the calculation process of the whole GRU model;
r t ＝σ(W r ·[h t-1 ,x t ]) (6)
z t ＝σ(W z ·[h t-1 ,x t ]) (7)
h t ＝(I-z t )×h t-1 +z t ×h t (9)
y t ＝σ(W o ·h t ) (10)
wherein x is t ,h t-1 ,h t ,r t ,z t ,
y t Respectively an input vector, a memory state at the previous moment, a state memory variable at the current moment, a state of an update gate, a state of a reset gate, a state of a current candidate set and an output vector at the current moment; w r ，W z ，W h ，W o Respectively, update gate, reset gate, candidate set, output vector and x t 、h t-1 The weight parameter multiplied by the formed connection matrix; i represents an identity matrix; []Representing a vector concatenation; represents a matrix dot product; x represents the matrix product; σ denotes a sigmoid activation function for scaling a value to [0, 1%]To (c) to (d); tanh is an activation function for scaling a value to [ -1,1]To (c) to (d);
GRU model prediction process: the time sequence value of a prediction variable is used as the input of a GRU model, and the prediction variable T wb 、Q c As output, firstly, normalizing the prediction variable data, then randomly initializing GRU model parameters, then enabling an input layer to pass through a GRU circulating neural network layer, and enabling a GRU unit to selectively reserve or forget information and continuously update the information in iteration;
s2.2, selection of prediction step size
According to the control period, the system quality parameter takes temperature or temperature difference as a representative, and the quantity parameter takes pressure or flow as a representative lag response time to preliminarily select a prediction step length; selecting an optimal prediction step length according to the prediction accuracy of the GRU model and whether the actual engineering requirements are met;
s2.3, parameter prediction
Predicting the prediction variable determined in the S1.4 according to the optimal prediction step length selected in the S2.2;
s3, building of each equipment model of refrigerating station based on mathematical physical model
S3.1 refrigerator model
Wherein PLR is the partial load rate of the refrigerator, T e And T c Respectively the rated evaporation temperature and the rated condensation temperature of the refrigerator; q ch For refrigerating capacity of refrigerator, Q ch,r The rated refrigerating capacity of the refrigerator is set; t is er And T cr Respectively is the rated evaporating temperature and the rated condensing temperature of the cooler; a is 1 -a 10 Is a fitting coefficient;
s3.2 model of chilled water pump
Wherein, P ch The power of a freezing water pump is kW; m is chw Flow rate of water pump, m chw,j The flow rate of the jth chilled water pump,
flow rate measured by sensor for jth chilled water pump, H chw Is the lift of a water pump H chw,j The head of the jth chilled water pump; g c Is a constant when m has the unit m 3 H, when the unit of H is m.H 2O, then g c 367.3; when m is in kg/s and H is in kPa, g is c Is 134.9; eta chw Efficiency of chilled water pump; n is a radical of chwp The total number of chilled water pumps, N d,chwp,j The running state of the jth chilled water pump is shown; s. the 0 -S 3 Is a fitting coefficient;
s3.3 Cooling Water Pump model
Wherein, P cwp kW is the power of a refrigeration water pump; m is a unit of cw,m Is the actual flow of the mth cooling water pump, m cw，nom,m The rated flow of the mth cooling water pump is set; PLR cwp,m The part load rate of the mth cooling water pump; p is cwp,nom,m Rated power of the mth cooling water pump; c 0,m -C 3,m Is a fitting coefficient; n is a radical of hydrogen cwp For the total number of cooling water pumps, N d,cwp,m The start-stop state of the mth cooling water pump is a discrete variable composed of 0 and 1, wherein 0 represents that the equipment is stopped, and 1 represents that the equipment is started;
s3.4 Cooling Tower model
Wherein Q is rej kW is the heat dissipation capacity of the cooling tower; g w For cooling water flow, m 3 /h；G a The air inlet volume of a cooling tower fan, the air inlet volume of Ga, r in a rated state of the cooling tower fan, m 3 /h；T wi For the inlet water temperature, T, of the cooling tower wo The temperature of the outlet water of the cooling tower is DEG C; t is a unit of wb The temperature of outdoor air wet bulb is DEG C; c 1 ,C 2 ,C 3 Are all fitting coefficients; c. C p,w Specific heat of water, J/(kg. DEG C.); rho is the density of water, kg/m 3 (ii) a Pr is the power of the cooling tower fan in a rated state, and P is the power of the cooling tower fan in any state, KW; regulating the air intake of a fan according to the water outlet temperature and calculating the operation energy consumption of the fan of the cooling tower by using formulas (18), (19) and (20);
s4, global optimization based on genetic algorithm
The flow of the genetic algorithm is as follows: firstly, before genetic algorithm calculation, setting population scale M, termination evolution algebra T of genetic algorithm and cross probability P c Mutation probability P m (ii) a Then randomly generating an initial population, adaptively calculating the initial population, and when the initial population does not meet the stop condition, cross-selecting to generate a child population until the stop condition is met;
s4.2, determining optimization variables of set values
Determining an optimization variable of a set value according to S1.5; optimizing the optimization variables through a genetic algorithm under the same load of the machine room and outdoor meteorological parameters to realize the minimum sum of energy consumption of all equipment;
s4.3, determining optimization variables of supply quantity distribution
The supply amount allocated to each equipment is in a constraint range, and the formula (21) is a constraint condition of the supply amount of the equipment;
S min≤ S≤S max ,S∈{Q ch,i ,m chw,j ,m cw,m ,G a,n } (21)
S min and S max Respectively providing an upper limit set and a lower limit set of the supply quantity which can be borne by each device; q ch,i The refrigerating capacity of the ith water chilling unit is set; m is a unit of chw.j The flow rate of the jth chilled water pump is determined; m is cw,m The flow of the mth cooling water pump; g a,n The air quantity of the nth cooling tower fan is set;
s4.4, global optimization result
After the constrained global optimization problem, namely the formula (21), is converted into an unconstrained optimization problem, a heuristic algorithm is used for optimization.
2. The air-conditioning refrigeration station global energy-saving optimization and regulation and control method based on the mathematical physical model is characterized in that an optimization database is established, and the specific steps of the optimization database are as follows;
step one, establishing a machine room load and outdoor meteorological parameter database meeting a prediction model according to historical data;
secondly, on the basis of training the database established in the first step, establishing a prediction model by using a GRU (generalized regression unit) method to predict future machine room load and future meteorological parameters;
thirdly, similarity comparison is carried out on the working condition to be optimized and historical data, set values and supply quantity parameters are selected according to the most similar principle, and set value optimization and supply quantity distribution are determined through a database without optimization each time;
and step four, communication and feedback of the global optimization result and the automatic control system.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211129220.5A CN115630331A (en) | 2022-09-16 | 2022-09-16 | Air-conditioning refrigeration station global energy-saving optimization and regulation and control method based on mathematical physical model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211129220.5A CN115630331A (en) | 2022-09-16 | 2022-09-16 | Air-conditioning refrigeration station global energy-saving optimization and regulation and control method based on mathematical physical model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115630331A true CN115630331A (en) | 2023-01-20 |
Family
ID=84902772
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211129220.5A Pending CN115630331A (en) | 2022-09-16 | 2022-09-16 | Air-conditioning refrigeration station global energy-saving optimization and regulation and control method based on mathematical physical model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115630331A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117053355A (en) * | 2023-06-20 | 2023-11-14 | 江苏中屹能源技术有限公司 | Energy-saving control device for refrigerating station and control method thereof |
-
2022
- 2022-09-16 CN CN202211129220.5A patent/CN115630331A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117053355A (en) * | 2023-06-20 | 2023-11-14 | 江苏中屹能源技术有限公司 | Energy-saving control device for refrigerating station and control method thereof |
| CN117053355B (en) * | 2023-06-20 | 2024-04-02 | 江苏中屹能源技术有限公司 | Energy-saving control device for refrigerating station and control method thereof |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111256294A (en) | Model prediction-based optimization control method for combined operation of water chilling unit | |
| CN103912966B (en) | A kind of earth source heat pump refrigeration system optimal control method | |
| CN112415924A (en) | Energy-saving optimization method and system for air conditioning system | |
| CN113326651B (en) | Refrigerating station load and energy efficiency ratio dynamic modeling method based on T-S fuzzy model | |
| CN114484731B (en) | Central air conditioner fault diagnosis method and device based on stacking fusion algorithm | |
| CN110392515A (en) | A kind of Cooling and Heat Source equipment room energy-conserving control method and system based on historical data | |
| CN112628956B (en) | Water chilling unit load prediction control method and system based on edge cloud cooperative framework | |
| CN109634121A (en) | More parent genetic algorithm air source heat pump multiobjective optimization control methods based on radial basis function neural network | |
| Cui | District heating load prediction algorithm based on bidirectional long short-term memory network model | |
| CN113268913B (en) | Intelligent building air conditioner cooling machine system operation optimization method based on PSO-ELM algorithm | |
| CN113762387B (en) | Multi-element load prediction method for data center station based on hybrid model prediction | |
| CN116989442A (en) | Central air conditioner load prediction method and system | |
| CN115630331A (en) | Air-conditioning refrigeration station global energy-saving optimization and regulation and control method based on mathematical physical model | |
| Yang et al. | Research on energy-saving optimization of commercial central air-conditioning based on data mining algorithm | |
| Wang et al. | Multi-criteria comprehensive study on predictive algorithm of heating energy consumption of district heating station based on timeseries processing | |
| Guo et al. | A thermal response time ahead energy demand prediction strategy for building heating system using machine learning methods | |
| CN111473480A (en) | Central air conditioner energy-saving control method based on decision tree classification | |
| CN116045461B (en) | Energy-saving control method and device for air-cooled air conditioner based on water supply and return temperature adjustment | |
| CN115577828A (en) | Air conditioner refrigerating station system group control method based on data-driven modeling and optimization | |
| Qiao | Intelligent building with multi-energy system planning method considering energy supply reliability | |
| CN116499023A (en) | Intelligent control method and system for geothermal coupling solar heating station | |
| CN115013863B (en) | Autonomous optimization regulation and control method for heat supply system of jet pump based on digital twin model | |
| Fang et al. | Optimization of Air Conditioning Energy Consumption Based on Indoor Comfort Degree | |
| Zheng | Research on energy-saving control and optimisation of air conditioning system based on genetic algorithm | |
| Du et al. | Prediction of HVAC Energy Consumption Using PSO Optimized Deep Neural Network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |