A multi-stage equipment energy efficiency optimization control system and method for high-efficiency machine rooms

Abstract

The application discloses a kind of multistage equipment energy efficiency optimization control system and method for high efficiency machine room, it is related to equipment optimization technical field, the environmental parameter data and design parameter data of refrigeration machine room are obtained to analyze environmental correction factor, based on environmental correction factor, dynamic correction is carried out to cold load rate;Based on the delay time of preestablished, the history correction cold load rate sequence is reconstructed in phase space, the final embedding dimension is analyzed, and a hybrid prediction model is constructed based on the final embedding dimension;Based on the physical resonance characteristics of equipment, the optimal operating parameters are matched with the comprehensive energy efficiency index of machine room;The optimal operating parameters and the corresponding machine room comprehensive energy efficiency index are compressed in data structure, the data points are screened and analyzed, and the parameter library is updated.Reduce invalid data interference, so that the system can realize self-adaptive optimization in long-term operation, improve data storage and computing efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of device optimization, and in particular to a multi-stage device energy efficiency optimization control system and method for high-efficiency machine rooms. BACKGROUND

[0002] According to relevant statistics, building energy consumption in China accounts for about 30% of total social energy consumption, and in China, the energy-saving compliance rate of 21.5 billion square meters of existing public buildings is less than 10%, and the proportion of high-energy-consumption buildings in nearly 20 billion square meters of newly built buildings is more than 90% every year. As a major consumer of building energy, the air conditioning system accounts for about 47% of the total building energy consumption, and the energy consumption of the refrigeration machine room accounts for as high as 75% of the air conditioning energy consumption, reaching 11% of the total social energy consumption. In the industrial field accounting for 65% of the total social energy consumption, the energy consumption of the refrigeration machine room also accounts for a high proportion. Energy saving of high-efficiency machine rooms is the top priority of building and industrial energy saving and emission reduction.

[0003] The traditional refrigeration machine room control method is currently limited to controlling single devices, and the chilled water main unit, the refrigeration pump, the cooling pump, and the cooling tower are adjusted separately. The control strategy of the refrigeration pump, the cooling water pump, and the cooling tower fan is only the start-stop control and interlocking of the device itself, and there is no attention to the high energy efficiency of the chilled water unit, and there is a lack of optimization control of the overall energy efficiency of the refrigeration system. The refrigeration machine room is often in a non-stop running state, generating a large amount of running data, which is easy to cause data stacking and reduce the calculation efficiency.

[0004] Therefore, the present application discloses a multi-stage device energy efficiency optimization control system and method for high-efficiency machine rooms to solve the above problems. SUMMARY

[0005] The present application aims to provide a multi-stage device energy efficiency optimization control system and method for high-efficiency machine rooms to solve the problems in the prior art.

[0006] To achieve the above-mentioned purpose, the present application provides the following technical scheme: a multi-stage device energy efficiency optimization control method for high-efficiency machine rooms, which comprises the following steps:

[0007] S1: Obtain the environmental parameter data and design parameter data of the refrigeration machine room, analyze the environmental correction factor, and dynamically correct the cooling load rate based on the environmental correction factor;

[0008] S2: Based on the preset delay time, reconstruct the historical corrected cooling load rate sequence in phase space, analyze the final embedding dimension, and construct a hybrid prediction model based on the final embedding dimension;

[0009] S3: Based on the physical resonance characteristics of the device, match the optimal operating parameters with the comprehensive energy efficiency index of the machine room;

[0010] S4: compress the optimal operation parameters and corresponding machine room comprehensive energy efficiency indicators into a data structure, analyze and update the parameter library by screening the data points.

[0011] According to the above scheme, in S1, the following is included:

[0012] S101: Obtain the environmental parameter data and design parameter data of the refrigeration machine room, wherein the environmental parameter data includes the temperature T in of the water supply pipe, the temperature T out of the return water pipe, the water flow Q, the solar radiation intensity S, the ground wind speed V, the environmental humidity H, and the power consumption of the equipment; analyze the environmental correction factor a based on the environmental parameter data; a = (S / S d ) × (V d / V) β × exp[-k H (H-H d )] × cos[π(T wet -T wetd ) / 6]; wherein S d , V d , H d , and T wetd represent the solar radiation intensity, the ground wind speed, the environmental humidity, and the wet-bulb temperature under the design working condition respectively; β represents a preset ground wind speed correction coefficient; k H represents a preset humidity attenuation coefficient.

[0013] S102: dynamically correct the cooling load rate L' based on the environmental correction factor; L' = Q × (T in -T out ) × a ÷

[0014] (Q d ×△T d ); wherein Q d and△T d represent the flow and the temperature difference between the supply water pipe and the return water pipe under the design working condition; generate a historical corrected cooling load rate sequence {L'(t)|t∈[1,T]} according to the time sequence of the corrected cooling load rate; wherein T represents the total number of time.

[0015] The present application improves the accuracy of cooling load calculation and the adaptability to dynamic environmental changes by introducing environmental variables such as solar radiation, wind speed, humidity, and wet-bulb temperature; uses cosine function and exponential decay term to correct humidity and temperature, making the calculation result more smooth and in line with the actual situation, avoiding large calculation deviation caused by single factor fluctuation. By introducing the correction factor, the system load demand under different meteorological conditions is compensated, and the stability and energy efficiency of system operation are optimized.

[0016] According to the above scheme, in S2, the following is included:

[0017] S201: Embedding the historical corrected cold load rate sequence into a j-dimensional phase space for phase space reconstruction based on a preset delay time τ, and taking the i-th phase space vector as X i j = (L'(i), L'(i+τ), …, L'(i+(j-1)τ)); analyze the corresponding embedding degree coefficient wherein i ∈ [1, I], I represents the total number of phase space vectors; X i j +1 represents the phase space vector X i j represents the phase space vector X i j represents the phase space vector X i j+1 represents the phase space vector X i j represents the phase space vector X i j+1 represents the Euclidean distance between the phase space vectors X i j represents the phase space vector X i j+1 represents the phase space vector X i j represents the phase space vector X i j+1 represents the phase space vector X

[0018] S202: Constructing a hybrid prediction model based on the final embedding dimension, wherein the hybrid prediction model comprises a master system model and a slave system model, and the master system model is expressed as: dL / dt=θ[L'(t-τ)-L'(t)]; the slave system model is expressed as: dL* / dt=θ[L'(t-τ)-L*(t)]+K[L'(t)-L*(t)]; fitting the hybrid coefficient θ by using the least square method based on the historical corrected cold load rate sequence, and optimizing the coupling strength coefficient K by using the gradient descent method.

[0019] By using the master-slave system design, the synchronization coupling strength is dynamically adjusted by using the historical data, the prediction error is reduced, the accuracy of short-term load prediction is improved, the nonlinear complex system is better adapted, and the stability of prediction is improved.

[0020] According to the above scheme, in S3, the following contents are included:

[0021] S301: Analyzing the resonance frequency f of the device based on the equivalent stiffness and the equivalent mass of the devicer ;f r = 1 / (2π) x (ke / je) 1 / 2 ; wherein ke represents the equivalent stiffness of the device, which is inversely proportional to the real-time power consumption; je represents the equivalent mass, which is proportional to the cumulative running time;

[0022] analyzing a machine room comprehensive energy efficiency index, wherein the machine room comprehensive energy efficiency index is equal to the total refrigeration capacity provided by the refrigeration system at time t divided by the sum of the power consumptions of the chiller, the chilled water pump, the cooling water pump and the cooling tower fan at time t;

[0023] S302: constructing a device running frequency constraint condition corresponding to the machine room comprehensive energy efficiency index and a chilled water supply and return water temperature difference constraint condition, wherein the device running frequency constraint condition is that the difference between the actual running frequency of the device and the resonance frequency is less than the corresponding preset threshold value, and the chilled water supply and return water temperature difference constraint condition is that the chilled water supply and return water temperature difference is greater than the corresponding preset threshold value; mapping the optimization problem of the machine room comprehensive energy efficiency index into the Hamiltonian form of the Ising model, and solving the ground state of the Ising model by quantum annealing algorithm according to the device start-stop state and the energy efficiency coupling coefficient between devices to obtain the optimal parameter combination {Nch, ffp, fcp, fct} corresponding to time t; wherein Nch represents the number of chiller units, ffp represents the running frequency of the chilled water pump, fcp represents the running frequency of the cooling water pump, and fct represents the running frequency of the cooling tower fan.

[0024] the start-stop or running state of each device is represented by a binary spin variable σ m ∈{-1, +1}; wherein σ m = +1 indicates that the device m is in the running state, and σ m = -1 indicates that the device m is in the shutdown state, and maximizing the machine room comprehensive energy efficiency is equivalent to minimizing the Hamiltonian HM;

[0025]

[0026] wherein J (m，n) represents the cooperative energy efficiency relationship between device m and device n; h m represents the energy efficiency contribution of the independent running of the device;

[0027] J (m，n) = [SCOP pair (m, n) - SCOP single (m)] ÷ SCOP max ;

[0028] h m = SCOP single (m) ÷ SCOP max ;

[0029] SCOP pair (m, n) represents the energy efficiency value of the cooperative operation of device m and device n, SCOP single (m) represents the system comprehensive energy efficiency when device m operates independently; the system comprehensive energy efficiency when device m operates independently is equal to the refrigerating capacity provided by device m alone divided by the power consumption of device m alone; SCOP max represents the maximum value of the machine room comprehensive energy efficiency index;

[0030] randomly generate an initial spin parameter combination {σ m | m e [1, M]}; M represents the total number of devices; the initial transverse magnetic field of the quantum annealing machine causes the system to be in a superposition state of all possible states; the transverse magnetic field strength is gradually reduced, causing the system to evolve from a quantum superposition state to a classical state. When the time tends to the annealing time, the system collapses to the ground state with the lowest energy, corresponding to the optimal spin parameter combination. The device operation frequency constraint condition and the chilled water supply and return water temperature difference constraint condition are checked, and if they are not satisfied, the penalty mechanism is triggered and the solution is recalculated. According to the decoded optimal spin parameter combination, the optimal parameter combination corresponding to time t is obtained.

[0031] The application realizes dynamic calculation of the device resonance frequency by establishing an equivalent stiffness and equivalent mass model, combining device operation time and power consumption data, and improves the accuracy of device operation state evaluation; resonance matching control is used to ensure that the device operating frequency is as close as possible to the resonance frequency, reduce mechanical loss and energy consumption, and improve the service life of the device; by preferentially adjusting the device with the smallest frequency deviation, the coordinated operation strategy of the device group is optimized, and the stability and operation efficiency of the overall system are improved.

[0032] The resonance frequency is constructed using device physical parameters, and the dynamics of the device during operation is described through a physical model; combined with operation constraints, the optimization problem is converted into an Ising model, and the optimal device parameter combination per hour is obtained by solving the quantum annealing algorithm. The application considers the physical resonance effect of the device, and simultaneously optimizes the global collaboration in energy efficiency.

[0033] According to the above scheme, in S4, the following content is included:

[0034] S401: Obtain the optimal parameter combination at each time and generate a parameter vector corresponding to the machine room comprehensive energy efficiency index, and analyze the optimal linear reconstruction weight corresponding to the local field based on the parameter vector: wherein y r represents the parameter vector corresponding to time r; y q represents the parameter vector corresponding to time q; w (r，q) represents the linear reconstruction weight between the parameter vector y r and the parameter vector y q ; N(r) represents the parameter vector y rn nearest neighbors of the data point, n being a preset constant; constructing a low-dimensional embedding based on the obtained optimal linear reconstruction weight, mapping the parameter vector to a two-dimensional manifold to generate a data point;

[0035] S402: preset neighborhood radius and minimum sample number; for each data point, calculate the number of data points in the neighborhood; if the number is not less than the minimum sample number, mark the data point as a core point; otherwise, if the data point falls within the neighborhood of a core point, mark it as a boundary point; otherwise, mark the data point as a noise point; group the core points directly or indirectly connected from a core point and the boundary points in its neighborhood into the same cluster; label each data point with the corresponding cluster label, and label the data points not belonging to any cluster as noise points;

[0036] S403: calculate the mean vector for each cluster, analyze the Euclidean distance of each data point in the cluster to the cluster center based on the mean vector, and analyze the standard deviation of the Euclidean distance of all data points; delete the data points whose Euclidean distance to the cluster center is greater than a preset multiple of the standard deviation;

[0037] A lightweight database is generated, which includes a load rate interval, a device parameter combination, a predicted energy efficiency index, a parameter confidence, and an index table; the load rate interval is divided into intervals based on the distribution of the predicted load rate; the device parameter combination represents the optimal parameter combination of each cluster center; the predicted energy efficiency index represents the predicted energy efficiency index of each device parameter combination in the optimization process; the parameter confidence is equal to the probability of reaching or exceeding the predicted energy efficiency in actual operation; the database is sorted by load rate and an index table is generated.

[0038] The present application extracts key operating modes through dimensionality reduction processing of high-dimensional parameter space, improves the visualization and interpretability of data analysis, identifies different categories in device operating modes, automatically removes outliers, and improves the robustness of the model; only the parameter combination of the cluster center is retained for updating, reducing invalid data interference, so that the system can realize self-adaptive optimization in long-term operation, and improve data storage and calculation efficiency.

[0039] Another aspect of the present application is a multi-stage device energy efficiency optimization control system for high-efficiency machine rooms, which is applied to the above-mentioned multi-stage device energy efficiency optimization control system for high-efficiency machine rooms, and the system comprises a data acquisition correction module, a prediction model construction module, an operating parameter optimization module, and a data screening optimization module.

[0040] The data acquisition correction module is used to acquire environmental parameter data and design parameter data of the refrigeration machine room to analyze environmental correction factors, and dynamically corrects the cold load rate based on the environmental correction factors.

[0041] The prediction model construction module reconstructs the historical corrected cooling load rate sequence based on a preset delay time, analyzes the final embedding dimension, and constructs a hybrid prediction model based on the final embedding dimension.

[0042] The operation parameter optimization module matches the optimal operation parameter based on the physical resonance characteristics of the equipment and the comprehensive energy efficiency index of the machine room.

[0043] The data screening optimization module compresses the data structure of the optimal operation parameter and the corresponding comprehensive energy efficiency index of the machine room, screens and analyzes the data points, and updates the parameter library.

[0044] According to the above scheme, the data acquisition correction module includes a correction factor analysis unit and a dynamic correction unit.

[0045] The correction factor analysis unit analyzes the environmental correction factor based on the environmental parameter data and the design parameter data of the refrigeration machine room.

[0046] The dynamic correction unit dynamically corrects the cooling load rate based on the environmental correction factor, and generates a historical corrected cooling load rate sequence according to the time sequence of the corrected cooling load rate.

[0047] According to the above scheme, the prediction model construction module includes an embedding dimension analysis unit and a prediction model construction unit.

[0048] The embedding dimension analysis unit is used to reconstruct the historical corrected cooling load rate sequence based on a preset delay time; analyze the corresponding embedding degree coefficient, and confirm the final embedding dimension based on the embedding degree coefficient.

[0049] The prediction model construction unit is used to construct a hybrid prediction model based on the final embedding dimension; adopt the least square method to fit the mixing coefficient, and adopt the gradient descent method to optimize the coupling strength coefficient.

[0050] According to the above scheme, the operation parameter optimization module is used for the index analysis unit and the optimal parameter analysis unit.

[0051] The index analysis unit analyzes the resonance frequency of the equipment based on the equivalent stiffness and the equivalent mass of the equipment, and analyzes the comprehensive energy efficiency index of the machine room.

[0052] The optimal parameter analysis unit is used to construct the constraint condition corresponding to the comprehensive energy efficiency index of the machine room; maps the optimization problem of the comprehensive energy efficiency index of the machine room into the Hamiltonian form of the Ising model, and according to the start-stop state of the equipment and the energy efficiency coupling coefficient between the equipment, solves the ground state of the Ising model through quantum annealing algorithm, and obtains the optimal parameter combination.

[0053] According to the above scheme, the data screening optimization module includes a data dimension reduction unit and a data screening unit.

[0054] The data dimension reduction unit is used for obtaining an optimal parameter combination at each time and generating a parameter vector of a corresponding machine room comprehensive energy efficiency index, analyzing the optimal linear reconstruction weight in a local field based on the parameter vector; constructing a low-dimensional embedding based on the obtained optimal linear reconstruction weight, and mapping the parameter vector to a two-dimensional manifold to generate a data point;

[0055] The data screening unit is used for grouping the data points to generate clusters, and further analyzing and screening the data points in the clusters to generate a lightweight database.

[0056] Compared with the prior art, the application has the beneficial effects that: the application introduces environmental variables of solar radiation, wind speed, humidity, and wet-bulb temperature, improves the accuracy of cold load calculation, and improves the adaptability to dynamic environmental changes; the cosine function and the exponential decay term are used for humidity and temperature correction, so that the calculation result is smoother and more in line with the actual situation, and the calculation deviation caused by single factor fluctuation is avoided. Through the introduction of the correction factor, the demand of the system load under different meteorological conditions is compensated, and the stability and energy efficiency of the system operation are optimized; through the master-slave system design, the historical data is used to dynamically adjust the synchronous coupling strength, the prediction error is reduced, and the accuracy of short-term load prediction is improved; the application better adapts to the nonlinear complex system and improves the stability of prediction; the application establishes an equivalent stiffness and equivalent mass model, combines the equipment running time and power consumption data, realizes dynamic calculation of the equipment resonance frequency, and improves the accuracy of equipment running state evaluation; resonance matching control is adopted to ensure that the equipment running frequency is as close to the resonance frequency as possible, reduce mechanical loss and energy consumption, and improve equipment running life; by preferentially adjusting the equipment with the smallest frequency deviation, the coordinated operation strategy of the equipment group is optimized, and the stability and operation efficiency of the overall system are improved; the application reduces the dimension of the high-dimensional parameter space, extracts the key operation mode, improves the visualization and interpretability of data analysis, identifies different categories in the equipment operation mode, automatically eliminates abnormal points, and improves the robustness of the model; only the parameter combination of the cluster center is retained for updating, invalid data interference is reduced, the system can realize self-adaptive optimization in long-term operation, and data storage and calculation efficiency is improved. BRIEF DESCRIPTION OF DRAWINGS

[0057] The accompanying drawings are included to provide a further understanding of the application, and constitute a part of the specification, illustrate the application together with the embodiments of the application, and do not constitute a limitation on the application. In the drawings:

[0058] Figure 1 A flowchart of a multi-stage equipment energy efficiency optimization control method for a high-efficiency machine room of the application;

[0059] Figure 2 A structural schematic diagram of a multi-stage equipment energy efficiency optimization control system for a high-efficiency machine room of the application. DETAILED DESCRIPTION

[0060] The technical solutions in the embodiments of the present application will be clearly and completely described with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0061] Please refer to Figure 1 , the present application provides technical solutions: a multi-stage equipment energy efficiency optimization control system and method for high-efficiency machine room, to solve the problems in the prior art.

[0062] To achieve the above object, the present application provides the following technical solutions: a multi-stage equipment energy efficiency optimization control method for high-efficiency machine room, the method comprising the following steps:

[0063] S1: obtaining the environmental parameter data and design parameter data of the refrigeration machine room, and analyzing the environmental correction factor, and dynamically correcting the cooling load rate based on the environmental correction factor;

[0064] In S1, the following contents are included:

[0065] S101: obtaining the environmental parameter data and design parameter data of the refrigeration machine room, the environmental parameter data including the temperature T in of the water supply pipe, the temperature T out of the return water pipe, the water flow Q, the solar radiation intensity S, the ground surface wind speed V, the environmental humidity H and the equipment power consumption; based on the environmental parameter data, analyzing the environmental correction factor a; a=(S / S d )×(V d / V) β ×exp[-k H (H-H d )]×cos[π(T wet -T wetd ) / 6]; wherein S d , V d , H d and T wetd represent the solar radiation intensity, the ground surface wind speed, the environmental humidity and the wet-bulb temperature under the design working condition respectively; β represents the preset ground surface wind speed correction coefficient; k H represents the preset humidity attenuation coefficient;

[0066] S102: dynamically correcting the cooling load rate L' based on the environmental correction factor; L'=Q×(T in -T out )×a÷

[0067] (Q d ×△Td ); wherein Q d and ΔT d represent the flow rate and the temperature difference between the supply pipe and the return pipe under the design working condition; a history corrected cooling load rate sequence {L'(t)|t∈[1,T]} is generated according to the time sequence integration of the corrected cooling load rate; wherein T represents the total number of time points.

[0068] S2: based on the preset delay time, reconstructing the history corrected cooling load rate sequence in phase space, analyzing the final embedding dimension, and constructing a hybrid prediction model based on the final embedding dimension;

[0069] In S2, the following contents are included:

[0070] S201: based on the preset delay time τ, embedding the history corrected cooling load rate sequence into a j-dimensional phase space for phase space reconstruction, and taking the i-th phase space vector as X i j =(L'(i), L'(i+τ), …, L'(i+(j-1)τ)); analyzing the corresponding embedding degree coefficient wherein i∈[1,I], I represents the total number of phase space vectors; X i j +1 represents the phase space vector X i j the phase space vector corresponding to the j+1-dimensional phase space; ||X i j -X i j+1 ||represents the Euclidean distance between the phase space vector X i j and the phase space vector X i j+1 ; R represents a preset distance threshold; A() represents an indicator function, if ||X i j -X i j+1 ||>R, then A(||X i j -X i j+1 ||>R) = 1, otherwise 0; if the embedding degree coefficient EDC is less than the corresponding threshold, the corresponding dimension is taken as the final embedding dimension.

[0071] S202: Construct a hybrid prediction model based on the final embedding dimension, the hybrid prediction model includes a master system model and a slave system model, the master system model expression is: dL / dt=θ[L'(t-τ)-L'(t)]; the slave system model expression is: dL* / dt=θ[L'(t-τ)-L*(t)]+K[L'(t)-L*(t)]; based on the historical corrected cooling load rate sequence, the hybrid coefficient θ is fitted by the least square method, and the coupling strength coefficient K is optimized by the gradient descent method.

[0072] S3: Matching optimal operation parameters based on device physical resonance characteristics and combining with machine room comprehensive energy efficiency index;

[0073] In S3, the following contents are included:

[0074] S301: Analyzing the resonance frequency f of the device based on the equivalent stiffness and the equivalent mass of the device r ; f r =1 / (2π)×(ke / je) 1 / 2 ; wherein ke represents the equivalent stiffness of the device, the equivalent stiffness of the device is inversely proportional to the real-time power consumption; je represents the equivalent mass, the equivalent mass is proportional to the cumulative running time; analyzing the machine room comprehensive energy efficiency index, the machine room comprehensive energy efficiency index is equal to the total refrigeration capacity provided by the refrigeration system at time t divided by the sum of the power consumption of the chiller, the chilled water pump, the cooling water pump and the cooling tower fan at time t;

[0075] S302: Constructing the device operation frequency constraint condition corresponding to the machine room comprehensive energy efficiency index and the chilled water supply and return water temperature difference constraint condition, the device operation frequency constraint condition is that the difference between the actual operation frequency of the device and the resonance frequency is less than the corresponding preset threshold, and the chilled water supply and return water temperature difference constraint condition is that the chilled water supply and return water temperature difference is greater than the corresponding preset threshold; mapping the optimization problem of the machine room comprehensive energy efficiency index into the Hamiltonian form of the Ising model, and solving the ground state of the Ising model by quantum annealing algorithm according to the device start-stop state and the energy efficiency coupling coefficient between devices, to obtain the optimal parameter combination {Nch, ffp, fcp, fct} corresponding to time t; wherein Nch represents the number of chiller units, ffp represents the operation frequency of the chilled water pump, fcp represents the operation frequency of the cooling water pump, and fct represents the operation frequency of the cooling tower fan.

[0076] S4: Data structure compression is performed on the optimal operation parameters and the corresponding machine room comprehensive energy efficiency index, and the data points are analyzed and the parameter library is updated.

[0077] In S4, the following contents are included:

[0078] S401: Obtaining the optimal parameter combination at each time and the corresponding machine room comprehensive energy efficiency index to generate a parameter vector, and analyzing the optimal linear reconstruction weight in the local field based on the parameter vector: wherein, y r represents the parameter vector corresponding to time r; y q represents the parameter vector corresponding to time q; w (r，q) represents the linear reconstruction weight between the parameter vectors y r and y q ; N(r) represents the n nearest neighbors of the parameter vector y r , n being a preset constant; based on the obtained optimal linear reconstruction weight, a low-dimensional embedding is constructed to map the parameter vector to a two-dimensional manifold to generate a data point;

[0079] S402: a preset neighborhood radius and a minimum sample number; for each data point, the number of data points in the neighborhood of the data point is calculated; if the number is not less than the minimum sample number, the data point is marked as a core point; otherwise, if the data point falls within the neighborhood of a core point, it is marked as a boundary point; otherwise, the data point is marked as a noise point; the core points directly or indirectly connected to a core point and the boundary points in the neighborhood of the core point are classified into the same cluster; the corresponding cluster label is marked for each data point, and the data points not belonging to any cluster are marked as noise points;

[0080] S403: the mean vector of each cluster is calculated, the Euclidean distance of each data point in the cluster to the cluster center is analyzed based on the mean vector, and the standard deviation of the Euclidean distances of all data points is analyzed; the data points with a Euclidean distance to the cluster center greater than a preset multiple of the standard deviation are deleted;

[0081] Embodiment 1: for any two data points, if data point 1 is a core point and data point 2 falls within the neighborhood of data point 1, then data point 1 is directly connected to data point 2;

[0082] There is a chain DP1, DP2, …, DP N , wherein DP1 is a core point; and for any n, DP n+1 is directly connected to DP n ; if DQ is the last point in the chain, then DQ is indirectly connected to DP1.

[0083] A lightweight database is generated, which includes a load rate interval, a device parameter combination, a predicted energy efficiency index, a parameter confidence, and an index table; the load rate interval is divided into intervals based on the distribution of the predicted load rate; the device parameter combination represents the optimal parameter combination of each cluster center; the predicted energy efficiency index represents the predicted energy efficiency index of each device parameter combination in the optimization process; the parameter confidence is equal to the probability of reaching or exceeding the predicted energy efficiency in actual operation; the database is sorted according to the load rate and an index table is generated.

[0084] Please refer to Figure 2The application provides a technical scheme: a multi-stage equipment energy efficiency optimization control system for an efficient machine room, which comprises a data acquisition correction module, a prediction model construction module, an operation parameter optimization module and a data screening optimization module.

[0085] The data acquisition correction module is used for acquiring environmental parameter data and design parameter data of the refrigeration machine room to analyze an environmental correction factor, and dynamically correcting a cooling load rate based on the environmental correction factor.

[0086] The prediction model construction module reconstructs a historical corrected cooling load rate sequence in phase space based on a preset delay time, analyzes a final embedding dimension, and constructs a hybrid prediction model based on the final embedding dimension.

[0087] The operation parameter optimization module matches optimal operation parameters based on device physical resonance characteristics and a machine room comprehensive energy efficiency index.

[0088] The data screening optimization module compresses data structures of the optimal operation parameters and the corresponding machine room comprehensive energy efficiency index, screens and analyzes data points, and updates a parameter library.

[0089] The data acquisition correction module comprises a correction factor analysis unit and a dynamic correction unit.

[0090] The correction factor analysis unit analyzes an environmental correction factor based on environmental parameter data and design parameter data of the refrigeration machine room.

[0091] The dynamic correction unit dynamically corrects a cooling load rate based on the environmental correction factor, and generates a historical corrected cooling load rate sequence by integrating the corrected cooling load rate according to time sequence.

[0092] The prediction model construction module comprises an embedding dimension analysis unit and a prediction model construction unit.

[0093] The embedding dimension analysis unit is used for reconstructing a historical corrected cooling load rate sequence in phase space based on a preset delay time, analyzing a corresponding embedding degree coefficient, and confirming a final embedding dimension based on the embedding degree coefficient.

[0094] The prediction model construction unit is used for constructing a hybrid prediction model based on the final embedding dimension, fitting a hybrid coefficient by using a least square method, and optimizing a coupling strength coefficient by using a gradient descent method.

[0095] The operation parameter optimization module comprises an index analysis unit and an optimal parameter analysis unit.

[0096] The index analysis unit analyzes a resonance frequency of the device based on equivalent stiffness and equivalent mass of the device, and analyzes a machine room comprehensive energy efficiency index.

[0097] The optimal parameter analysis unit is used for constructing constraint conditions corresponding to the computer room comprehensive energy efficiency index; the optimization problem of the computer room comprehensive energy efficiency index is mapped into a Hamiltonian form of an Ising model, and the ground state of the Ising model is solved by a quantum annealing algorithm according to the equipment start-stop state and the energy efficiency coupling coefficient between equipment, so as to obtain an optimal parameter combination.

[0098] The data screening optimization module comprises a data dimension reduction unit and a data screening unit.

[0099] The data dimension reduction unit is used for obtaining an optimal parameter combination at each moment and generating a parameter vector corresponding to the computer room comprehensive energy efficiency index, analyzing optimal linear reconstruction weights in a local field based on the parameter vector, constructing a low-dimensional embedding based on the obtained optimal linear reconstruction weights, and mapping the parameter vector to a two-dimensional manifold to generate a data point.

[0100] The data screening unit is used for grouping data points to generate clusters, and further analyzing and screening data points in the clusters to generate a lightweight database.

[0101] It should be noted that, in this document, the terms such as first and second are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprise", "contain" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such a process, method, article or device.

[0102] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application. Therefore, the embodiments should be regarded as exemplary and non-limiting, and the scope of the present application is defined by the appended claims rather than the above description, and all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present application. Any reference signs in the claims should not be regarded as limiting the claims involved.

Claims

1. A method for multi-level equipment energy efficiency optimization control in high-efficiency computer rooms, characterized in that, The method includes the following steps: S1: Obtain environmental parameter data and design parameter data of the refrigeration room, analyze environmental correction factors, and dynamically correct the cooling load rate based on the environmental correction factors; S1 includes the following: S101: Obtain environmental parameter data and design parameter data for the refrigeration room. The environmental parameter data includes the water supply pipe temperature T. in , return water pipe temperature T out The environmental parameters include water flow rate Q, solar radiation intensity S, surface wind speed V, ambient humidity H, and equipment power consumption; an environmental correction factor α is analyzed based on these environmental parameter data. S102: Dynamically adjust the cooling load rate L' based on the environmental correction factor; L'=Q×(T in -T out )×α÷(Q) d ×△T d ); where Q d and △T d This represents the flow rate and the temperature difference between the supply and return water pipes under the design conditions; the historical corrected cooling load rate sequence {L'(t)|t∈[1,T]} is generated by integrating the corrected cooling load rate according to the time sequence; where T represents the total number of time points. S2: Based on the preset delay time, the historical corrected cooling load rate sequence is reconstructed in phase space, the final embedding dimension is analyzed, and a hybrid prediction model is constructed based on the final embedding dimension. S2 contains the following: S201: Based on a preset delay time τ, the historical corrected cooling load rate sequence is embedded into the j-dimensional phase space for phase space reconstruction, and the i-th phase space vector is denoted as X. i j = (L'(i), L'(i+τ), ..., L'(i+(j-1)τ)); Analyze the corresponding embedding degree coefficient EDC; if the embedding degree coefficient EDC is less than the corresponding threshold, the corresponding dimension is used as the final embedding dimension; S202: Construct a hybrid prediction model based on the final embedding dimension. The hybrid prediction model includes a master system model and a slave system model. Fit the hybrid coefficients using the least squares method based on the historical corrected cooling load rate sequence, and optimize the coupling strength coefficients using the gradient descent method. S3: Matching optimal operating parameters based on the physical resonance characteristics of the equipment and the comprehensive energy efficiency index of the computer room; S4: Compress the optimal operating parameters and corresponding comprehensive energy efficiency indicators of the computer room into a data structure, filter and analyze the data points, and update the parameter library; S4 includes the following: S401: Obtain the optimal parameter combination and corresponding comprehensive energy efficiency index of the data center at each time point to generate a parameter vector; analyze the optimal linear reconstruction weight in the local domain based on the parameter vector; construct a low-dimensional embedding based on the obtained optimal linear reconstruction weight, and map the parameter vector to a two-dimensional manifold to generate data points. S402: Preset neighborhood radius and minimum sample size; For each data point, calculate the number of data points in its neighborhood; If the number is not less than the minimum sample size, mark the data point as a core point; Otherwise, if the data point falls within the neighborhood of a core point, mark it as a boundary point; Otherwise, mark the data point as a noise point; Group core points directly or indirectly connected to a core point and their neighborhood boundary points into the same cluster; Mark each data point with its corresponding cluster label, and mark data points that do not belong to any cluster as noise points; S403: Calculate the mean vector for each cluster, analyze the Euclidean distance from each data point in the cluster to the cluster center based on the mean vector, and analyze the standard deviation of the Euclidean distances for all data points; delete data points whose Euclidean distance to the cluster center is greater than a preset number of standard deviations. A lightweight database is generated, comprising load rate intervals, equipment parameter combinations, predicted energy efficiency indicators, parameter confidence levels, and an index table. The load rate intervals are divided into intervals based on the distribution of predicted load rates. The equipment parameter combinations represent the optimal parameter combinations for each cluster center. The predicted energy efficiency indicators represent the predicted energy efficiency indicators for each equipment parameter combination during the optimization process. The parameter confidence level is equal to the probability of achieving or exceeding the predicted energy efficiency in actual operation. The database is then sorted according to load rate, and an index table is generated.

2. The method for multi-level equipment energy efficiency optimization control in a high-efficiency computer room according to claim 1, characterized in that: S3 includes the following: S301: Analysis of the resonant frequency f of the equipment based on its equivalent stiffness and equivalent mass. r ;Analyze the comprehensive energy efficiency index of the computer room, which is equal to the total cooling capacity provided by the refrigeration system at time t divided by the sum of the power consumption of the chiller, chilled water pump, cooling water pump and cooling tower fan at time t; S302: Construct the equipment operating frequency constraint and chilled water supply and return temperature difference constraint corresponding to the comprehensive energy efficiency index of the computer room. The equipment operating frequency constraint is that the difference between the actual operating frequency and the resonant frequency of the equipment is less than the corresponding preset threshold. The chilled water supply and return temperature difference constraint is that the chilled water supply and return temperature difference is greater than the corresponding preset threshold. The optimization problem of the comprehensive energy efficiency index of the computer room is mapped to the Hamiltonian form of the Ising model. Based on the start-up and shutdown status of the equipment and the energy efficiency coupling coefficient between the equipment, the ground state of the Ising model is solved by the quantum annealing algorithm to obtain the optimal parameter combination {Nch, ffp, fcp, fct} at time t; where Nch represents the number of chiller units, ffp represents the operating frequency of the chilled water pump, fcp represents the operating frequency of the cooling water pump, and fct represents the operating frequency of the cooling tower fan.

3. A multi-level equipment energy efficiency optimization control system for high-efficiency data centers, wherein the system is implemented using the multi-level equipment energy efficiency optimization control method for high-efficiency data centers as described in any one of claims 1-2, characterized in that... The system includes a data acquisition and correction module, a prediction model construction module, an operating parameter optimization module, and a data filtering and optimization module; The data acquisition and correction module is used to acquire environmental parameter data and design parameter data of the refrigeration room, analyze environmental correction factors, and dynamically correct the cooling load rate based on the environmental correction factors. The prediction model construction module reconstructs the phase space of the historical corrected cooling load rate sequence based on a preset delay time, analyzes the final embedding dimension, and constructs a hybrid prediction model based on the final embedding dimension. The operating parameter optimization module matches the optimal operating parameters based on the physical resonance characteristics of the equipment and the comprehensive energy efficiency index of the computer room. The data filtering and optimization module compresses the optimal operating parameters and corresponding comprehensive energy efficiency indicators of the data center into a data structure, filters and analyzes the data points, and updates the parameter database.

4. A multi-level equipment energy efficiency optimization control system for high-efficiency computer rooms according to claim 3, characterized in that: The data acquisition and correction module includes a correction factor analysis unit and a dynamic correction unit; The correction factor analysis unit analyzes environmental correction factors based on environmental parameter data and design parameter data of the refrigeration room; The dynamic correction unit dynamically corrects the cooling load rate based on the environmental correction factor, and generates a historical corrected cooling load rate sequence by integrating the corrected cooling load rates according to the time sequence.

5. A multi-level equipment energy efficiency optimization control system for high-efficiency computer rooms according to claim 3, characterized in that: The prediction model construction module includes an embedded dimension analysis unit and a prediction model construction unit; The embedded dimension analysis unit is used to reconstruct the phase space of the historical corrected cooling load rate sequence based on a preset delay time. Analyze the corresponding embedding degree coefficients and determine the final embedding dimension based on the embedding degree coefficients; The prediction model building unit is used to construct the hybrid prediction model with the final embedding dimension; the least squares method is used to fit the hybrid coefficients, and the gradient descent method is used to optimize the coupling strength coefficient.

6. A multi-level equipment energy efficiency optimization control system for high-efficiency computer rooms according to claim 3, characterized in that: The operating parameter optimization module is used for the indicator analysis unit and the optimal parameter analysis unit; The index analysis unit analyzes the resonance frequency of the equipment based on its equivalent stiffness and equivalent mass, and analyzes the comprehensive energy efficiency index of the computer room. The optimal parameter analysis unit is used to construct the constraints corresponding to the comprehensive energy efficiency index of the data center; the optimization problem of the comprehensive energy efficiency index of the data center is mapped into the Hamiltonian form of the Ising model, and the ground state of the Ising model is solved by quantum annealing algorithm according to the equipment start-up and shutdown state and the energy efficiency coupling coefficient between equipment to obtain the optimal parameter combination.

7. A multi-level equipment energy efficiency optimization control system for high-efficiency computer rooms according to claim 3, characterized in that: The data filtering and optimization module includes a data dimensionality reduction unit and a data filtering unit; The data dimensionality reduction unit is used to obtain the optimal parameter combination and the corresponding comprehensive energy efficiency index of the data center at each time point to generate a parameter vector. Based on the parameter vector, the optimal linear reconstruction weight in the local domain is analyzed. Based on the obtained optimal linear reconstruction weight, a low-dimensional embedding is constructed, and the parameter vector is mapped to a two-dimensional manifold to generate data points. The data filtering unit is used to group data points into clusters, and further analyze and filter the data points within the clusters to generate a lightweight database.

Images