An intelligent center comprehensive energy power distribution control system and control method

Abstract

The application discloses a kind of intelligent calculation center comprehensive energy distribution control system and control method, it is related to intelligent calculation center energy management technical field, it solves the technical problem that the prior art is often based on the priority of distribution area and distributes, lack of considering the energy type applied to the distribution selection of region, leading to the poor distribution effect of distribution area;By generating regional power consumption priority based on regional data;Based on regional data and energy data, generate regional energy priority;Based on regional power consumption priority and regional energy priority, generate distribution scheme;Based on the energy supply of each region in distribution scheme, while considering regional priority and regional energy priority, so that each region has optimal distribution scheme when distributing, improve the distribution effect of distribution area.

Description

Technical Field

[0001] This application belongs to the field of energy management technology for intelligent computing centers, specifically a comprehensive energy distribution control system and control method for intelligent computing centers. Background Technology

[0002] Intelligent computing centers are a new generation of computing infrastructure that integrates artificial intelligence computing, high-performance computing, and big data processing capabilities. They aim to provide powerful intelligent computing support for complex scenarios such as smart cities, autonomous driving, intelligent manufacturing, and scientific research. With the rapid development of artificial intelligence, big data, and high-performance computing, intelligent computing centers, as a new type of infrastructure, are accelerating their evolution towards large-scale, intensive, and intelligent operation. However, while providing powerful computing support, intelligent computing centers also face prominent challenges such as high energy consumption, complex power supply requirements, and high operating costs.

[0003] Existing technologies often distribute power based on the priority of the distribution area, without considering the applicable energy type of the area when selecting power distribution, resulting in poor power distribution performance in the distribution area; therefore, the integrated energy distribution control method for intelligent computing centers still needs further improvement. Summary of the Invention

[0004] This application aims to solve at least one of the technical problems existing in the prior art; to this end, this application proposes an integrated energy distribution control system and control method for intelligent computing centers, which is used to solve the technical problem that the prior art often distributes power according to the priority of the distribution area and lacks consideration of the energy type applicable to the area for power distribution selection, resulting in poor power distribution effect of the distribution area.

[0005] To achieve the above objectives, the first aspect of this application provides a method for integrated energy distribution control in a smart computing center, comprising:

[0006] Acquire regional and energy data from the intelligent computing center;

[0007] Regional electricity consumption priorities are generated based on regional data; these regional electricity consumption priorities quantify the priority of each region when distributing electricity.

[0008] Regional energy priorities are generated based on regional data and energy data; the regional energy priorities refer to the quantification of priority power distribution of several energy sources within a region.

[0009] Power distribution schemes are generated based on regional electricity consumption priorities and regional energy priorities.

[0010] Energy is supplied to various regions based on the power distribution plan.

[0011] This application constructs a multi-dimensional evaluation mechanism through the above steps, integrating two core indicators: regional priority and energy priority, to formulate personalized power distribution strategies for each zone, thereby significantly improving the power supply efficiency and resource utilization of the power distribution method.

[0012] Furthermore, the generation of regional electricity priority based on regional data includes:

[0013] The regional data includes the region ID, electricity consumption impact parameters, and time characteristics;

[0014] The electricity consumption impact parameters and time characteristics corresponding to the region ID are combined to form electricity consumption analysis data;

[0015] Extract historical electricity consumption analysis data corresponding to several region IDs;

[0016] The historical electricity consumption analysis data corresponding to the region ID are integrated into an electricity consumption prediction sequence corresponding to the region ID in chronological order.

[0017] The region ID and its corresponding electricity consumption forecast sequence are input into the load forecast model to obtain the regional electricity consumption forecast load corresponding to the region ID; the load forecast model is constructed using an LSTM model to predict the electricity consumption of different regions in the future.

[0018] Regional electricity consumption priorities are generated based on the predicted load of regional electricity consumption.

[0019] Furthermore, the load forecasting model is constructed using an LSTM model, including:

[0020] Obtain several historical electricity consumption forecast sequences and historical electricity loads corresponding to the region ID;

[0021] Several historical electricity consumption prediction sequences and historical electricity loads corresponding to the region ID are divided into training data, validation data and test data corresponding to the region ID; and the training data, validation data and test data corresponding to the region ID are preprocessed to obtain the training set, validation set and test set corresponding to the region ID.

[0022] The LSTM model was chosen as the base model for the region ID.

[0023] The base model corresponding to the region ID is trained on the training set corresponding to the region ID, and the learning rate and other hyperparameters are adjusted on the validation set corresponding to the region ID to obtain the pre-trained model corresponding to the region ID.

[0024] By validating the pre-trained model corresponding to the region ID on the test set corresponding to the region ID, a load prediction model is finally obtained with the region ID and its corresponding electricity consumption prediction sequence as input and the region electricity consumption prediction load corresponding to the region ID as output.

[0025] Furthermore, the generation of regional electricity priority based on regional electricity consumption forecast load includes:

[0026] Obtain several historical concurrent electricity loads corresponding to the predicted electricity load of the region with the region ID;

[0027] The historical load average LFJ corresponding to the region ID is calculated based on several historical electricity loads during the same period.

[0028] Calculate the regional electricity priority QYY using the formula. i The formula satisfies:

[0029] Where α1, α2, and α3 represent the load regulation coefficient, level regulation coefficient, and stability regulation coefficient, respectively, and α1, α2, and α3 are all greater than 0; i represents the number corresponding to the region ID, QYYF i Let QZX represent the predicted regional electricity load for the i-th region. i Let QYWX represent the regional importance of the i-th region. i This represents the regional power stability of the i-th region; LFJ i It is represented as the historical average load of the i-th region.

[0030] Furthermore, the regional power stability is obtained through the following methods:

[0031] Obtain the voltage values ​​V of several nodes within region ID;

[0032] Construct a voltage stability function DYF(V) to characterize the nonlinear relationship between voltage value and voltage stability. j The voltage stability function satisfies the following formula:

[0033] Where j represents the index of several nodes; max() represents the maximum value operation; max(j) represents the total number of nodes; V nom This is expressed as the rated voltage;

[0034] The voltage values ​​of several nodes are substituted into the voltage stability function to calculate the voltage stability VX corresponding to region ID;

[0035] Get the real-time frequency PL within the region ID;

[0036] A frequency stabilization function PLF(PL) is constructed to characterize the nonlinear relationship between real-time frequency and frequency stability; the frequency stabilization function satisfies the following formula:

[0037] Where k represents the attenuation coefficient, k>0; PLnom Represented as standard frequency;

[0038] Substitute the real-time frequency within region ID into the frequency stability function to calculate the frequency stability PWX corresponding to region ID.

[0039] The regional power consumption stability QYWX is obtained by weighted fusion calculation based on voltage stability VX and frequency stability PWX.

[0040] This application employs a comprehensive analysis of regional power consumption stability when calculating regional power consumption priority. Based on the voltage and frequency changes of each detection node within the region, it calculates regional power consumption stability, thereby improving the accuracy of regional power consumption stability and providing data support for the accurate calculation of regional power consumption priority.

[0041] Furthermore, the generation of regional energy priorities based on regional data and energy data includes:

[0042] The energy data includes energy ID and energy type; the energy type includes mains power, photovoltaic power, and energy storage.

[0043] Energy Dimension Scoring (NWP) Based on Energy Data n ;

[0044] Through formula Calculate the regional energy priority QNY; where i represents the region ID, m represents the energy ID, n represents the energy dimension score, and N represents the total number of energy dimension scores; QX m,n This represents the dimension weight coefficient corresponding to the energy dimension score, QP. m >0; NWP n This is represented as the score for the nth energy dimension.

[0045] Furthermore, the generation of energy dimension scores based on energy data includes:

[0046] The energy data includes energy price NJ, response time XS, carbon emission intensity TQ, and energy availability NYL;

[0047] An economic scoring function JJF(NJ) is constructed to characterize the nonlinear relationship between energy prices and economic scores; the economic scoring function satisfies the following formula:

[0048] Where t represents the price sensitivity coefficient, t>0; NJ base This is represented as the benchmark electricity price;

[0049] The economic score corresponding to the energy ID is calculated by substituting the energy price into the economic scoring function;

[0050] A reliability scoring function KKF(XS, NYL) is constructed to characterize the nonlinear relationship between response time, energy availability, and reliability score, respectively; the reliability scoring function satisfies the following formula:

[0051] Among them, XS max This is expressed as the maximum tolerable response time;

[0052] The reliability score corresponding to the energy ID is calculated by substituting the response time and energy availability into the reliability scoring function;

[0053] A cleanliness score function QJF(TQ) is constructed to characterize the nonlinear relationship between carbon emission intensity and cleanliness score; the cleanliness score function satisfies the following formula:

[0054] Among them, TQ max This is expressed as the maximum carbon emission standard;

[0055] The cleanliness score corresponding to the energy ID is calculated by substituting the carbon emission intensity into the cleanliness scoring function;

[0056] The energy dimension score includes an economic score, a reliability score, and a cleanliness score.

[0057] This application analyzes energy data from multiple perspectives to find the optimal selection relationship between regions and energy types, providing accurate data support for regional energy prioritization.

[0058] Furthermore, the dimension weight coefficients are generated in the following ways:

[0059] Obtain data on energy prices, regional electricity stability, carbon allowance balances, and the environment;

[0060] Integrate energy prices, regional electricity stability, carbon allowance margins, and environmental data into dimensional weighted analysis data;

[0061] The dimensional weight analysis data is input into the dimensional weight generation model to obtain several dimensional weight coefficients; the dimensional weight generation model is constructed through a machine learning model and is used to predict the dimensional weight coefficients corresponding to several energy dimensional scores.

[0062] Furthermore, the dimension weight generation model is constructed using a machine learning model, including:

[0063] Obtain historical dimensional weight analysis data and their corresponding dimensional weight coefficients;

[0064] The historical dimensional weight analysis data and their corresponding dimensional weight coefficients are divided into training data, validation data, and test data; and the training data, validation data, and test data are preprocessed to obtain the training set, validation set, and test set.

[0065] Choose a machine learning model as the base model;

[0066] The base model is trained on the training set, and the learning rate and other hyperparameters are adjusted on the validation set to obtain the pre-trained model.

[0067] By validating the pre-trained model on the test set, a dimensional weight generation model is finally obtained, which takes dimensional weight analysis data as input and outputs several dimensional weight coefficients.

[0068] This application considers multiple dimensions of energy data when calculating regional energy priority, and dynamically adjusts the corresponding dimension weight coefficients for each dimension score through a pre-trained dimension weight generation model. This enables the calculation of regional energy priority to improve the energy priority of each region based on real-time regional data, providing accurate data support for subsequent regional and energy supply.

[0069] Furthermore, the generation of a power distribution scheme based on regional electricity consumption priority and regional energy priority includes:

[0070] The region ID and its corresponding regional electricity priority, as well as the energy priorities of several regions corresponding to the region ID, are integrated into power distribution analysis data.

[0071] Input the power distribution analysis data into the scheme generation model to obtain the power distribution scheme corresponding to each area ID;

[0072] The proposed solution generation model is constructed using a large model, including:

[0073] Obtain historical power distribution analysis data and their corresponding power distribution schemes;

[0074] Several historical power distribution analysis data and their corresponding power distribution schemes are divided into training data, validation data, and test data; and the training data, validation data, and test data are preprocessed to obtain training sets, validation sets, and test sets.

[0075] Choose a large model as the base model;

[0076] The base model is trained on the training set, and the learning rate and other hyperparameters are adjusted on the validation set to obtain the pre-trained model.

[0077] By validating the pre-trained model on the test set, a scheme generation model is finally obtained, which takes power distribution analysis data as input and outputs power distribution schemes corresponding to region IDs.

[0078] Another aspect of the present invention provides an integrated energy distribution control system for an intelligent computing center, comprising: a data acquisition module, a data analysis module, and a control module; the data acquisition module and the data analysis module are connected; the data analysis module and the control module are connected.

[0079] The data acquisition module acquires data from several areas of the intelligent computing center and energy data through data acquisition equipment.

[0080] The data analysis module: generates regional electricity consumption priorities based on regional data; generates regional energy priorities based on regional data and energy data; and generates power distribution schemes based on regional electricity consumption priorities and regional energy priorities.

[0081] The control module supplies energy to each area based on the power distribution scheme.

[0082] Compared with the prior art, the beneficial effects of this application are:

[0083] 1. This application generates regional electricity priority based on regional data; generates regional energy priority based on regional data and energy data; generates a power distribution scheme based on regional electricity priority and regional energy priority; and supplies energy to each region based on the power distribution scheme, while considering both regional priority and regional energy priority, so that each region has an optimal power distribution scheme when distributing electricity, thereby improving the power distribution effect of the distribution area.

[0084] 2. When calculating the power priority of each region, this application not only considers the total load of each region in the future, but also the importance of each region itself and the power stability of each region. It analyzes each region from multiple perspectives to accurately calculate the regional power priority and provide accurate data support for subsequent power allocation.

[0085] 3. When calculating the regional energy priority for each region, this application not only considers the compatibility between each energy type and the region, but also analyzes the energy data from multiple dimensions and dynamically adjusts the weight coefficients between each dimension, so as to make the compatibility between each energy type and the region better and improve the accuracy of regional energy priority. Attached Figure Description

[0086] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0087] Figure 1 This is a flowchart of a smart computing center integrated energy distribution control method according to this application;

[0088] Figure 2 This is a schematic diagram of the principle of an integrated energy distribution control system for a smart computing center according to this application. Detailed Implementation

[0089] The technical solutions of this application will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0090] Please see Figure 1 The first aspect of this application provides a method for integrated energy distribution control in a smart computing center, comprising:

[0091] Acquire regional and energy data from the intelligent computing center;

[0092] Regional electricity consumption priorities are generated based on regional data; regional electricity consumption priorities quantify the priority of each region when distributing electricity.

[0093] Regional energy priorities are generated based on regional and energy data; regional energy priorities refer to the quantification of prioritizing the distribution of several energy sources within a region.

[0094] Power distribution schemes are generated based on regional electricity consumption priorities and regional energy priorities.

[0095] Energy is supplied to various regions based on the power distribution plan.

[0096] In this embodiment, the generation of regional electricity consumption priorities based on regional data includes:

[0097] Regional data includes region ID, electricity consumption impact parameters, and time characteristics;

[0098] The electricity consumption impact parameters and time characteristics corresponding to the region ID are combined to form electricity consumption analysis data;

[0099] Extract historical electricity consumption analysis data corresponding to several region IDs;

[0100] Several historical electricity consumption analysis data corresponding to a region ID are integrated into an electricity consumption forecast sequence corresponding to the region ID in chronological order; the electricity consumption forecast sequence refers to the impact data required to predict the future regional electricity load.

[0101] The region ID and its corresponding electricity consumption forecast sequence are input into the load forecast model to obtain the region electricity consumption forecast load corresponding to the region ID; the load forecast model is constructed using an LSTM model to predict the electricity consumption of different regions in the future.

[0102] Regional electricity consumption priorities are generated based on the predicted load of regional electricity consumption.

[0103] The load forecasting model in this embodiment is constructed using an LSTM model, including:

[0104] Obtain several historical electricity consumption forecast sequences and historical electricity loads corresponding to the region ID;

[0105] Several historical electricity consumption prediction sequences and historical electricity loads corresponding to the region ID are divided into training data, validation data, and test data corresponding to the region ID; and the training data, validation data, and test data corresponding to the region ID are preprocessed to obtain the training set, validation set, and test set corresponding to the region ID; the ratio between the training set, test set, and validation set is 7:2:1.

[0106] The LSTM model was chosen as the base model for the region ID.

[0107] The base model corresponding to the region ID is trained on the training set corresponding to the region ID, and the learning rate and other hyperparameters are adjusted on the validation set corresponding to the region ID to obtain the pre-trained model corresponding to the region ID.

[0108] By validating the pre-trained model corresponding to the region ID on the test set corresponding to the region ID, a load prediction model is finally obtained with the region ID and its corresponding electricity consumption prediction sequence as input and the region electricity consumption prediction load corresponding to the region ID as output.

[0109] In this embodiment, the generation of regional electricity consumption priorities based on regional electricity consumption forecast load includes:

[0110] Obtain several historical concurrent electricity loads corresponding to the predicted electricity load of the region with the region ID;

[0111] The historical load average LFJ corresponding to the region ID is calculated based on several historical same-period electricity loads. In this implementation, 10 historical same-period electricity loads are selected for calculating the historical load average, and these 10 historical same-period electricity loads are the 10 historical same-period loads closest to the current one.

[0112] Calculate the regional electricity priority QYY using the formula. i The formula satisfies:

[0113] Wherein, α1, α2, and α3 represent the load regulation coefficient, level regulation coefficient, and stability regulation coefficient, respectively. α1, α2, and α3 are all greater than 0, and their specific values ​​are set based on experience. In this embodiment, α1, α2, and α3 are set to 1.3, 0.8, and 0.7, respectively. The purpose of setting α1, α2, and α3 is to control the sensitivity of each parameter. For example, α1 amplifies the impact of load fluctuations, α2 suppresses high-level monopolies, and α3 enhances the weight of power stability. i represents the number corresponding to the region ID, QYYF. i Let QZX represent the predicted regional electricity load for the i-th region. i Let represent the regional importance of the i-th region. The regional importance is set empirically. In this embodiment, the intelligent computing center is divided into three regions: a GPU computing region, a liquid cooling system region, and a network server room region. The regional importance of each region is set to 0.95, 0.8, and 0.7, respectively. i This represents the regional power stability of the i-th region; LFJ i It is represented as the historical average load of the i-th region.

[0114] In calculating the electricity priority of each region, this embodiment comprehensively evaluates the total load demand for future periods, the importance level of the region, and power stability parameters. By constructing a multi-dimensional analysis model, it accurately quantifies the electricity priority of each region, providing a high-precision decision-making basis for the intelligent power distribution system.

[0115] The regional power supply stability in this embodiment is achieved through the following methods:

[0116] Obtain the voltage values ​​V of several nodes within region ID;

[0117] Construct a voltage stability function DYF(V) to characterize the nonlinear relationship between voltage value and voltage stability. j The voltage stability function satisfies the following formula:

[0118] Where j represents the index of several nodes; max() represents the maximum value operation; max(j) represents the total number of nodes; V nom This is expressed as the rated voltage; the specific value is set based on experience. In this embodiment, V is... nom Set to 415 volts; the smaller the voltage deviation at several nodes, the closer the voltage stability is to 1, and the more stable the voltage.

[0119] The voltage values ​​of several nodes are substituted into the voltage stability function to calculate the voltage stability VX corresponding to region ID;

[0120] Get the real-time frequency PL within the region ID;

[0121] A frequency stabilization function PLF(PL) is constructed to characterize the nonlinear relationship between real-time frequency and frequency stability; the frequency stabilization function satisfies the following formula:

[0122] Where k represents the attenuation coefficient, k>0, and the specific value is set based on experience; in this embodiment, k is set to 10; PL nom Represented as the standard frequency, in this embodiment, PL is... nom Set to 50Hz;

[0123] Substitute the real-time frequency within region ID into the frequency stability function to calculate the frequency stability PWX corresponding to region ID.

[0124] The regional power consumption stability QYWX is obtained by weighted fusion calculation based on voltage stability VX and frequency stability PWX.

[0125] The specific calculation formula is as follows: Wherein, β1 and β2 are weighting coefficients, β1 and β2∈(0,1), and the specific values ​​are set according to experience. In this embodiment, β1 and β2 are set to 0.6 and 0.4 respectively.

[0126] This embodiment introduces a multi-dimensional dynamic evaluation model when assessing regional power stability. The real-time voltage fluctuation amplitude and frequency offset of each monitoring node in the region are the main influencing parameters. This model quantifies the power quality stability of the region, improves the quantitative accuracy of power stability assessment, provides high-confidence data support for the differentiated calculation of regional power priority, and effectively reduces the subjective error rate of traditional assessment methods.

[0127] In this embodiment, generating regional energy priorities based on regional data and energy data includes:

[0128] Energy data includes energy ID and energy type; energy types include grid power, solar power, and energy storage.

[0129] Energy Dimension Scoring (NWP) Based on Energy Data n ;

[0130] Through formula Calculate the regional energy priority QNY; where i represents the region ID, m represents the energy ID, n represents the energy dimension score, and N represents the total number of energy dimension scores; QX m,n This represents the dimension weight coefficient corresponding to the energy dimension score, QP. m Represented as the regional matching coefficient, QP m >0; The specific value is set based on experience. In this embodiment, the region matching coefficient is obtained from the matching coefficient table; NWPn This represents the score for the nth energy dimension; the matching coefficient table is based on pre-set values ​​from experience, as shown in the table below;

[0131]

[0132] In calculating the regional energy priority for each region, this embodiment not only comprehensively considers the compatibility between different energy types and regions, but also conducts multi-dimensional analysis from multiple dimensions of energy data. Based on this, the weight coefficients between each dimension are dynamically adjusted to more accurately reflect the matching effect between various energy types and regions, significantly improving the accuracy and scientific nature of regional energy priority assessment.

[0133] In this embodiment, the generation of energy dimension scores based on energy data includes:

[0134] Energy data includes energy price (NJ), response time (XS), carbon emission intensity (TQ), and energy availability (NYL).

[0135] An economic scoring function JJF(NJ) is constructed to characterize the nonlinear relationship between energy prices and economic scores; the economic scoring function satisfies the following formula:

[0136] Where t represents the price sensitivity coefficient, t>0, and the specific value is set based on experience. In this embodiment, the price sensitivity coefficient is related to the region ID. In the GPU computing region, liquid cooling system region, and network server room region, t is set to 0.8, 1.2, and 1, respectively; NJ base This is represented as the benchmark electricity price; in this embodiment, the benchmark electricity price is set as the average electricity price within the region.

[0137] The economic score corresponding to the energy ID is calculated by substituting the energy price into the economic scoring function;

[0138] A reliability scoring function KKF(XS, NYL) is constructed to characterize the nonlinear relationships between response time, energy availability, and reliability score, respectively; the reliability scoring function satisfies the following formula:

[0139] Among them, XS max This is expressed as the maximum tolerable response time, and the specific value is set based on experience. In this embodiment, XS max The maximum tolerable response time varies depending on the region ID. In this embodiment, the maximum tolerable response time for the GPU computing region, liquid cooling system region, and network equipment room region is set to 0.05 seconds, 0.5 seconds, and 0.1 seconds, respectively.

[0140] The reliability score corresponding to the energy ID is calculated by substituting the response time and energy availability into the reliability scoring function;

[0141] A cleanliness score function QJF(TQ) is constructed to characterize the nonlinear relationship between carbon emission intensity and cleanliness score; the cleanliness score function satisfies the following formula:

[0142] Among them, TQ max This is expressed as the maximum carbon emission standard, and the specific value is set based on experience. In this embodiment, the maximum carbon emission standard is set to 1.0 kgCO2 / kWh.

[0143] The cleanliness score corresponding to the energy ID is calculated by substituting the carbon emission intensity into the cleanliness scoring function;

[0144] The energy dimension score includes an economic score, a reliability score, and a cleanliness score.

[0145] The dimensional weight coefficients in this embodiment are generated in the following ways:

[0146] Obtain data on energy prices, regional electricity stability, carbon allowance balances, and the environment;

[0147] Energy prices, regional electricity stability, carbon allowance margins, and environmental data are integrated into dimensional weight analysis data; dimensional weight analysis data is the data that influences the generation of weight coefficients for several dimensions.

[0148] The dimensional weight analysis data is input into the dimensional weight generation model to obtain several dimensional weight coefficients; the dimensional weight generation model is constructed through a machine learning model to predict the dimensional weight coefficients corresponding to several energy dimension scores.

[0149] The dimension weight generation model in this embodiment is constructed using a machine learning model, including:

[0150] Obtain historical dimensional weight analysis data and their corresponding dimensional weight coefficients;

[0151] The historical dimensional weight analysis data and their corresponding dimensional weight coefficients are divided into training data, validation data, and test data; and the training data, validation data, and test data are preprocessed to obtain training set, validation set, and test set; the ratio between training set, test set, and validation set is 7:2:1.

[0152] Choose a machine learning model as the base model; machine learning models include the BP model, etc.

[0153] The base model is trained on the training set, and the learning rate and other hyperparameters are adjusted on the validation set to obtain the pre-trained model.

[0154] By validating the pre-trained model on the test set, a dimensional weight generation model is finally obtained, which takes dimensional weight analysis data as input and outputs several dimensional weight coefficients.

[0155] This embodiment comprehensively considers multiple dimensions of energy data when calculating regional energy priorities and introduces a pre-trained dimension weight generation model to dynamically score and adjust the weights of each dimension. It can flexibly optimize the weight coefficients of each dimension according to the real-time data changes of the region, thereby more accurately reflecting the demand characteristics of different regions for various types of energy. Through the dynamic calculation and updating of energy priorities, it provides reliable data support and decision-making basis for the efficient matching between subsequent regions and energy supply.

[0156] The power distribution scheme generated in this embodiment based on regional electricity consumption priority and regional energy priority includes:

[0157] The region ID and its corresponding regional electricity priority, as well as several regional energy priorities corresponding to the region ID, are integrated into power distribution analysis data; the power distribution analysis data is the data that affects the generation of power distribution schemes.

[0158] Input the power distribution analysis data into the scheme generation model to obtain the power distribution scheme corresponding to each area ID;

[0159] The solution generation model is constructed using a large model, including:

[0160] Obtain historical power distribution analysis data and their corresponding power distribution schemes;

[0161] Several historical power distribution analysis data and their corresponding power distribution schemes are divided into training data, validation data, and test data; and the training data, validation data, and test data are preprocessed to obtain training sets, validation sets, and test sets; the ratio between the training set, test set, and validation set is 7:2:1.

[0162] Choose a large model as the base model; large models include DeepSeek large models, etc.

[0163] The base model is trained on the training set, and the learning rate and other hyperparameters are adjusted on the validation set to obtain the pre-trained model.

[0164] By validating the pre-trained model on the test set, a scheme generation model is finally obtained, which takes power distribution analysis data as input and outputs power distribution schemes corresponding to region IDs.

[0165] Please see Figure 2Another embodiment of this application provides an integrated energy distribution control system for a smart computing center, including: a data acquisition module, a data analysis module, and a control module; the data acquisition module and the data analysis module are electrically and / or communicatively connected; the data analysis module and the control module are electrically and / or communicatively connected.

[0166] Data acquisition module: Acquires regional data and energy data from the intelligent computing center through data acquisition equipment; the data acquisition equipment includes several sensors, etc.

[0167] Data analysis module: Generates regional electricity consumption priorities based on regional data; generates regional energy priorities based on regional data and energy data; generates power distribution plans based on regional electricity consumption priorities and regional energy priorities;

[0168] Control module: Provides energy to each area based on the power distribution plan.

[0169] Some of the data in the above formula are calculated by removing dimensions and taking their numerical values. The formula is the closest to the real situation obtained by software simulation of a large amount of collected data. The preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.

[0170] The working principle of this application is as follows: It acquires regional data and energy data from a smart computing center; generates regional electricity consumption priorities based on the regional data; generates regional energy priorities based on the regional data and energy data; generates a power distribution plan based on the regional electricity consumption priorities and regional energy priorities; and supplies energy to each region based on the power distribution plan, taking into account both regional priorities and regional energy priorities. This ensures that each region has an optimal power distribution plan, improving the power distribution effect of the distribution area. It avoids the problem that existing technologies often distribute power based on the priority of the distribution area, lacking consideration of the applicable energy type for the region, resulting in poor power distribution performance.

[0171] The above embodiments are only used to illustrate the technical methods of this application and are not intended to limit it. Although this application has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of this application without departing from the spirit and scope of the technical methods of this application.

Claims

1. A method for integrated energy distribution control in an intelligent computing center, characterized in that, include: Acquire regional and energy data from the intelligent computing center; Generate regional electricity consumption priorities based on regional data; The regional electricity priority is a quantification of the priority of each region when distributing electricity. Regional energy priorities are generated based on regional data and energy data; the regional energy priorities refer to the quantification of priority power distribution of several energy sources within a region. Power distribution schemes are generated based on regional electricity consumption priorities and regional energy priorities. Energy supply to various regions is based on the power distribution plan; The generation of regional electricity consumption priorities based on regional electricity consumption forecast load includes: Obtain several historical concurrent electricity loads corresponding to the predicted electricity load of the region with the region ID; The historical load average LFJ corresponding to the region ID is calculated based on several historical electricity loads during the same period. Calculate the regional electricity priority QYY using the formula. i The formula satisfies: Where α1, α2, and α3 represent the load regulation coefficient, level regulation coefficient, and stability regulation coefficient, respectively, and α1, α2, and α3 are all greater than 0; i represents the number corresponding to the region ID, QYYF i Let QZX represent the predicted regional electricity load for the i-th region. i Let QYWX represent the regional importance of the i-th region. i This represents the regional power stability of the i-th region; LFJ i This is represented as the historical average load of the i-th region; The generation of regional energy priorities based on regional data and energy data includes: The energy data includes energy ID and energy type; the energy type includes mains power, photovoltaic power, and energy storage. Energy Dimension Scoring (NWP) Based on Energy Data n ; Through formula Calculate the regional energy priority QNY; where i represents the region ID, m represents the energy ID, n represents the energy dimension score, and N represents the total number of energy dimension scores; QX m,n This represents the dimension weight coefficient corresponding to the energy dimension score, QP. m Represented as the regional matching coefficient, QP m >0; NWP n This is represented as the score for the nth energy dimension; The generation of energy dimension scores based on energy data includes: The energy data includes energy price NJ, response time XS, carbon emission intensity TQ, and energy availability NYL; Construct an economic scoring function JJF(NJ) to characterize the nonlinear relationship between energy prices and economic scores; The economic score corresponding to the energy ID is calculated by substituting the energy price into the economic scoring function; Construct a reliability scoring function KKF(XS, NYL) to characterize the nonlinear relationship between response time, energy availability and reliability score, respectively; The reliability score corresponding to the energy ID is calculated by substituting the response time and energy availability into the reliability scoring function; A cleanliness score function QJF(TQ) is constructed to characterize the nonlinear relationship between carbon emission intensity and cleanliness score; The cleanliness score corresponding to the energy ID is calculated by substituting the carbon emission intensity into the cleanliness scoring function; The energy dimension score includes an economic score, a reliability score, and a cleanliness score.

2. The integrated energy distribution control method for a smart computing center according to claim 1, characterized in that, The generation of regional electricity consumption priorities based on regional data includes: The regional data includes the region ID, electricity consumption impact parameters, and time characteristics; The electricity consumption impact parameters and time characteristics corresponding to the region ID are combined to form electricity consumption analysis data; Extract historical electricity consumption analysis data corresponding to several region IDs; The historical electricity consumption analysis data corresponding to the region ID are integrated into an electricity consumption prediction sequence corresponding to the region ID in chronological order. The region ID and its corresponding electricity consumption forecast sequence are input into the load forecast model to obtain the regional electricity consumption forecast load corresponding to the region ID; the load forecast model is constructed using an LSTM model to predict the electricity consumption of different regions in the future. Regional electricity consumption priorities are generated based on the predicted load of regional electricity consumption.

3. The integrated energy distribution control method for a smart computing center according to claim 2, characterized in that, The load forecasting model is constructed using an LSTM model, including: Obtain several historical electricity consumption forecast sequences and historical electricity loads corresponding to the region ID; Several historical electricity consumption prediction sequences and historical electricity loads corresponding to the region ID are divided into training data, validation data and test data corresponding to the region ID; and the training data, validation data and test data corresponding to the region ID are preprocessed to obtain the training set, validation set and test set corresponding to the region ID. The LSTM model was chosen as the base model for the region ID. The base model corresponding to the region ID is trained on the training set corresponding to the region ID, and the learning rate and other hyperparameters are adjusted on the validation set corresponding to the region ID to obtain the pre-trained model corresponding to the region ID. By validating the pre-trained model corresponding to the region ID on the test set corresponding to the region ID, a load prediction model is finally obtained with the region ID and its corresponding electricity consumption prediction sequence as input and the region electricity consumption prediction load corresponding to the region ID as output.

4. The integrated energy distribution control method for a smart computing center according to claim 1, characterized in that, The regional power stability is obtained through the following methods, including: Obtain the voltage values ​​V of several nodes within region ID; Construct a voltage stability function DYF(V) to characterize the nonlinear relationship between voltage value and voltage stability. j ); The voltage stability corresponding to region ID is calculated by substituting the voltage values ​​of several nodes into the voltage stability function. Get the real-time frequency PL within the region ID; Construct a frequency stability function PLF(PL) to characterize the nonlinear relationship between real-time frequency and frequency stability; The frequency stability corresponding to the region ID is calculated by substituting the real-time frequency within the region ID into the frequency stability function. Regional power supply stability is obtained by weighted fusion calculation based on voltage stability and frequency stability.

5. The integrated energy distribution control method for a smart computing center according to claim 1, characterized in that, The dimension weight coefficients are generated in the following ways: Obtain data on energy prices, regional electricity stability, carbon allowance balances, and the environment; Integrate energy prices, regional electricity stability, carbon allowance margins, and environmental data into dimensional weighted analysis data; The dimensional weight analysis data is input into the dimensional weight generation model to obtain several dimensional weight coefficients; the dimensional weight generation model is constructed through a machine learning model and is used to predict the dimensional weight coefficients corresponding to several energy dimensional scores.

6. The integrated energy distribution control method for a smart computing center according to claim 1, characterized in that, The generation of a power distribution scheme based on regional electricity consumption priority and regional energy priority includes: The region ID and its corresponding regional electricity priority, as well as the energy priorities of several regions corresponding to the region ID, are integrated into power distribution analysis data. Input the power distribution analysis data into the scheme generation model to obtain the power distribution scheme corresponding to each area ID; The proposed solution generation model is constructed using a large model, including: Obtain historical power distribution analysis data and their corresponding power distribution schemes; Several historical power distribution analysis data and their corresponding power distribution schemes are divided into training data, validation data, and test data; and the training data, validation data, and test data are preprocessed to obtain training sets, validation sets, and test sets. Choose a large model as the base model; The base model is trained on the training set, and the learning rate and other hyperparameters are adjusted on the validation set to obtain the pre-trained model. By validating the pre-trained model on the test set, a scheme generation model is finally obtained, which takes power distribution analysis data as input and outputs power distribution schemes corresponding to region IDs.

7. A smart computing center integrated energy distribution control system, applied to the smart computing center integrated energy distribution control method according to any one of claims 1-6, characterized in that, include: Data acquisition module, data analysis module, and control module; The data acquisition module is connected to the data analysis module; the data analysis module is connected to the control module. The data acquisition module acquires data from several areas of the intelligent computing center and energy data through data acquisition equipment. The data analysis module generates regional electricity consumption priorities based on regional data. Generate regional energy priorities based on regional and energy data; Power distribution schemes are generated based on regional electricity consumption priorities and regional energy priorities. The control module supplies energy to each area based on the power distribution scheme.

Images