Word folding method and system under direct current interconnection reliable power supply facing computing power load

Abstract

This invention discloses a token conversion method and system for reliable power supply under DC interconnection of computing loads, relating to the technical field of collaborative planning of new energy power supply and computing infrastructure. The method includes: mapping and classifying power consumption relationships; establishing an equivalent configuration model for the DC interconnection system of photovoltaic-storage computing centers; determining the minimum capacity configuration of photovoltaic and energy storage to meet the reliable power supply requirements of computing centers based on the power consumption mode and reliable power supply requirements of computing centers; establishing an integrated operating cost model; and, in conjunction with the periodic token output of computing centers, calculating unit token cost, break-even token price, and token suggested price decision. This invention can solve the problems of difficulty in accurately characterizing the dynamic load characteristics of computing centers, unclear relationship between photovoltaic and energy storage capacity configuration under random photovoltaic power supply conditions, and difficulty in corresponding the energy supply cost with the token output.

Description

Technical Field

[0001] This invention relates to the field of collaborative planning technology for new energy power supply and computing infrastructure, and in particular to a word conversion method and system for reliable power supply under DC interconnection for computing loads. Background Technology

[0002] Currently, research on the coordinated operation of computing centers and new energy sources mainly includes microgrid energy management, photovoltaic energy storage capacity planning, workload scheduling, operating cost optimization, and carbon emission assessment. However, problems still exist, such as a disconnect between configuration models, a separation between power consumption and output, and a separation between cost accounting. Most studies place computing centers within a grid-connected framework, focusing on power purchase and sale strategies or external electricity price responses, failing to directly provide the minimum configuration boundary for photovoltaic and energy storage capacities to meet continuous power supply requirements, especially neglecting power balance and random fluctuation constraints under direct DC bus coupling. Furthermore, existing power consumption models only describe power changes and do not establish a correlation with token output capabilities. In addition, energy investment and computing service pricing belong to different research fields, lacking a unified framework from reliable power supply configuration to token cost accounting, making it difficult to support capacity planning, reliability verification, and cost calculation in integrated operations. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the purpose of this invention is to provide a token conversion method and system for reliable power supply under DC interconnection for computing loads, solving the problems in existing technologies such as the difficulty in accurately characterizing the dynamic load characteristics of computing centers, the unclear relationship between photovoltaic and energy storage capacity configuration under random photovoltaic power supply conditions, and the difficulty in corresponding and calculating the energy supply side cost with the token output.

[0004] To achieve the above objectives, the present invention is implemented through the following technical solution: The first aspect of this invention provides a word conversion method for reliable power supply under DC interconnection for computing loads, comprising the following steps: Obtain operational data and token measurement data of the target computing center, and perform relationship mapping and classification of power consumption based on the aggregated business behavior of the computing center to obtain the power consumption mode of the computing center; An equivalent configuration model of the DC interconnection system of the photovoltaic and energy storage computing center is established. Based on the power consumption mode and reliable power supply requirements of the computing center, the minimum capacity configuration of photovoltaic and energy storage that meets the reliable power supply requirements of the computing center is determined. Based on the minimum optical storage configuration, an integrated operating cost model is established to calculate the comprehensive cost within the billing cycle; and combined with the periodic token output of the computing center, unit token cost accounting, break-even token price calculation, and token suggested price decision are performed.

[0005] Furthermore, the operational data is collected by the target computing center's operation monitoring system according to a preset time granularity, including the computing node operating status, task load, IT equipment power consumption, cooling equipment power consumption, total power consumption of the computing center, and power usage efficiency; the token measurement data is obtained from the call logs or metering module of the large model service platform, including the number of input tokens, the number of output tokens, and the total number of tokens corresponding to each time period.

[0006] Furthermore, based on operational data and token measurement data, a mapping relationship is established between token measurement, typical power consumption mode, and total power consumption of the computing center, resulting in the dynamic load curve of the computing center, token output under different power consumption modes, and basic operating power consumption.

[0007] Furthermore, based on the power consumption pattern and reliable power supply requirements of the computing center, the specific steps for determining the minimum capacity configuration of photovoltaic and energy storage to meet the reliable power supply requirements of the computing center are as follows: The equivalent configuration model of the DC interconnection system of the photovoltaic-storage computing center takes the maximum load power determined by the dynamic load curve of the computing center as the reliable power supply requirement, and takes the photovoltaic installed capacity, energy storage power capacity and energy storage capacity as the parameters to be configured. Based on the balance relationship between random photovoltaic output, dynamic load of computing center, energy storage state of charge, energy storage charging and discharging power and DC bus power, and in accordance with the reliable power supply requirements of computing center, the constraints of photovoltaic and energy storage that meet the reliable power supply requirements of computing center are determined. Establish a minimum capacity configuration model for photovoltaic and energy storage to determine the minimum photovoltaic and energy storage configuration.

[0008] Furthermore, the minimum capacity configuration model for photovoltaic and energy storage takes the minimum reliable power supply standard of the computing center as a constraint and aims to minimize the weighted sum of photovoltaic configuration capacity, energy storage capacity, and energy storage power capacity to determine the minimum photovoltaic and energy storage configuration.

[0009] Furthermore, the integrated operation cost model includes a fixed cost model and an operation variable cost model. The fixed cost model includes photovoltaic construction costs, energy storage construction costs, DC / DC converter construction costs, DC line construction costs, and fixed operation and maintenance costs. The operation variable cost model includes energy storage cycle degradation costs, DC interconnection system loss costs, and computing center operation costs.

[0010] A second aspect of the present invention provides a word conversion system for reliable power supply under DC interconnection for computing loads, comprising: The data acquisition and power consumption decomposition module is used to acquire the operating data and token measurement data of the target computing center, and to perform relational mapping and division of power consumption based on the aggregated business behavior of the computing center to obtain the power consumption mode of the computing center. The minimum configuration determination module is used to establish an equivalent configuration model of the DC interconnection system of the photovoltaic and energy storage computing center. Based on the power consumption mode and reliable power supply requirements of the computing center, it determines the minimum capacity configuration of photovoltaic and energy storage that meets the reliable power supply requirements of the computing center. The cost accounting and decision generation module establishes an integrated operating cost model based on the minimum optical storage configuration to calculate the comprehensive cost within the billing cycle; and combines the token output of the computing center in a cycle to perform unit token cost accounting, break-even token price calculation, and token suggested price decision.

[0011] A third aspect of the present invention provides a computer-readable storage medium storing a computer program adapted to be loaded by a processor and to execute steps in the word conversion method for reliable power supply under DC interconnection for computing loads as described in the first aspect of the present invention.

[0012] A fourth aspect of the present invention provides a computer device comprising: A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the word conversion method for reliable power supply under DC interconnection for computing loads as described in the first aspect of the present invention.

[0013] A fifth aspect of the present invention provides a computer program product or computer program comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the steps in the word conversion method for reliable power supply under DC interconnection for computing loads as described in the first aspect of the present invention.

[0014] The above one or more technical solutions have the following beneficial effects: This invention discloses a token conversion method and system for reliable power supply under DC interconnection for computing load. Starting from the integrated construction and operation of photovoltaic-storage-computing power centers, it establishes a dynamic load model for computing power centers based on the aggregated load behavior of computing power centers. Under the constraints of stochastic photovoltaic output and energy storage operation, it determines the photovoltaic-storage constraints that meet the continuous power supply requirements of computing power centers. Furthermore, it performs capacity configuration based on reliable power supply constraints, and calculates the unit token cost, break-even token parameters, and token recommendation decisions under a given profit rate in combination with the periodic token output of computing power centers.

[0015] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0016] 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.

[0017] Figure 1 This is a flowchart of the word conversion method under reliable power supply for DC interconnection of computing load in Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the physical structure of the DC interconnection system for photovoltaic, energy storage, and computing centers in Embodiment 1 of the present invention. Detailed Implementation

[0018] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0019] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof. The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. 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.

[0020] As artificial intelligence (AI) technology accelerates from the model development stage to large-scale application, the demand for computing infrastructure is continuously growing due to the need for large-scale model training, inference services, and intelligent industry applications. As a critical infrastructure for AI computing tasks, the energy consumption and power supply reliability of computing centers have become significant factors affecting the development and stable operation of the AI ​​industry. On the one hand, the load on computing centers continues to expand, placing high demands on power supply continuity and quality during operation. On the other hand, the supply of new energy sources is volatile and uncertain, and traditional construction methods relying on external power supply expansion cannot fully reflect the needs of local clean energy consumption and low-carbon operation of computing facilities. Therefore, developing a local energy supply method that coordinates photovoltaic, energy storage, and computing load configuration for computing centers is of great significance for improving the energy supply level of computing infrastructure, enhancing the support capability of high-reliability computing services, and establishing a quantifiable cost accounting mechanism for computing energy consumption.

[0021] Photovoltaic power generation is characterized by its clean resources, flexible deployment, and suitability for grid connection around industrial parks and buildings, providing a low-carbon power source for computing centers. Energy storage systems can release energy when photovoltaic output is insufficient and absorb electricity when photovoltaic output is abundant, thus alleviating the contradiction between the randomness of photovoltaic power generation and the continuity of computing load. For application scenarios such as local consumption of new energy, independent power supply at the park level, edge computing nodes, and limited AC access capabilities, simply relying on external power compensation cannot reflect the true support capacity of local photovoltaic and energy storage resources for the continuous operation of computing centers. It is necessary to evaluate the capacity configuration relationship between photovoltaic and energy storage based on the local energy supply structure.

[0022] Photovoltaic modules, energy storage batteries, and a large number of IT devices within computing centers all exhibit DC-side access or DC power consumption characteristics. Adopting a DC interconnection method for photovoltaic-storage-computing centers can reduce the multi-stage AC-DC conversion links in traditional AC power supply paths, lower energy transmission losses, and establish a direct power coupling relationship between the photovoltaic side, energy storage side, and computing load side via a DC bus. In this structure, the random fluctuations in photovoltaic output, changes in the state of charge of energy storage, and the dynamic load of the computing center interact with each other, requiring the energy storage system to play a power balancing role between photovoltaic supply and computing load demand. Therefore, capacity configuration is no longer simply a matter of selecting a single device, but a comprehensive configuration problem that simultaneously considers random photovoltaic output, energy storage power and energy constraints, DC bus power balance, and the continuous power supply requirements of the computing center.

[0023] Existing research has focused on the coordinated operation of computing centers and new energy sources, primarily encompassing microgrid energy management, photovoltaic energy storage capacity planning, workload scheduling, operating cost optimization, and carbon emission assessment. This research provides a crucial foundation for characterizing the energy consumption features of computing centers, describing the charging and discharging process of energy storage, and constructing photovoltaic-energy storage capacity optimization models. However, existing methods often place computing centers under the constraints of grid-connected microgrids, active distribution networks, or points of common coupling, focusing on power purchase and sale strategies, day-ahead power planning, external electricity price response, carbon emission intensity tracking, or workload migration. While these methods are suitable for analyzing the participation of computing centers in grid dispatch or reducing grid-connected operating costs, they remain insufficient for providing a direct minimum configuration boundary for reliable power supply between photovoltaic and energy storage capacities in local photovoltaic-energy storage DC power supply scenarios.

[0024] On the other hand, existing computing center power consumption models are typically built around factors such as server utilization, workload queues, IT equipment power consumption, and cooling power consumption, reflecting the basic principle that the total power consumption of a computing center is composed of both computing and auxiliary equipment. However, when serving large-scale models, business output is closely related to token metering. The token production capacity, energy consumption per token, and runtime of the computing center under different power consumption modes directly affect the energy cost allocation results. If only power changes are described without establishing the relationship between power consumption modes and token production, it is difficult to further answer the question of how much energy cost each unit of token should bear after the local optical energy storage system is built.

[0025] Existing token pricing research focuses more on platform billing rules, package design, user needs, and model version differences, while research on photovoltaic (PV) and energy storage capacity planning focuses more on energy equipment investment, operation and maintenance, energy storage degradation, and supply-demand balance. For integrated PV-energy storage computing center operators, both need to be considered within the same technical framework: PV and energy storage configurations determine the energy costs required for reliable operation of the computing center, and the token output under different power consumption modes of the computing center determines the cost-sharing basis. Without a unified method from reliable power supply configuration to token cost accounting, it is difficult to simultaneously support capacity planning, reliability verification, and operational cost assessment.

[0026] To address the aforementioned problems, this invention discloses a word conversion method and system for reliable power supply under DC interconnection for computing loads. Details are provided in the following embodiments.

[0027] Example 1: Embodiment 1 of the present invention provides a word conversion method for reliable power supply under DC interconnection for computing loads, such as Figure 1As shown, firstly, a dynamic load model is established based on the aggregated business behavior of the computing center, and then the minimum configuration for reliable power supply is determined under the constraints of random photovoltaic output and energy storage status. On the basis of meeting the requirements for reliable power supply, the economic configuration is determined in combination with the investment and operating costs of photovoltaic, energy storage and DC conversion equipment. Finally, based on the token output under the typical power consumption mode of the computing center, the unit token cost, the break-even price and the suggested price under a given profit rate are calculated, thereby providing methodological support for the planning, construction and operation evaluation of local photovoltaic-storage DC power supply computing centers.

[0028] In this embodiment, a token refers to a basic processing and measurement unit formed by a large language model according to preset text segmentation and encoding rules. Its granularity is determined by the encoder of the model used and can correspond to character fragments, sub-words, word fragments, punctuation marks, or combinations thereof. The number of tokens is used to characterize the scale of text processed or generated by the model's inference service. Without establishing a correlation between power consumption and token output capability, it is impossible to support a reasonable evaluation of the energy consumption per unit token.

[0029] Specifically, the following steps are included: S1: Obtain the operating data and token measurement data of the target computing center, and perform relationship mapping and division of power consumption based on the aggregated business behavior of the computing center to obtain the power consumption mode of the computing center.

[0030] In one specific implementation, operational data is collected by the target computing center's operational monitoring system at a preset time granularity, including the computing node operating status, task load, IT equipment power consumption, cooling equipment power consumption, total computing center power consumption, and energy efficiency. Token metering data is obtained from the call logs or metering module of the large model service platform, including the number of input tokens, output tokens, and total tokens corresponding to each time period. Based on the operational data and token metering data, a mapping relationship is established between token metering, typical power consumption patterns, and the total power consumption of the computing center, resulting in the computing center's dynamic load curve, token output under different power consumption patterns, and basic operating power consumption. The computing center's dynamic load curve is used for subsequent DC bus power balance calculations and optical storage capacity configuration calculations based on the computing center's stable power supply. The token output is used to make token decisions based on the comprehensive cost within the integrated operation cycle and the computing center's periodic token output.

[0031] Specifically: S1.1: Establish a power consumption decomposition model for the computing center, dividing the power consumption of the computing center into two parts: AI running token power consumption and basic running power consumption.

[0032] (1).

[0033] Wherein, PUE(t) represents the power utilization efficiency of the computing center in the t-th time period. This represents the total power consumption of the computing center during the t-th time period. This represents the basic power consumption of the computing center, which includes server standby power consumption, network communication equipment power consumption, and non-token power consumption of the computing center's electricity consumption. This indicates the power consumption of the token used to run the AI.

[0034] S1.2: The power consumption of AI running tokens is determined based on the number of AI inference requests in the t-th time period, the average input token length of a single AI inference request, the average output token length, the energy consumption coefficient per unit input token, and the energy consumption coefficient per unit output token, so as to map the token processing volume in the AI ​​inference business to the average power consumption on the power side.

[0035] In one specific implementation, an AI runtime token power consumption model is established. During the t-th time period, let R(t) be the number of AI inference requests handled by the computing center, and let the average input token length and average output token length for a single AI inference request be respectively... and The corresponding number of input tokens and the number of output tokens are as follows: (2).

[0036] in, This represents the number of tokens input within the t-th time period. This represents the number of output tokens within the t-th time period. The input tokens correspond to the prompt word reading, context understanding, and pre-filling calculation stages in the model's inference process, while the output tokens correspond to the progressive generation stage of the model's response content.

[0037] Considering the differences between input and output tokens in terms of computation phase, memory access, and unit power consumption, energy consumption coefficients are set for unit input tokens and unit output tokens respectively, resulting in the power consumption of the AI ​​running tokens: (3).

[0038] in, This indicates the energy consumption per unit input token. This represents the energy consumption per unit output token, where Δt is the time interval between two adjacent discrete time periods. This formula is used to map the token processing volume in AI inference operations to the average power consumption on the electricity side.

[0039] S1.3: Establish a computing center aggregated load model.

[0040] In one specific implementation, substituting equations (2) and (3) into equation (1) yields the computing center aggregated load model: (4).

[0041] S1.4: Perform power consumption mode division for computing centers.

[0042] In one specific implementation, the typical power consumption mode of this embodiment is divided according to the power consumption contribution of AI running token. The power consumption contribution of AI running token is used to characterize the proportion of AI running token power consumption in the power consumption of the IT side of the computing center. According to the preset threshold, the power consumption mode of the computing center is divided into basic maintenance power consumption mode, collaborative operation power consumption mode and intelligent inference-dominated power consumption mode.

[0043] To characterize the differences in load behavior of the computing center at different times, the power consumption contribution of the AI ​​running token is defined as follows: (5).

[0044] in, This represents the contribution of AI token power consumption to the total power consumption on the IT side of the computing center during the t-th time period. When the load is relatively low, it indicates that the load on the computing center is mainly determined by basic operating power consumption and non-AI business load; when When the value is large, it indicates that the computing center load is mainly driven by AI inference requests and the token generation process.

[0045] according to The value of is used to classify the power consumption mode of the computing center into three types: basic maintenance power consumption mode, collaborative operation power consumption mode, and intelligent inference-driven power consumption mode. (6).

[0046] Where M(t) represents the power consumption pattern of the computing center in the t-th time period. Indicates the basic sustain power mode. Indicates a cooperative operation power consumption mode. This indicates a power consumption mode dominated by intelligent inference. and Define a threshold for the preset power consumption mode, and satisfy (0 < ... < <1).

[0047] S2: Establish an equivalent configuration model for the DC interconnection system of the photovoltaic and energy storage computing center. Based on the power consumption mode and reliable power supply requirements of the computing center, determine the minimum capacity configuration of photovoltaic and energy storage that meets the reliable power supply requirements of the computing center.

[0048] S2.1: Establish an equivalent configuration model of the DC interconnection system of the optical storage computing center based on the DC interconnection system of the optical storage computing center.

[0049] In one specific implementation, such as Figure 2 As shown, the DC interconnection system of the photovoltaic-storage computing center includes photovoltaic power generation units, energy storage units, a DC bus, DC / DC converters, and computing center load units. The photovoltaic power generation units, energy storage units, and computing center load units are connected to the DC bus through their respective DC / DC converters. The three converters are controlled using MPPT control, droop control, and constant voltage control, respectively. During system operation, the photovoltaic power generation units prioritize supplying power to the computing center. When the photovoltaic output exceeds the computing center's load demand, the excess energy is used to charge the energy storage units. When the photovoltaic output is insufficient, the energy storage units discharge to compensate for the power deficit, thereby maintaining the power balance of the DC bus and ensuring a continuous and stable power supply to the computing center. This structure demonstrates the DC coupling relationship between photovoltaics, energy storage, and the computing center loads, providing a physical basis for subsequent reliable power supply configuration, minimum capacity boundary determination, and token cost calculation.

[0050] An equivalent configuration model for DC interconnection between photovoltaic and energy storage computing centers is established. To ensure that the computing centers can obtain stable power supply under different power consumption modes, this embodiment treats the computing centers as time-varying loads connected to the DC bus and constructs a DC interconnection model between photovoltaic, energy storage and computing centers.

[0051] Specifically, the photovoltaic power generation unit, energy storage unit, and computing center load unit are all interconnected via DC / DC converters. This DC interconnection system does not have an external AC grid interface, and the computing center load is entirely shared by the photovoltaic power generation unit and the energy storage unit.

[0052] The DC bus power balance relationship includes the power injected into the DC bus by the photovoltaic unit, the power injected into the DC bus by the energy storage unit, the total power consumption of the computing center, and the power loss of the DC line and each converter. The injected power is positive when the energy storage unit is discharging and negative when the energy storage unit is charging.

[0053] Under steady-state operating conditions, the DC bus power balance relationship can be expressed as follows: (7).

[0054] in, This indicates the power injected into the DC bus by the photovoltaic unit. This represents the power injected into the DC bus by the energy storage unit; it is positive during energy discharge and negative during charging. This indicates the power loss of the DC line and each converter.

[0055] S2.2: Based on the power consumption mode and reliable power supply requirements of the computing center, determine the minimum capacity configuration of photovoltaic and energy storage that meets the reliable power supply requirements of the computing center.

[0056] S2.2.1: The equivalent configuration model of the DC interconnection system of the photovoltaic-storage computing center uses the maximum load power determined by the dynamic load curve of the computing center as the reliable power supply requirement, and uses the photovoltaic installed capacity, energy storage power capacity, and energy storage capacity as parameters to be configured. Among them, the maximum load power of the computing center is the computing center capacity.

[0057] In one specific implementation, an equivalent operating model for the photovoltaic unit and the energy storage unit is established, where the available output power of the photovoltaic unit is determined jointly by the photovoltaic configuration capacity and the photovoltaic output per unit capacity. (8).

[0058] in, Indicates the photovoltaic configuration capacity. This represents the available photovoltaic power output per unit capacity during the t-th time period. This indicates the photovoltaic conversion efficiency.

[0059] The energy storage state of charge update relationship is expressed as follows: (9).

[0060] Where SOC(t) represents the state of charge of the energy storage unit. Indicates the energy storage configuration capacity. Indicates the energy storage discharge power. , indicates the energy storage charging power. and These represent the energy storage charging efficiency and the discharging efficiency, respectively.

[0061] S2.2.2: Based on the balance relationship between random photovoltaic output, dynamic load of computing center, energy storage state of charge, energy storage charging and discharging power and DC bus power, and in accordance with the reliable power supply requirements of computing center, determine the constraints of photovoltaic and energy storage that meet the reliable power supply requirements of computing center.

[0062] In one specific implementation, when determining whether the combination of photovoltaic and energy storage capacity meets the requirements for reliable power supply, it is necessary to ensure that the photovoltaic and energy storage can meet the reliable power supply requirements of the computing center under different power operation conditions. To this end, corresponding control strategies need to be configured for the photovoltaic, energy storage and computing center sides respectively.

[0063] Specifically, the control strategy for the DC-DC converters in the optical storage computing center is set according to the functions of each DC / DC converter. To maintain stable power supply under different power consumption modes in different computing centers, each DC / DC converter undertakes different control responsibilities.

[0064] The photovoltaic-side DC / DC converter adopts maximum power point tracking control, which enables the photovoltaic unit to output the maximum available power under the premise of meeting the DC bus voltage constraint and energy storage charging capacity constraint. When the photovoltaic output exceeds the load demand of the computing center and the energy storage reaches the charging limit, the photovoltaic-side DC / DC converter switches to power limiting control to avoid DC bus overvoltage.

[0065] The DC / DC converter on the energy storage side adopts droop control. When the photovoltaic output is insufficient to cover the power consumption of the computing center and system losses, the energy storage unit discharges to compensate for the power deficit; when there is a surplus in photovoltaic output, the energy storage unit absorbs the excess power for charging.

[0066] The DC / DC converter on the computing center side adopts a constant voltage control method to ensure a stable power supply to the computing center when the power consumption of the computing center changes.

[0067] S2.2.3: Establish a minimum capacity configuration model for photovoltaic and energy storage, taking the minimum reliable power supply standard of the computing center as a constraint, and taking the minimum weighted sum of photovoltaic configuration capacity, energy storage capacity and energy storage power capacity as the objective to determine the minimum photovoltaic and energy storage configuration.

[0068] In one specific implementation, a minimum photovoltaic-storage capacity configuration model is established. To obtain the minimum photovoltaic-storage configuration, with the goal of minimizing the weighted sum of photovoltaic configuration capacity, energy storage capacity, and energy storage power capacity, the following configuration model is established: (10).

[0069] Where J represents the evaluation index for the scale of photovoltaic and energy storage configuration. , and These represent the weighting coefficients for photovoltaic capacity, energy storage capacity, and energy storage power capacity, respectively. P ES max This indicates the energy storage capacity.

[0070] This configuration model must meet the following stable power supply constraints: (11).

[0071] in, and These represent the lower and upper limits of the energy storage state of charge, respectively. and These represent the lower and upper limits of the allowable voltage for the DC bus, respectively.

[0072] To ensure that the energy storage capacity can cover the energy deficit under conditions of continuous low photovoltaic output or high computing center power consumption, the energy storage capacity must meet the following requirements: (13).

[0073] in, , where x represents the input quantity.

[0074] Energy storage capacity must meet the requirements for supporting the maximum instantaneous power deficit: (14).

[0075] Photovoltaic capacity must meet long-term energy supply requirements: (15).

[0076] In summary, equations (7) to (15) together constitute the minimum capacity configuration model for photovoltaic and energy storage. Here, photovoltaic installed capacity, energy storage power capacity, and energy storage energy capacity are the configuration variables to be determined. Under given photovoltaic output and dynamic load conditions of the computing center, the photovoltaic output, energy storage charging and discharging power, and energy storage state of charge for each time period are determined according to the aforementioned DC / DC converter control strategy. The constraints are DC bus power balance, energy storage state of charge boundary, energy storage charging and discharging power constraint, energy storage energy support constraint, and long-term energy supply requirements. By solving an optimization problem with the objective of minimizing the weighted sum of photovoltaic configuration capacity, energy storage energy capacity, and energy storage power capacity, the minimum capacity configuration of photovoltaic and energy storage that meets the reliable power supply requirements of the computing center is obtained. This minimum capacity configuration serves as the standard for subsequent integrated operation cost accounting and token decision-making.

[0077] S3: Based on the minimum optical storage configuration, establish an integrated operating cost model to calculate the comprehensive cost within the billing cycle; and combine the token output of the computing center in a cycle to perform unit token cost accounting, break-even token price calculation, and token suggested price decision.

[0078] S3.1: Establish an integrated operating cost model based on the minimum optical storage configuration to calculate the comprehensive cost within the billing cycle.

[0079] In one specific implementation, an integrated operating cost model is established based on the minimum photovoltaic-storage configuration. This model comprehensively considers photovoltaics, energy storage, DC-DC converter equipment, operation and maintenance, energy storage degradation, conversion losses, and the opportunity cost of curtailment to determine the operating cost. The integrated operating cost model includes a fixed cost model and a variable operating cost model. The fixed cost model includes photovoltaic construction costs, energy storage construction costs, DC / DC converter construction costs, DC line construction costs, and fixed operation and maintenance costs. The variable operating cost model includes energy storage cyclic degradation costs, DC interconnection system loss costs, and computing center operating costs.

[0080] This embodiment constructs an integrated operation cost model for a photovoltaic-storage-computing center from the perspective of an integrated operator of such centers.

[0081] Specifically: S3.1.1: Construct a fixed cost model.

[0082] In one specific implementation, the fixed cost model includes photovoltaic construction costs, energy storage construction costs, DC / DC converter construction costs, DC line construction costs, and fixed operation and maintenance costs, expressed by the following formula: (16).

[0083] in, This represents the equivalent fixed cost of a photovoltaic-storage computing center, where k represents the equivalent conversion factor. , , and These represent the construction costs per unit of photovoltaic capacity, per unit of energy storage capacity, per unit of converter power, and per unit of DC line size, respectively. PPCS represents the total configured power of the DC / DC converter. Indicates the length of a DC line or its equivalent line size. This represents fixed maintenance costs.

[0084] S3.1.2: Construct a variable cost model for the operation of the optical storage computing center.

[0085] In one specific implementation, during the t-th time period, the variable operating cost of the optical-storage computing center mainly consists of the energy storage cycle degradation cost, the DC interconnect system loss cost, and the computing center operating cost: (17).

[0086] in, This represents the variable cost of operation during the t-th time period. This represents the cost per unit of energy storage that degrades over time. This represents the cost of electricity loss in a DC interconnect system. Indicates the system power loss. This represents the unit energy consumption operating cost of a computing center.

[0087] S3.1.3: Calculate the total cost within the billing cycle.

[0088] In one specific implementation, the total cost of the optical storage computing center within the billing period T is the sum of fixed costs and variable costs for each time period: (1) in, This represents the total cost of the optical storage computing center within the billing period.

[0089] S3.2: Based on the comprehensive cost within the integrated operation cycle and the token output of the computing center during the cycle, calculate the unit token cost, calculate the break-even token price, and make a token suggested price decision.

[0090] In one specific implementation, based on the integrated operation cost model of the optical storage computing center and the total amount of tokens, the unit token cost, the break-even token parameter, and token decision suggestions under a given profit rate are calculated.

[0091] Specifically: S3.1.4: Calculate the cost per unit token.

[0092] In one specific implementation, the unit token cost is calculated based on the total cost of the optical storage computing center within the billing period and the total number of tokens within the billing period.

[0093] Fixed costs are allocated based on the amount of billing tokens, resulting in the fixed cost allocation item for the t-th time period: (18).

[0094] in, This represents the unit token cost within the t-th time period.

[0095] S3.1.5: Determine the token recommendation decision.

[0096] In one specific implementation, it is recommended that token decisions be determined based on the unit token cost, combined with an operational revenue coefficient and a service value correction coefficient related to the power consumption mode of the computing center. The decision can be a suggested price for the token.

[0097] Based on the unit token cost, the suggested price for time-sharing transactions is expressed as follows: (19).

[0098] in, This represents the suggested price of the token in the t-th time period, where μ represents the operating profit coefficient. This represents the service value correction coefficient related to the power consumption mode of the computing center.

[0099] This invention combines aggregated business behavior, token measurement, and total power consumption modeling in the computing center, enabling it to characterize the service load changes of the large model from the overall level of the computing center, avoiding the estimation of the computing center load based solely on device power consumption or service request power consumption.

[0100] This invention addresses the DC interconnection structure of photovoltaic-storage computing centers, establishing a reliable power supply constraint relationship between random photovoltaic output, energy storage status, and dynamic load of the computing center, thereby determining the minimum capacity configuration that meets the continuous power supply requirements of the computing center.

[0101] This invention combines the energy supply costs of photovoltaic, energy storage, and DC conversion equipment with the periodic token output of the computing center to obtain the unit token cost, providing a cost accounting basis for the integrated operation of photovoltaic-storage computing centers.

[0102] Example 2: Embodiment 2 of the present invention provides a word conversion system for reliable power supply under DC interconnection for computing loads, comprising: The data acquisition and power consumption decomposition module is used to acquire the operating data and token measurement data of the target computing center, and to perform relational mapping and division of power consumption based on the aggregated business behavior of the computing center to obtain the power consumption mode of the computing center. The minimum configuration determination module is used to establish an equivalent configuration model of the DC interconnection system of the photovoltaic and energy storage computing center. Based on the power consumption mode and reliable power supply requirements of the computing center, it determines the minimum capacity configuration of photovoltaic and energy storage that meets the reliable power supply requirements of the computing center. The cost accounting and decision generation module establishes an integrated operating cost model based on the minimum optical storage configuration to calculate the comprehensive cost within the billing cycle; and combines the token output of the computing center in a cycle to perform unit token cost accounting, break-even token price calculation, and token suggested price decision.

[0103] Example 3: Embodiment 3 of the present invention provides a computer-readable storage medium storing a computer program adapted for loading by a processor and executing the steps in the word conversion method for reliable power supply under DC interconnection for computing loads as described in Embodiment 1 of the present invention.

[0104] Example 4: Embodiment 4 of the present invention provides a computer device, the device comprising: A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the steps in the word conversion method for reliable power supply under DC interconnection for computing loads as described in Embodiment 1 of the present invention.

[0105] Example 5: Embodiment 5 of the present invention provides a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the steps in the word conversion method for reliable power supply under DC interconnection for computing loads as described in Embodiment 1 of the present invention.

[0106] The steps and methods involved in Examples 2, 3, 4 and 5 above correspond to those in Example 1. For specific implementation methods, please refer to the relevant description section of Example 1.

[0107] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0108] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in or transmitted through a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The computer-readable storage medium can be any available medium that a computer can access or a data processing device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium, an optical medium, or a semiconductor medium, etc.

[0109] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A word conversion method for reliable power supply under DC interconnection for computing loads, characterized in that, Includes the following steps: Obtain operational data and token measurement data of the target computing center, and perform relationship mapping and classification of power consumption based on the aggregated business behavior of the computing center to obtain the power consumption mode of the computing center; An equivalent configuration model of the DC interconnection system of the photovoltaic and energy storage computing center is established. Based on the power consumption mode and reliable power supply requirements of the computing center, the minimum capacity configuration of photovoltaic and energy storage that meets the reliable power supply requirements of the computing center is determined. Based on the minimum optical storage configuration, an integrated operating cost model is established to calculate the comprehensive cost within the billing cycle; and combined with the periodic token output of the computing center, unit token cost accounting, break-even token price calculation, and token suggested price decision are performed.

2. The word conversion method for reliable power supply under DC interconnection for computing load as described in claim 1, characterized in that, Operational data is collected by the target computing center's operation monitoring system at preset time granularity, including computing node operating status, task load, IT equipment power consumption, cooling equipment power consumption, total computing center power consumption, and power usage efficiency; token measurement data is obtained from the call logs or metering module of the large model service platform, including the number of input tokens, output tokens, and total tokens corresponding to each time period.

3. The word conversion method for reliable power supply under DC interconnection for computing load as described in claim 1, characterized in that, Based on operational data and token measurement data, a mapping relationship is established between token measurement, typical power consumption mode and total power consumption of the computing center, resulting in the dynamic load curve of the computing center, token output under different power consumption modes and basic operating power consumption.

4. The word conversion method for reliable power supply under DC interconnection for computing load as described in claim 1, characterized in that, Based on the power consumption pattern and reliable power supply requirements of computing centers, the specific steps to determine the minimum capacity configuration of photovoltaic and energy storage that meets the reliable power supply requirements of computing centers are as follows: The equivalent configuration model of the DC interconnection system of the photovoltaic-storage computing center takes the maximum load power determined by the dynamic load curve of the computing center as the reliable power supply requirement, and takes the photovoltaic installed capacity, energy storage power capacity and energy storage capacity as the parameters to be configured. Based on the balance relationship between random photovoltaic output, dynamic load of computing center, energy storage state of charge, energy storage charging and discharging power and DC bus power, and in accordance with the reliable power supply requirements of computing center, the constraints of photovoltaic and energy storage that meet the reliable power supply requirements of computing center are determined. Establish a minimum capacity configuration model for photovoltaic and energy storage to determine the minimum photovoltaic and energy storage configuration.

5. The word conversion method for reliable power supply under DC interconnection for computing load as described in claim 4, characterized in that, The minimum capacity configuration model for photovoltaic and energy storage is constrained by the minimum reliable power supply standard for computing centers, and aims to determine the minimum photovoltaic and energy storage configuration by minimizing the weighted sum of photovoltaic configuration capacity, energy storage capacity, and energy storage power capacity.

6. The word conversion method for reliable power supply under DC interconnection for computing load as described in claim 1, characterized in that, The integrated operation cost model includes a fixed cost model and a variable operation cost model. The fixed cost model includes photovoltaic construction costs, energy storage construction costs, DC / DC converter construction costs, DC line construction costs, and fixed operation and maintenance costs. The variable operation cost model includes energy storage cycle degradation costs, DC interconnection system loss costs, and computing center operation costs.

7. A word conversion system for reliable power supply under DC interconnection for computing loads, characterized in that, include: The data acquisition and power consumption decomposition module is used to acquire the operating data and token measurement data of the target computing center, and to perform relational mapping and division of power consumption based on the aggregated business behavior of the computing center to obtain the power consumption mode of the computing center. The minimum configuration determination module is used to establish an equivalent configuration model of the DC interconnection system of the photovoltaic and energy storage computing center. Based on the power consumption mode and reliable power supply requirements of the computing center, it determines the minimum capacity configuration of photovoltaic and energy storage that meets the reliable power supply requirements of the computing center. The cost accounting and decision generation module establishes an integrated operating cost model based on the minimum optical storage configuration to calculate the comprehensive cost within the billing cycle; and combines the token output of the computing center in a cycle to perform unit token cost accounting, break-even token price calculation, and token suggested price decision.

8. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the word conversion method for reliable power supply under DC interconnection for computing loads as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted to be loaded by a processor and executed as described in any one of claims 1-6 for a reliable power supply under DC interconnect for computing loads.

10. A computer device, characterized in that, include: A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the word conversion method under reliable power supply for DC interconnection of computing loads as described in any one of claims 1-6.

Images