A method for converting distributed energy into distributed computing power

Abstract

The present application provides a method for converting distributed energy into distributed computing power, which relates to the field of distributed energy and distributed computing power conversion technology, and is used to improve the utilization efficiency of distributed energy, reduce the cost of acquiring computing power, and achieve coordinated optimization of energy and computing power. The method includes: obtaining energy information of a distributed energy system, wherein the distributed energy includes one or more of solar energy, wind energy, and hydropower, and the energy information includes energy output and energy reserves; obtaining status information of a distributed computing power node, wherein the distributed computing power node includes an electronic device with computing capabilities, and the status information includes the computing capabilities and task execution status of the distributed computing power node; matching a target node based on the energy information, the status information, and the target task, wherein the target node is the distributed computing power node with the highest matching degree with the target task.

Description

Technical Field

[0001] The embodiments of the present application relate to the technical fields of distributed energy and distributed computing power, and in particular, to a method for converting distributed energy into distributed computing power. Background Art

[0002] With the acceleration of digitalization, global demand for computing power is exploding. However, traditional centralized computing power provision models, such as large data centers, face numerous challenges. These data centers not only consume enormous amounts of energy, have high construction and operating costs, but also have stringent requirements for location and infrastructure. Meanwhile, distributed energy sources, such as solar, wind, and tidal power, while offering advantages such as widespread distribution and environmental friendliness, face challenges in energy storage and utilization due to their intermittent and volatile nature. A significant amount of distributed energy is wasted during the production process due to inability to effectively utilize it in a timely manner. Furthermore, a significant portion of energy is lost during the conversion process from DC to AC, resulting in significant resource loss.

[0003] Traditional centralized computing centers also face challenges such as high energy consumption (requiring AC to DC conversion), high construction costs, and geographical restrictions. Meanwhile, distributed energy sources such as solar, wind, and small hydropower are widely distributed, but their energy utilization requires DC to AC conversion before being fed into the grid. This efficiency needs to be further improved, and there are also challenges with energy consumption. Therefore, effectively integrating distributed energy and converting it into usable computing resources has become a pressing technical challenge to address the shortcomings of traditional centralized computing.

[0004] In existing technologies, computing centers generally rely on centralized power grids, which leads to high energy consumption, large carbon emissions, and high deployment costs in remote areas. Furthermore, distributed energy sources (such as wind and photovoltaics) have intermittent power supply characteristics, making it difficult to directly match the demand for high-stability computing power.

[0005] Therefore, there is an urgent need for a technical solution that can deeply integrate distributed energy and computing resources to achieve dynamic matching of energy and computing power. Summary of the Invention

[0006] The embodiments of the present application provide a method for converting distributed energy into distributed computing power, so as to improve the utilization efficiency of distributed energy, reduce the cost of obtaining computing power, and achieve coordinated optimization of energy and computing power.

[0007] To achieve the above objectives, the embodiments of the present application adopt the following technical solutions:

[0008] The present application provides a method for converting distributed energy into distributed computing power, the method comprising: obtaining energy information of a distributed energy system, the distributed energy comprising one or more of solar energy, wind energy and hydropower, the energy information comprising energy output and energy reserves; obtaining status information of a distributed computing power node, the distributed computing power node comprising an electronic device with computing capabilities, the status information comprising the computing capabilities and task execution status of the distributed computing power node; matching a target node based on the energy information, the status information and the target task, the target node being the distributed computing power node with the highest matching degree with the target task.

[0009] In one possible implementation, before obtaining the energy information of the distributed energy system, the method further includes: constructing the distributed energy system based on the distributed energy and the energy storage unit, the energy storage unit including at least one of a battery energy storage unit and a supercapacitor energy storage unit; constructing the distributed computing power node based on an electronic device with computing capability and a configuration strategy, the configuration strategy at least including assigning an identifier and permissions to each of the distributed computing power nodes.

[0010] In a possible implementation, before obtaining the energy information of the distributed energy system, the method further includes:

[0011] Collecting the distributed energy through a preset energy collection device in the distributed energy system;

[0012] The collected distributed energy is preprocessed, and the preprocessing includes at least any one of voltage adjustment, frequency conversion, and rectification and filtering.

[0013] In one possible implementation, the target node is matched based on the energy information, the status information and the target task, including: extracting the task information of the target task, the task information including at least energy consumption, computational complexity and time constraints; and dynamically allocating the target task to the target node through a multi-objective scheduling algorithm based on the status information, the energy information and the task information.

[0014] In one possible implementation, the method further includes: obtaining a characteristic factor of computing power demand growth; constructing a prediction model based on the characteristic factor, the prediction model being used to predict computing power demand; obtaining a matching degree between current energy parameters and historical energy parameters; generating first predicted computing power data based on the matching degree and historical computing power data, the first predicted computing power data including computing power demand for multiple time periods; obtaining an initial value of the characteristic factor based on the first predicted computing power data, current actual computing power data and the prediction model; and predicting computing power demand based on the prediction model corresponding to the initial value of the characteristic factor.

[0015] In one possible implementation, the characteristic factor includes a first characteristic factor and a second characteristic factor. The first characteristic factor is used to characterize the driving effect of energy supply capacity on computing power demand, and the second characteristic factor is used to characterize the coordinated growth of computing power demand caused by network effects.

[0016] In a possible implementation, the prediction model is expressed as:

[0017]

[0018] Among them, C(t) is the computing power demand at time t, E(t) is the energy supply capacity at time t, and E max is the maximum available energy capacity of the system, p is the first characteristic factor, and q is the second characteristic factor.

[0019] In one possible implementation, the initial value of the characteristic factor is obtained based on the first predicted computing power data, the current actual computing power data and the prediction model, including: constructing an intermediate model based on historical energy data; generating second predicted computing power data through the intermediate model; correcting the predicted value according to the deviation between the current actual computing power data and the second predicted computing power data to obtain third predicted computing power data; and obtaining the initial value of the characteristic factor based on the data of the trend node in the third predicted computing power data and the prediction model.

[0020] In one possible implementation, the predicted value is corrected according to the deviation between the current actual computing power data and the second predicted computing power data to obtain the third predicted computing power data, including: obtaining the average deviation between first data in the second predicted computing power data and the current actual computing power data, the first data being the data in the second predicted computing power data corresponding to the current actual computing power data; and correcting the second predicted computing power data according to the average deviation to obtain the third predicted computing power data.

[0021] In one possible implementation, the trend nodes include the time point when energy supply is saturated, the time point when computing power demand changes from linear growth to exponential growth, and the critical point when network effects trigger the coordinated growth of computing power demand.

[0022] In one possible implementation, the method further includes: when it is detected that the energy supply fluctuation rate exceeds a threshold, triggering a dynamic scheduling protocol, the dynamic scheduling protocol including shutting down non-essential high-power consumption computing nodes, migrating tasks to nodes with stable energy supply, and correcting one or more of the second characteristic factors based on the migration results. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A schematic flow chart of the method provided for some embodiments of the present application;

[0024] Figure 2 for Figure 1 Schematic diagram of the process of S400 in some embodiments;

[0025] Figure 3 for Figure 1 FIG. 4 is a flow chart of S450 in some embodiments. DETAILED DESCRIPTION

[0026] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments.

[0027] Hereinafter, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature identified with "first," "second," etc., may explicitly or implicitly include one or more of the features. In the description of this application, unless otherwise specified, "plurality" means two or more.

[0028] In addition, in this application, directional terms such as "up", "down", "left", and "right" may be defined including but not limited to the orientation relative to the schematic placement of the components in the drawings. It should be understood that these directional terms may be relative concepts. They are used for relative descriptions and clarifications, and they may change accordingly according to changes in the orientation of the components in the drawings.

[0029] In this application, unless otherwise specified or limited, the term "connection" should be understood broadly. For example, "connection" can mean fixed connection, detachable connection, or integration; it can mean direct connection or indirect connection through an intermediate medium. In addition, the term "electrical connection" can refer to the method of electrical connection that enables signal transmission.

[0030] As used herein, “about,” “substantially,” or “approximately” includes the stated value and referenced values ​​that are within an acceptable range of deviation from the particular value, where the acceptable range of deviation is determined by one of ordinary skill in the art taking into account the measurements in question and the errors associated with the measurement of the particular quantity (i.e., the limitations of the measurement method).

[0031] In the embodiments of this application, words such as "exemplarily" or "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design described in the embodiments of this application as "exemplarily" or "for example" should not be interpreted as being more preferred or advantageous than other embodiments or designs. Specifically, the use of words such as "exemplarily" or "for example" is not intended to be construed as being more preferred or advantageous than other embodiments or designs.

[0032] With the acceleration of digitalization, global demand for computing power is exploding. However, traditional centralized computing power provision models, such as large data centers, face numerous challenges. These data centers not only consume enormous amounts of energy, have high construction and operating costs, but also have stringent requirements for location and infrastructure. Meanwhile, distributed energy sources, such as solar, wind, and tidal power, while offering advantages such as widespread distribution and environmental friendliness, face challenges in energy storage and utilization due to their intermittent and volatile nature. A significant amount of distributed energy is wasted during the production process due to inability to effectively utilize it in a timely manner. Furthermore, a significant portion of energy is lost during the conversion process from DC to AC, resulting in significant resource loss.

[0033] Traditional centralized computing centers also face challenges such as high energy consumption (requiring AC to DC conversion), high construction costs, and geographical restrictions. Meanwhile, distributed energy sources such as solar, wind, and small hydropower are widely distributed, but their energy utilization requires DC to AC conversion before being fed into the grid. This efficiency needs to be further improved, and there are also challenges with energy consumption. Therefore, effectively integrating distributed energy and converting it into usable computing resources has become a pressing technical challenge to address the shortcomings of traditional centralized computing.

[0034] In existing technologies, computing centers generally rely on centralized power grids, which leads to high energy consumption, large carbon emissions, and high deployment costs in remote areas. Furthermore, distributed energy sources (such as wind and photovoltaics) have intermittent power supply characteristics, making it difficult to directly match the demand for high-stability computing power.

[0035] The embodiment of the present application provides a method for converting distributed energy into distributed computing power, so as to improve the utilization efficiency of distributed energy, reduce the cost of obtaining computing power, and achieve the coordinated optimization of energy and computing power. Figure 1 、 Figure 2 as well as Figure 3 As shown, the method includes:

[0036] S010. Construct the distributed energy system based on the distributed energy and the energy storage unit, wherein the energy storage unit includes at least one of a battery energy storage unit and a supercapacitor energy storage unit.

[0037] For example, in this embodiment, the distributed energy system includes solar energy and wind energy, and the energy storage units include battery energy storage units and supercapacitor energy storage units. Solar energy is converted into DC power by photovoltaic panels, and wind energy is converted into AC power by wind turbines. After both are converted into DC power through rectifiers, a portion is directly supplied to the computing power nodes, and the remaining portion is stored in the battery energy storage units and supercapacitor energy storage units. The battery energy storage unit is used to store large amounts of electricity to meet long-term energy needs, while the supercapacitor energy storage unit is used to store small amounts of electricity, but has the characteristics of rapid charging and discharging to meet high power needs in a short period of time.

[0038] S020. Construct the distributed computing power node based on electronic devices with computing capabilities and a configuration strategy, wherein the configuration strategy at least includes assigning an identifier and permissions to each of the distributed computing power nodes.

[0039] For example, distributed computing nodes consist of electronic devices with computing capabilities, such as servers, personal computers, and industrial control computers. Through policy configuration, each computing node is assigned a unique identifier and permissions. Identifiers can be used to distinguish different computing nodes, and permissions can be used to control access and operational permissions for computing nodes. For example, computing node A is assigned the identifier "Node001" and the permission "High-level Permission," allowing it to access and process high-level computing tasks; while computing node B is assigned the identifier "Node002" and the permission "Intermediate Permission," allowing it to access and process intermediate-level computing tasks.

[0040] S100: Obtain energy information of a distributed energy system. The distributed energy includes one or more of solar energy, wind energy, and hydropower, and the energy information includes energy output and energy reserves.

[0041] For example, in this embodiment, distributed energy sources include solar energy, wind energy, and hydropower. Different distributed energy sources can be collected using different energy collection devices pre-set in the distributed energy system. Specifically, for solar energy, photovoltaic panels are installed on building roofs, open spaces, and other locations to convert light energy into direct current (DC) electricity. The output power of the photovoltaic panels is monitored in real time by power sensors to obtain solar energy output information. Simultaneously, energy storage devices (such as battery storage units) store solar energy, and the remaining capacity of the energy storage devices is monitored in real time by a battery management system to obtain energy reserve information. For wind energy, wind turbines are used to convert wind energy into AC electricity, and the output power of the wind turbines is monitored in real time by an electric meter to obtain wind energy output information. The energy storage devices are also used to store wind energy, and their remaining capacity is also monitored in real time by the battery management system. For hydropower, a turbine drives a generator to generate electricity, and the generator's output power is monitored in real time by a power sensor to obtain hydropower energy output information. The energy storage devices are used to store hydropower electricity, and their remaining capacity is also monitored in real time by the battery management system.

[0042] It is understandable that after collecting the relevant energy, in order to ensure normal use, the collected distributed energy must be preprocessed. The preprocessing includes at least one of voltage adjustment, frequency conversion and rectification filtering to make it meet the requirements of subsequent energy storage and use. At the same time, it is also convenient to use sensors to monitor the operating status and energy output data of the energy collection equipment in real time.

[0043] In some embodiments, when monitoring the status of an energy harvesting device, various methods can be used. For example, the temperature, impedance, voltage, current, and other status data of the energy harvesting device are first normalized, and then the normalized values ​​are used as their grayscale values. The status data are then filled into a preset image area as pixels according to the grayscale values ​​to obtain a status image. The status image is then input into a pre-trained neural network, and a status evaluation value is obtained through its sofxmax function. When the status evaluation value is within a preset range, the status of the device is determined to be normal; otherwise, it is determined to be abnormal. Alternatively, a status matrix is ​​constructed based on the normalized status data, and the correlation value between the current status matrix and the historical status matrix at the same time is calculated. When the correlation value is within a preset range, the device is determined to be normal; otherwise, it is determined to be abnormal. The information exchange efficiency between the energy storage device and the energy collection device can also be calculated to determine the status of the energy collection device and the energy storage device. For example, the information exchange frequency of the information exchange interface corresponding to the energy storage device and the energy collection device, the amount of information in the communication packet, and the envelope composition of the recorded information packet are calculated. Then, a communication state matrix is ​​constructed based on the information exchange frequency, amount of information, and envelope composition. Then, the correlation value or similarity value between the communication state matrix and the standard state matrix is ​​calculated. When the correlation value or similarity value is within a preset range, it is determined that the energy storage device and the energy collection device are normal, otherwise it is determined to be abnormal. Of course, in order to meet the monitoring requirements, those skilled in the art can also choose other feasible monitoring methods besides the examples described in this application.

[0044] S200. Obtain status information of a distributed computing power node, where the distributed computing power node includes an electronic device with computing capabilities, and the status information includes the computing capabilities and task execution status of the distributed computing power node.

[0045] For example, distributed computing nodes include servers, personal computers, industrial control computers, and other electronic devices with computing capabilities. By installing performance monitoring software on the computing nodes, computing power information of the computing nodes, such as CPU utilization, memory occupancy, and graphics card performance, can be obtained in real time. At the same time, the task execution status of the computing nodes can be obtained in real time, including the number of tasks running, task type, and task progress. For example, for a server computing node, the performance monitoring software shows that its CPU utilization is 60% and its memory utilization is 70%. The task management module shows that it is running 5 tasks, including 2 data processing tasks and 3 model training tasks. Among them, one data processing task is 50% complete, two model training tasks are 30% complete, and the other two tasks have just started.

[0046] It is understandable that when distributing computing power, the hardware and software of the electronic equipment can be upgraded and adapted first so that it can operate stably in a distributed energy supply environment. The adapted devices can then be used as distributed computing power nodes and connected to the distributed computing power network through wired or wireless communication networks. A unique identification and permissions are assigned to each node to ensure communication security and smooth data interaction between nodes; or through software-defined methods, the computing resources can be integrated into a computing power network that can be uniformly scheduled to achieve efficient application of computing power.

[0047] S300, matching a target node based on the energy information, the state information, and the target task, wherein the target node is the distributed computing power node with the highest matching degree with the target task. S300 includes:

[0048] S310: Extract task information of the target task, where the task information at least includes energy consumption, computational complexity, and time constraints.

[0049] For example, when a user submits a target task, the task parsing module extracts key information about the task. For example, for a deep learning model training task, the estimated energy consumption is 100 kWh, the computational complexity is 10^12 floating-point operations, and the time constraint is to complete within 48 hours.

[0050] S320 , dynamically allocating the target task to the target node through a multi-objective scheduling algorithm according to the state information, the energy information, and the task information.

[0051] For example, based on the computing node's status information (such as computing power and task execution status), the energy system's energy information (such as energy output and energy reserves), and the target task's task information (such as energy consumption, computational complexity, and time constraints), a multi-objective scheduling algorithm is used to calculate the matching degree between each computing node and the target task. The computing node with the highest matching degree is designated as the target node, and the target task is dynamically assigned to the target node for execution.

[0052] For example, computing node A has a computing capacity of 80 TFLOPS, is currently performing relatively light tasks, and has sufficient energy reserves. Computing node B has a computing capacity of 120 TFLOPS, but is currently performing relatively busy tasks and has low energy reserves. The target task requires 100 TFLOPS of computing power, with an estimated energy consumption of 50 kWh and a time constraint of 24 hours. The multi-objective scheduling algorithm calculates that computing node A is more compatible with the target task than computing node B. Therefore, computing node A is selected as the target node and the target task is assigned to it for execution.

[0053] In some examples, the matching degree of each parameter is calculated as follows:

[0054] Computing capability matching Among them, C node is the node’s current available computing power (TFLOPS), C task The computing power requirement for the task (TFLOPS).

[0055] Energy reserve matching Among them, E node is the remaining available energy of the node (kWh), E task Estimate the energy consumption (kWh) for the task.

[0056] Task load matching Among them, T running is the number of tasks being executed by the node, T max The maximum concurrent task capacity of the node.

[0057] Time constraint matching Among them, t predicted is the estimated time for the node to complete the task (hours), t task is the task time constraint (hours).

[0058] Comprehensive matching degree S z This can be obtained by weighted summation:

[0059] S z =w1×S compute +w2×S energy +w3×S load +w4× time

[0060] Where w1-w4 are weight coefficients, and their sum is 1.

[0061] By quantifying the matching degree of computing power, energy reserves, task load and time constraints, and combining them with weight distribution, this method can accurately select the optimal node to ensure efficient task execution and rational energy utilization.

[0062] It can be understood that calculating the matching degree between each performance parameter and the task parameter and weighting to obtain the comprehensive matching degree is a conventional technical means for technical personnel in this field. This application does not limit this. Technical personnel in this field can choose the scheduling algorithm for allocating target nodes on their own under the guidance of this application.

[0063] Computing power demand forecasting technology is a crucial component of distributed energy management. Its purpose is to predict computing power demand trends by analyzing historical data and energy factors, thereby providing a scientific basis for the rational allocation of computing power resources. Existing computing power demand forecasting methods primarily include empirical prediction, statistical model prediction, and machine learning prediction.

[0064] While existing computing power demand forecasting technologies can provide a certain level of support for distributed energy management, they often suffer from shortcomings that lead to poor prediction results. For example, existing forecasting models often rely on simple extrapolation of historical data, resulting in significant deviations between the predicted results and actual events. Furthermore, many forecasting models rely heavily on the accumulation of extensive historical data, limiting their applicability and accuracy in areas with incomplete or lacking historical data. Furthermore, the growth of computing power demand is influenced by multiple factors, such as energy supply capacity and network effects, and existing models often struggle to accurately simulate the dynamic impact of these factors on computing power demand.

[0065] Therefore, the prediction method in the prior art has a poor prediction effect on computing power demand. The method of the embodiment of the present application also includes the following steps to improve the problem that the prediction method in the prior art has a poor prediction effect on computing power demand.

[0066] S410: Obtain characteristic factors of computing power demand growth. The characteristic factors include a first characteristic factor and a second characteristic factor, wherein the first characteristic factor is used to characterize the driving effect of energy supply capacity on computing power demand, and the second characteristic factor is used to characterize the coordinated growth of computing power demand caused by network effects.

[0067] The growth of computing power demand is influenced by multiple factors, the most critical of which are energy supply capacity and network effects. Energy supply capacity determines the upper limit of computing power resources, while network effects affect the coordinated growth of computing power demand.

[0068] Energy supply capacity is the foundation for the growth of computing power demand. Increased energy supply capacity can support the operation of more computing equipment, directly driving the growth of computing power demand. For example, increases in distributed energy generation and improvements in the remaining capacity of energy storage systems will enhance energy supply capacity, thereby promoting the growth of computing power demand.

[0069] Network effects refer to the mutual influence and coordinated growth of computing power demand among users. As more and more users connect to the computing network, the computing resources and service quality available to each user improve, attracting even more users and leading to a coordinated growth in computing power demand. For example, an increase in the number of users on a cloud computing platform will drive an overall increase in the platform's computing power demand. Improved task collaboration efficiency among computing nodes can also lead to a positive feedback effect, increasing overall computing power demand.

[0070] Therefore, this application uses the driving effect of energy supply capacity on computing power demand as the first characteristic factor and the coordinated growth of computing power demand caused by network effects as the second characteristic factor. The first and second characteristic factors jointly determine the growth trend of computing power demand.

[0071] S420: Build a prediction model based on the characteristic factors, where the prediction model is used to predict computing power requirements:

[0072] For example, a prediction model is constructed based on the extracted characteristic factors. The prediction model is expressed as:

[0073]

[0074] This prediction model comprehensively considers the impact of energy supply capacity and network effect on computing power demand. Among them, C(t) is the computing power demand at time t (TFLOPS / hour), E(t) is the energy supply capacity at time t (kW), and E max is the maximum available energy capacity of the system, p is the first characteristic factor, which reflects the driving effect of energy supply capacity on computing power demand, and q is the second characteristic factor, which reflects the coordinated growth of computing power demand caused by network effects.

[0075] This prediction model describes the dynamic changes in computing power demand over time, as well as the contribution of energy supply capacity and network effects to the growth of computing power demand. Specifically, the first term p·(E max -E(t)) describes the driving effect of energy supply capacity on computing power demand. As energy supply capacity increases, unused energy capacity (E max -E(t)) decreases, the growth rate of computing power demand also changes accordingly. The second This model describes the coordinated growth of computing power demand driven by network effects. As energy supply capacity increases, network effects gradually emerge, influencing the growth rate of computing power demand. Furthermore, this forecasting model considers the growth of computing power demand as a dynamic system influenced by both energy supply capacity and network effects. It accurately captures the nonlinear characteristics of computing power demand and reflects the complex relationships involved in computing power demand growth.

[0076] S430: Obtain the degree of match between current energy parameters and historical energy parameters. These energy parameters include distributed energy generation power, remaining energy storage system capacity, energy volatility, and other factors. Changes in these parameters directly impact energy supply capacity, which in turn affects the growth of computing power demand.

[0077] For example, by deploying various sensors and monitoring equipment in a distributed energy system, parameters such as the generated power of distributed energy, the remaining capacity of the energy storage system, and the energy volatility can be obtained in real time. For example, the solar power generation power is obtained through the power sensor on the photovoltaic panel, the remaining capacity of the energy storage system is obtained through the energy storage battery management system, and the energy volatility is calculated using data analysis software. The currently acquired energy parameter data is compared with the historical energy parameter data, and the similarity between the two is quantified using the Pearson correlation coefficient to generate a matching score. For example, the Pearson correlation coefficient between the current solar power generation curve and the solar power generation curve for each day in the past week is calculated to obtain a matching score to assess the similarity between the current energy status and the historical energy status. A higher score indicates a higher similarity between the current energy parameters and the historical energy parameters.

[0078] For example, in some embodiments, the calculation method is:

[0079]

[0080] Among them, x i and y i are the i-th data point of current energy parameters and historical energy parameters respectively, and are the average values ​​of current energy parameters and historical energy parameters, respectively, and ρ is the similarity.

[0081] S440: Generate first predicted computing power data based on the matching degree and historical computing power data, where the first predicted computing power data includes computing power requirements for multiple time periods.

[0082] Exemplarily, based on the degree of match between current and historical energy parameters, combined with historical computing power data, a first predicted computing power data is generated through weighted averaging or other suitable prediction methods to predict computing power demand for multiple time periods in the future. For example, based on an 80% match between current and historical energy parameters, combined with computing power demand data for corresponding time periods within the past week, a first predicted computing power data is generated through a weighted averaging method to predict computing power demand for each hour in the next 24 hours. Alternatively, based on the matching score, the computing power data corresponding to the historical energy parameters with the highest degree of match with the current energy parameters is selected as reference data. The computing power demand in the reference data is used as the first predicted computing power data. It can be understood that this data includes computing power demand for multiple time periods and is used to preliminarily predict the computing power demand trend under the current circumstances.

[0083] S450: Obtaining an initial value of the characteristic factor based on the first predicted computing power data, the current actual computing power data, and the prediction model. S450 includes:

[0084] S451. Build an intermediate model based on historical energy data.

[0085] For example, historical energy data is used to construct an intermediate model for predicting computing power demand through time series analysis or other appropriate modeling methods. For example, based on the energy data of the past year, an ARIMA model is used to construct an intermediate model to predict the computing power demand trend over the next period of time. In other examples, the model can also be composed of multiple weighted prediction models, such as BP neural network models, fuzzy neural network models, and adaptive probabilistic neural network models. The weight of each model is determined by the accuracy of its prediction results, with the higher the accuracy, the greater the weight.

[0086] S452: Generate second predicted computing power data using the intermediate model.

[0087] Exemplarily, current energy data is input into the intermediate model to generate second predicted computing power data. For example, current solar power generation data is input into the intermediate model to obtain hourly computing power demand forecast data for the next 24 hours.

[0088] S453: Correct the predicted value based on the deviation between the current actual computing power data and the second predicted computing power data to obtain third predicted computing power data. S453 includes:

[0089] S4531. Obtain an average deviation between first data in the second predicted computing power data and the current actual computing power data, where the first data is data in the second predicted computing power data corresponding to the current actual computing power data.

[0090] For example, data corresponding to the current actual computing power data is selected from the second predicted computing power data, and the average deviation between the two is calculated. For example, data for each hour in the past 24 hours from the second predicted computing power data is selected and compared with the current actual computing power data, and the average deviation is calculated to be 8 TFLOPS / hour.

[0091] S4532. Correct the second predicted computing power data according to the average deviation to obtain the third predicted computing power data.

[0092] For example, a weighted algorithm is used to dynamically compensate for the calculated average deviation to generate third predicted computing power data. For example, based on an average deviation of 8 TFLOPS / hour, the second predicted computing power data is compensated by a weighted algorithm to generate third predicted computing power data, thereby improving the accuracy of the prediction. In some examples, the following algorithm can be used for calculation:

[0093] C 修正 (t) = C 预测 (t)+α·deviation

[0094] Among them, C 修正 (t) is the corrected computing power requirement, C 预测 (t) is the second predicted computing power data, α is a weighting coefficient, which is formulated by technical personnel in this field according to needs. In this embodiment, it can be 0-1, such as 1. The deviation is the average deviation between the second predicted computing power data and the current actual computing power data.

[0095] S454. Obtain an initial value of the characteristic factor based on the data of the trend node in the third predicted computing power data and the prediction model.

[0096] Exemplarily, the trend nodes include the time point when energy supply is saturated, the time point when computing power demand changes from linear growth to exponential growth, and the critical point when network effects trigger the coordinated growth of computing power demand.

[0097] The time point of energy supply saturation refers to the time point when the energy supply reaches its maximum value. The time point when computing power demand changes from linear growth to exponential growth refers to the time point when the growth rate of computing power demand changes from linear growth to exponential growth. The critical point when network effect triggers the coordinated growth of computing power demand refers to the time point when network effect begins to significantly affect the growth of computing power demand.

[0098] Substitute the data from these nodes into the prediction model to calculate the initial values ​​of the characteristic factors. For example, extract the energy supply inflection point data and the peak computing power demand data from the third predicted computing power data, substitute them into the prediction model, and calculate the initial values ​​of the characteristic factors.

[0099] S460: Predict computing power requirements based on the prediction model corresponding to the initial value of the characteristic factor.

[0100] For example, the initial values ​​of the characteristic factors are substituted into the prediction model to predict the computing power demand in the future. For example, the initial values ​​of the characteristic factors (such as p = 0.6, q = 0.4) are substituted into the prediction model to predict the computing power demand for each day in the next 7 days, providing a basis for the scheduling and allocation of distributed computing power resources.

[0101] S500: When it is detected that the energy supply fluctuation rate exceeds a threshold, a dynamic scheduling protocol is triggered.

[0102] For example, the monitoring system monitors energy supply fluctuations in real time, and when the fluctuation exceeds a preset threshold, the dynamic dispatch protocol is triggered. For example, when the wind power generation fluctuation exceeds 30%, the dynamic dispatch protocol is triggered.

[0103] For example, based on the importance of the computing nodes' tasks and their energy consumption, non-essential, high-power computing nodes can be shut down to reduce overall energy consumption and ensure the normal operation of critical tasks. For example, high-power computing nodes performing non-urgent tasks, such as those performing low-priority data analysis tasks, can be shut down.

[0104] For example, tasks originally executed on high-power computing nodes can be migrated to computing nodes with stable energy supply based on the characteristics and requirements of the tasks, ensuring task continuity and stability. For example, an ongoing AI training task can be migrated from a computing node affected by energy fluctuations to another computing node with stable energy supply.

[0105] For example, based on the distribution of computing nodes and task execution after task migration, the coordinated growth of computing power demand caused by network effects is reassessed, and the second characteristic factor (network effect coefficient) is corrected. For example, after task migration, the collaboration pattern between computing power nodes changes. By analyzing the new task distribution and collaboration frequency, as well as the change in computing power demand after task migration, a new network effect coefficient is calculated and substituted into the prediction model to improve the accuracy of the prediction model.

[0106] In summary, the present application can achieve at least one or more of the following beneficial effects:

[0107] 1. Improve energy utilization efficiency: Distributed green energy such as solar energy, wind energy, and tidal energy, which were originally difficult to effectively utilize, can be efficiently converted into valuable computing power resources, reducing energy waste, maximizing benefits, and improving the comprehensive utilization efficiency of distributed green energy.

[0108] 2. Reduce computing power costs: Low cost is a basic requirement for the popularization of artificial intelligence technology. Using distributed energy and ordinary equipment to build a distributed computing power network avoids the high costs of large-scale construction and operation of centralized data centers. This distributed computing power equipment is easier to deploy. Distributed computing power construction solves the one-time large investment problem of centralized computing power construction (obtaining large benefits with small capital investment), lowering the threshold for obtaining computing power.

[0109] 3. Enhance system flexibility and stability: The distributed energy and computing power architecture enables the system to be flexibly adjusted according to changes in energy and demand, enhancing the system's risk resistance and stability.

[0110] 4. Promote the integrated development of energy and information technology: This application provides new computing power approaches and methods for the deep integration of the energy and information technology fields, improves the efficiency of distributed green energy use, and makes up for the lack of centralized computing power resources.

[0111] Through the description of the above embodiments, those skilled in the art can clearly understand that the diagnostic method in the above embodiments can be implemented by means of software plus a necessary general hardware platform, or of course by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the relevant technology, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods of various embodiments of the present invention.

[0112] Through the description of the above implementation methods, technical personnel in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0113] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of modules or units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0114] Units described as separate components may or may not be physically separate, and components shown as units may be one physical unit or multiple physical units, that is, they may be located in one place or distributed in multiple places. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0115] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated units may be implemented in the form of hardware.

[0116] The above content is only a specific embodiment of this application, but the scope of protection of this application is not limited to this. Any changes or replacements within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A method for converting distributed energy into distributed computing power, characterized in that: The method comprises: Acquiring energy information of a distributed energy system, wherein the distributed energy includes one or more of solar energy, wind energy, and hydropower, and the energy information includes energy output and energy reserves; Obtaining status information of a distributed computing power node, wherein the distributed computing power node includes an electronic device with computing capabilities, and the status information includes the computing capabilities and task execution status of the distributed computing power node; Matching a target node based on the energy information, the state information, and the target task, the target node being the distributed computing power node with the highest matching degree with the target task; Before obtaining the energy information of the distributed energy system, the method further includes: Constructing the distributed energy system based on the distributed energy and the energy storage unit, wherein the energy storage unit includes at least one of a battery energy storage unit and a supercapacitor energy storage unit; Constructing the distributed computing power nodes based on electronic devices with computing capabilities and a configuration strategy, wherein the configuration strategy at least includes assigning an identifier and permissions to each of the distributed computing power nodes; The matching of target nodes based on the energy information, the state information, and the target task includes: Extracting task information of the target task, wherein the task information at least includes energy consumption, computational complexity, and time constraint; Dynamically assigning the target task to the target node through a multi-objective scheduling algorithm according to the state information, the energy information and the task information; The method further comprises: Obtain characteristic factors of computing power demand growth; A prediction model is constructed based on the characteristic factors. The prediction model is used to predict computing power requirements. The expression of the prediction model is: Among them, C(t) is the computing power demand at time t, E(t) is the energy supply capacity at time t, and E max is the maximum available energy capacity of the system, p is the first characteristic factor, and q is the second characteristic factor.

2. The method according to claim 1, characterized in that Before obtaining the energy information of the distributed energy system, the method further includes: Collecting the distributed energy through a preset energy collection device in the distributed energy system; The collected distributed energy is preprocessed, and the preprocessing includes at least any one of voltage adjustment, frequency conversion, and rectification and filtering.

3. The method according to claim 1, characterized in that The method further comprises: Obtain the matching degree between current energy parameters and historical energy parameters; generating first predicted computing power data based on the matching degree and the historical computing power data, wherein the first predicted computing power data includes computing power requirements for multiple time periods; Obtaining an initial value of the characteristic factor according to the first predicted computing power data, the current actual computing power data, and the prediction model; The computing power demand is predicted according to the prediction model corresponding to the initial value of the characteristic factor.

4. The method according to claim 3, characterized in that The characteristic factors include a first characteristic factor and a second characteristic factor. The first characteristic factor is used to characterize the driving effect of energy supply capacity on computing power demand, and the second characteristic factor is used to characterize the coordinated growth of computing power demand caused by network effects.

5. The method according to claim 3, characterized in that Obtaining the initial value of the characteristic factor according to the first predicted computing power data, the current actual computing power data, and the prediction model includes: Build an intermediate model based on historical energy data; Generate second predicted computing power data using the intermediate model; Correct the predicted value based on the deviation between the current actual computing power data and the second predicted computing power data to obtain the third predicted computing power data; An initial value of the characteristic factor is obtained based on the data of the trend node in the third predicted computing power data and the prediction model.

6. The method according to claim 5, characterized in that The step of correcting the predicted value according to the deviation between the current actual computing power data and the second predicted computing power data to obtain the third predicted computing power data includes: Obtaining an average deviation between first data in the second predicted computing power data and the current actual computing power data, where the first data is data in the second predicted computing power data corresponding to the current actual computing power data; The second predicted computing power data is corrected according to the average deviation to obtain the third predicted computing power data.

7. The method according to claim 5, characterized in that The trend nodes include the time point when energy supply is saturated, the time point when computing power demand changes from linear growth to exponential growth, and the critical point when network effects trigger the coordinated growth of computing power demand.

Images