Hydropower station fusion computing power intelligent server system

Abstract

The invention provides a hydropower station fusion computing power intelligent server system, and relates to the technical field of carbon emission. The method comprises the following steps: an acquisition correction module acquires hydropower station generation power data and server cluster instantaneous energy consumption data in real time, establishes a time sequence corresponding relation between a power generation side and a power utilization side, and obtains clean energy supply quantity data; the prediction scheduling module generates a non-tampering carbon footprint tracking chain number, and predicts a clean energy supply fluctuation trend sequence by adopting a long short-term memory network algorithm; training a random forest model to predict a computing power demand trend sequence; optimizing task scheduling according to a carbon emission minimization target, and generating dynamic balance configuration data of a computing power task and clean energy supply; and the monitoring evaluation module outputs computing power task level carbon emission and a clean energy use ratio in real time through an application programming interface according to the dynamic balance configuration data, and generates quantitative evaluation data of a computing power service carbon neutralization target by adopting a data visualization technology monitoring technology.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of carbon emissions, in particular to a hydropower station fusion computing power intelligent server system. BACKGROUND

[0002] With the rapid development of digital economy, the energy consumption and carbon emissions of computing power infrastructure have become increasingly prominent. To achieve the "double carbon" goal, it is urgent to accurately quantify and track the carbon footprint of computing power services.

[0003] The existing mainstream carbon emission monitoring and calculation methods are as follows: 1. The current average carbon emission factor of regional power grid is widely relied on for estimation. The "one-size-fits-all" method completely ignores the essential difference between clean energy such as hydropower and thermal power in carbon emissions, resulting in overestimation of the carbon footprint of computing power services using clean energy, which cannot truly reflect the environmental value of low-carbon operation, and the data lacks accuracy and credibility. 2. The traditional monitoring system regards the power generation link of the hydropower station and the energy consumption link of the data center as independent islands, lacking effective correlation mechanism. The clean energy output on the power generation side is volatile, while the energy consumption on the demand side of computing power changes dynamically. The mismatch between the two in time and space makes it extremely difficult to establish a full-link carbon footprint mapping relationship from "green electricity" production to computing power consumption. 3. A computing power service is usually completed by multiple tasks in a server cluster. The resource occupation and energy consumption patterns of different tasks differ greatly. Existing technologies can only measure the overall energy consumption of servers or clusters, and cannot accurately allocate carbon emissions to specific single computing tasks. The ambiguity of the mapping relationship between tasks and carbon emissions makes it impossible for users to know the actual carbon cost of their specific computing tasks, hindering fine carbon management and green scheduling. SUMMARY

[0004] The main purpose of the present application is to provide a hydropower station fusion computing power intelligent server system, which solves the technical problems of inaccurate carbon emission factor, incomplete whole process link and inaccurate task level allocation in the prior art.

[0005] To solve the above technical problems, the technical solution adopted by the present application is: a hydropower station fusion computing power intelligent server system, comprising: The acquisition and correction module is used to collect the power generation data of the hydropower station and the instantaneous energy consumption data of the server cluster in real time through the Internet of Things sensor, synchronize the data timestamp by using the network time protocol, establish the time sequence correspondence relationship between the power generation side and the power consumption side, calculate the transmission loss coefficient according to the line impedance, voltage level and geographical distance parameters of the power grid, deduce the transmission efficiency of clean energy, and obtain the corrected clean energy supply data. The prediction scheduling module is configured to generate unforgeable carbon footprint tracking chain data based on the corrected clean energy supply data; predict a clean energy supply fluctuation trend sequence in a preset time period based on the unforgeable carbon footprint tracking chain data and by using a long short-term memory network algorithm; train a random forest model to predict a computing power demand trend sequence based on server historical energy consumption data and computing power task carbon footprint quantification data; and if the predicted clean energy supply fluctuation trend sequence is lower than the computing power demand trend sequence, optimize task scheduling based on a carbon emission minimization objective, and generate dynamic balance configuration data of computing power tasks and clean energy supply. The monitoring and evaluation module is configured to output computing power task level carbon emissions and clean energy usage ratio in real time through an application programming interface based on the dynamic balance configuration data of computing power tasks and clean energy supply, and generate carbon footprint monitoring dashboard data including line graphs and column charts by using a data visualization technology; if the carbon emissions exceed a preset daily carbon emission threshold, trigger an alarm and adjust the priority of computing power tasks, and generate quantification evaluation data of the carbon neutralization objective of computing power services.

[0006] In a preferred embodiment, the transmission loss coefficient is calculated based on the line impedance, voltage level and geographical distance parameters of the power grid, the clean energy transmission efficiency is derived, and the corrected clean energy supply data is obtained, specifically including: The total resistance of the transmission line is calculated according to the formula: ; In the formula, is the total resistance of the transmission line, is the unit length resistance value of the cable material at standard temperature, is the geographical distance of the power transmission path from the hydropower station to the server cluster; The power loss in the form of heat energy consumed on the line is calculated based on the total resistance and the average current flowing through the line, according to the formula: ; In the formula, is the average current value through the transmission line; Based on the power loss and the original power generation power of the hydropower station, the loss coefficient is calculated, and the complement of the transmission efficiency is calculated, according to the formula: Transmission loss coefficient: ; Clean energy transmission efficiency: ; In the formula, is the original power generation power collected in real time by the hydropower station; The transmission efficiency is applied to the original power generation data of the hydropower station to obtain the corrected clean energy quantity after transmission loss, according to the formula: .

[0007] In the preferred solution, the generating of the non-tamperable carbon footprint tracking chain data in the prediction scheduling module comprises: According to the modified clean energy supply data and the server cluster instantaneous energy consumption data, a dynamic carbon emission factor calculation model is constructed; If the modified clean energy supply is greater than or equal to the server cluster instantaneous energy consumption, the dynamic carbon emission factor is set to the preset clean energy carbon emission coefficient; If the modified clean energy supply is less than the server cluster instantaneous energy consumption, according to the proportion of clean energy supply and thermal power supplement, the hybrid dynamic carbon emission factor is calculated by weighting average based on the preset clean energy carbon emission coefficient and the thermal power carbon emission coefficient; and a real-time dynamic carbon emission factor sequence is generated.

[0008] In the preferred solution, the specific formula for calculating the hybrid carbon emission factor is: ; In the formula, is the hybrid carbon emission factor, is the hydropower supply, is the thermal power supplement, is the total energy consumption.

[0009] In the preferred solution, the generating of the non-tamperable carbon footprint tracking chain data in the prediction scheduling module further comprises: according to the real-time dynamic carbon emission factor sequence, the central processor utilization rate, the memory occupancy rate and the network bandwidth usage rate are obtained through the server performance monitoring interface, a nonlinear mapping model of hardware resource consumption and power consumption is constructed by using a polynomial regression algorithm, and the real-time power consumption distribution value sequence of a single computing power task is obtained by associating and calculating the task unique identifier.

[0010] In the preferred solution, the specific formula for constructing the nonlinear mapping model of hardware resource consumption and power consumption is: ; In the formula, is the power consumption of the i-th task, A is the central processor utilization rate, B is the memory occupancy rate, D is the network bandwidth usage rate, a is the central processor quadratic power consumption response coefficient, b is the memory linear power consumption weight coefficient, and c is the network bandwidth power consumption contribution coefficient.

[0011] In the preferred solution, the calculation formula of the central processor quadratic power consumption response coefficient a is: ; In the formula, a is a central processor quadratic power consumption response coefficient, T is the duration length of collecting data and evaluating the system state, S is the number of arithmetic logic units participating in calculation during task execution, R is the voltage value maintained by the server central processor core during task execution, G is the average stress of instruction parallel computing resources during the execution of the current task, and t is the time.

[0012] In a preferred scheme, the calculation formula of the memory linear power consumption weight coefficient b is: ; In the formula, b is a memory linear power consumption weight coefficient, U is the average memory usage of the current task, is the page fault rate change amount, V is the memory fragmentation rate, and Y is the hot page proportion. c = 1-a-b; In the formula, c is the network bandwidth power consumption contribution coefficient.

[0013] In a preferred scheme, the specific formula for calculating the instantaneous carbon emission of the computing power task level is: ; In the formula, E is the instantaneous carbon emission of the i-th task, P i is the task power consumption, and F is the carbon emission factor at time t.

[0014] In a preferred scheme, in the prediction scheduling module, generating the non-tamperable carbon footprint tracking chain data further includes: According to the real-time power consumption distribution value sequence of a single computing power task and the dynamic carbon emission factor sequence at the corresponding time, the instantaneous carbon emission of the computing power task level is calculated through product operation, the total carbon emission within the task execution period is accumulated and calculated through time integration algorithm, and the computing power task carbon footprint quantification data is obtained; According to the computing power task carbon footprint quantification data, the hash values of the hydropower station power generation data, the server instantaneous energy consumption data and the computing power task carbon emission are recorded by using the block chain distributed ledger technology, the data integrity is verified by the smart contract, and the on-chain hash value is mapped with the off-chain original data based on the timestamp and digital signature, to generate the non-tamperable carbon footprint tracking chain data.

[0015] This invention provides a hydropower station integrated computing power intelligent server system, comprising: a data acquisition and correction module that collects real-time hydropower generation power data and instantaneous energy consumption data of the server cluster, synchronizes data timestamps, and establishes a time-series correspondence between the power generation side and the power consumption side; acquiring corrected clean energy supply data; a prediction and scheduling module that generates an immutable carbon footprint tracking chain number based on the corrected clean energy supply data, and then uses a long short-term memory network algorithm to predict the clean energy supply fluctuation trend sequence within a preset time period; training a random forest model to predict the computing power demand trend sequence; if the predicted clean energy supply fluctuation trend sequence is lower than the computing power demand trend sequence, optimizing task scheduling according to the carbon emission minimization target, and generating dynamic balance configuration data between computing power tasks and clean energy supply; and a monitoring and evaluation module that... The dynamic balance configuration data of computing power tasks and clean energy supply is output in real time through the application programming interface (API) to measure carbon emissions and clean energy usage ratios at the computing power task level. Data visualization technology is used to generate a carbon footprint monitoring dashboard containing line charts and bar charts. Based on preset daily carbon emission thresholds, if carbon emissions exceed the threshold, an alarm is triggered and the priority of computing power tasks is adjusted, generating quantitative assessment data of the carbon neutrality target for computing power services. A dynamic carbon emission factor model is constructed, changing the extensive calculation mode that relies on fixed factors, accurately reflecting the actual carbon reduction effect of clean energy use, and improving the measurement accuracy of energy consumption and carbon emissions. Intelligent collaborative scheduling of computing power and clean energy achieves a dynamic balance between computing power supply and clean energy supply in time and space, improving the real-time performance and operability of carbon management. Attached Figure Description

[0016] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a system module connection diagram of the present invention. Detailed Implementation

[0017] Example 1 like Figure 1 As shown, a hydropower station integrated computing power intelligent server system includes: The data acquisition and correction module is used to collect real-time power generation data of hydropower stations and instantaneous energy consumption data of server clusters through IoT sensors. It uses the Network Time Protocol to synchronize data timestamps and establish a time sequence correspondence between the power generation side and the power consumption side. Based on the power grid line impedance, voltage level and geographical distance parameters, it calculates the transmission loss coefficient, derives the clean energy transmission efficiency, and obtains the corrected clean energy supply data.

[0018] The prediction scheduling module is configured to generate unforgeable carbon footprint tracking chain data based on the corrected clean energy supply data, predict a clean energy supply fluctuation trend sequence in a preset time period based on the unforgeable carbon footprint tracking chain data by using a long short-term memory network algorithm, train a random forest model to predict a computing power demand trend sequence based on historical energy consumption data of the server and carbon footprint quantification data of the computing power task, and generate dynamic balance configuration data of the computing power task and the clean energy supply based on task scheduling optimization according to a carbon emission minimization target if the predicted clean energy supply fluctuation trend sequence is lower than the computing power demand trend sequence.

[0019] The monitoring and evaluation module is configured to output computing power task level carbon emissions and clean energy usage ratio in real time through an application programming interface based on the dynamic balance configuration data of the computing power task and the clean energy supply, generate carbon footprint monitoring instrument panel data including line graphs and column graphs by using a data visualization technology, and generate quantified evaluation data of the carbon neutralization target of the computing power service based on a preset daily carbon emission threshold and by triggering an alarm and adjusting the priority of the computing power task if the carbon emissions exceed the threshold.

[0020] The embodiment integrates the clean energy collection, energy consumption calculation, prediction scheduling and carbon footprint evaluation functions of the hydropower station by using a modular structure to form a computing power intelligent service system.

[0021] In the collection correction module, the Internet of Things sensors are distributed and installed at the power supply interfaces of the main power generation equipment of the hydropower station and the server room, and the power generation power output value and the instantaneous energy consumption value of the server cluster are collected respectively. All data are uploaded to the edge computing node through the local area network or the 5G Internet of Things gateway, and the timestamps of all sensor data are calibrated by using the Network Time Protocol (NTP) to ensure that different data sources have a unified time reference. On this basis, the transmission loss coefficient of each power supply path is calculated by using a linear approximation model or a complex impedance model in combination with the power grid topology structure and the physical transmission parameters (including resistance, inductance, distance, load density, etc.). The corrected clean energy supply data is obtained by multiplying the power generation end data and the transmission efficiency to realize dynamic quantification of the real available energy.

[0022] The prediction scheduling module uses the corrected supply data as input to generate carbon footprint chain basic data by using an encryption hash algorithm (such as SHA-256), and constructs an unforgeable carbon footprint tracking structure in combination with the time sequence hash identifier. Subsequently, the clean energy supply time sequence of the chain is taken as a training sample and input into a long short-term memory network (LSTM) model for 24-hour rolling prediction. The LSTM model structure includes an input layer, a bidirectional LSTM hidden layer and a time-dense connection output layer, which can capture the short-term fluctuation and long-term trend characteristics of the energy supply.

[0023] In terms of computing power demand prediction, historical server energy consumption logs and past computing power task carbon emission calibration data are used to construct a feature vector set, which is input into a random forest model to train a predictor. This model uses multiple decision tree ensembles to determine future computing power task intensity at different times, achieving multivariate nonlinear trend prediction. The prediction results form a supply-demand trend comparison matrix, based on which it is determined whether there is a supply gap.

[0024] If the energy supply is insufficient, a scheduling model based on heuristic or evolutionary optimization algorithms (such as particle swarm optimization or genetic algorithms) is started, which reallocates the execution time, frequency, and resource priority of computing power tasks under the objective function of minimizing carbon emissions, generating a scheduling strategy that matches the clean energy curve. This strategy data is then input into the monitoring and evaluation module.

[0025] The monitoring and evaluation module outputs real-time data to the data display front end through the RESTful API interface, and generates a multi-dimensional carbon footprint dashboard using visualization tools such as D3.js or ECharts, displaying task-level carbon emissions, clean energy usage ratio, and excessive alarm records. Once it detects that carbon emissions exceed the set threshold, it adjusts the priority of the corresponding task through the task manager, delaying, reducing the frequency, or migrating it to a low-carbon node for operation, while generating a quantitative evaluation report of the adjustment for subsequent auditing and strategic planning.

[0026] The data collection correction module, supply-demand prediction module, task scheduling optimization module, and carbon emission monitoring and evaluation module work together to achieve dynamic matching between hydropower clean energy and server computing power, carbon emission control, and task optimization configuration.

[0027] 1. Data collection and clean energy supply correction First, voltage and current sensors are deployed at the output port of the hydropower station power generation equipment to collect real-time voltage and current values. The actual power generation at that moment is calculated by multiplying these two real-time data, with the unit being kilowatts. For example, if the power generation voltage is 220 volts and the current is 45 amperes, the power generation at that moment is 220 multiplied by 45, which is 9900 watts, or 9.9 kilowatts.

[0028] In the preferred scheme, the transmission loss coefficient is calculated according to the grid line impedance, voltage level, and geographical distance parameters, the clean energy transmission efficiency is derived, and the corrected clean energy supply data is obtained, which specifically includes: The total resistance of the transmission line is calculated by the formula: ; In the formula, is the total resistance of the transmission line, the unit length resistance value of the cable material at standard temperature, Geographical distance of the power transmission path from the hydropower station to the server cluster.

[0029] According to the total resistance and the average current flowing in the line, the power loss consumed in the form of heat energy on the line is calculated, and the formula is: ; In the formula, is the average current value passing through the power transmission line.

[0030] Based on the power loss and the original power generation of the hydropower station, the loss coefficient is calculated, and its complement is the transmission efficiency, and the formula is: Transmission loss coefficient: ; Clean energy transmission efficiency: ; In the formula, is the original power generation collected in real time by the hydropower station.

[0031] Apply the transmission efficiency to the original power generation data of the hydropower station to obtain the clean energy amount for the server to use after correcting the transmission loss, and the formula is: .

[0032] In order to obtain the final clean energy value available for the server, the power loss in the transmission process also needs to be considered in this embodiment; by measuring the length of the power transmission path (such as 5 kilometers) and the cable material (such as copper core cable), the resistance value per unit length (such as about 0.0175 ohms per kilometer) is queried, and combined with the average current value (such as 20 amperes), the loss power of energy on the line is calculated using the resistance law, that is, the total resistance of the line is multiplied by the square of the current to obtain the loss power. For example, the total resistance of 5 kilometers of copper wire is 0.0875 ohms, and the loss power is 0.0875 multiplied by the square of 20, which is equal to 35 watts. The available supply amount is obtained by deducting the loss power from the power generation power.

[0033] In addition, all collection devices are time calibrated by the network time protocol in this embodiment, which ensures that the data on the power consumption side and the power generation side have a unified time reference, guarantees the time sequence consistency and synchronization of the data, and provides an accurate basis for subsequent supply and demand prediction.

[0034] 2. Clean energy carbon footprint chain construction and supply trend prediction In order to ensure the traceability and tamper resistance of clean energy data, a hash algorithm is used to encrypt each energy supply record. Each record contains supply power, timestamp and hash value of the previous record, which is combined and encrypted to form a unique identification code of the current record, thereby constructing a chain structure of the data record system, ensuring that each energy supply data cannot be tampered with, forming a complete and reliable carbon footprint chain.

[0035] Based on the historical record of chain-based energy supply, a long short-term memory network model is used for prediction. The input is the sequence of clean energy supply in the past 24 hours (sampled every hour, a total of 24 data points), which learns the time trend through the neural network structure, and predicts the clean energy supply value of each hour in the next 24 hours. The neural network contains an input layer, a bidirectional recurrent structure (memory historical trend), and a fully connected output layer; and uses the Adam optimizer for parameter iteration, with the mean square error as the loss function, to ensure that the prediction result fits well with the actual historical data. The prediction value will be used in the subsequent scheduling logic.

[0036] 3. Computing power demand prediction Using historical server running records, including hourly energy consumption values (in kilowatts) and executed computing power task types, combined with task type corresponding carbon footprint data, an input feature set is constructed. Task carbon footprint data is obtained through previous measurement and energy consumption quantification, representing the average carbon emission level per unit time of a task.

[0037] A random forest model is used to predict the computing power demand in the next 24 hours. Specifically, historical energy consumption data is divided by hour, and factors such as task classification, time period, and workday flag are introduced as input variables to build a set of decision trees. Each tree is trained according to different variables, and the final prediction value is obtained through majority voting; effectively identifying the computing power demand trend in different time periods and outputting the required computing power per hour in the future, in kilowatts.

[0038] 4. Scheduling optimization of computing power tasks and energy supply After obtaining the prediction values of clean energy supply and computing power demand for the next 24 hours, the two sequences are compared hour by hour. If the predicted computing power demand for a certain hour is higher than the clean energy supply for that hour, it is judged as an "energy gap period".

[0039] At this time, the scheduling optimization algorithm is started, and the execution time, priority, and resource consumption of the computing power task are redistributed under the constraint condition. The goal is to move high-energy consumption tasks to periods with sufficient clean energy to run, in order to achieve the lowest overall carbon emissions; the process uses a particle swarm optimization algorithm to simulate multiple combinations of task scheduling strategies, and iteratively evaluates their total carbon emissions and execution delay costs to find the optimal scheduling configuration. The final output scheduling strategy includes the start time, resource consumption, and execution order of each task.

[0040] 5. Carbon footprint monitoring and carbon emission threshold setting According to the scheduling result, the actual carbon emissions generated by each task are calculated in real time, specifically: the task running power is multiplied by the carbon factor of its energy source to obtain the carbon emissions per unit time of the task. For example, if a task consumes 3 kilowatts per hour, and the energy used is hydropower (carbon factor is 10 grams of carbon per kilowatt-hour), then the carbon emissions of the task in 1 hour is 30 grams.

[0041] In this embodiment, the carbon emissions of all tasks in a day are accumulated and compared with the preset daily carbon emission threshold; the threshold is calculated according to the system monthly electricity budget, the target clean energy usage ratio, and the energy carbon factor. For example: if the monthly energy budget is 30,000 kilowatt-hours, the target usage of clean energy is 80%, and the average carbon factor is 30 grams of carbon emissions per kilowatt-hour, then the daily carbon emission allowance is 30,000 multiplied by 0.8 multiplied by 30, and finally divided by 30, which is 24,000 grams of carbon emissions per day.

[0042] If it is detected that the carbon emissions of a day are close to or exceed this threshold, an alarm mechanism is triggered immediately, and an emergency scheduling program is started to adjust the execution time or resources of the remaining tasks to ensure that the final running result is controlled within the acceptable carbon emission range.

[0043] This embodiment realizes the dynamic matching of clean energy and computing power resources, avoids energy waste while optimizing the computing power allocation strategy; significantly improves the system energy efficiency and carbon emission management capability, ensures that the server running carbon footprint is controllable and meets the carbon neutralization target. The use of tamper-proof chain improves the transparency of the energy supply process, which helps audit and compliance verification; the monitoring and evaluation mechanism enables the system to adjust task priorities in real time, strengthens the carbon emission response capability, and improves the green intelligent level of system operation.

[0044] In the preferred scheme, in the prediction scheduling module, the tamper-proof carbon footprint tracking chain data includes: According to the corrected clean energy supply amount data and the server cluster instantaneous energy consumption data, a dynamic carbon emission factor calculation model is constructed; If the corrected clean energy supply amount is greater than or equal to the instantaneous energy consumption of the server cluster, the dynamic carbon emission factor is set to the fixed carbon emission intensity of hydropower, and the specific value is 0.01 kilograms of carbon dioxide equivalent per kilowatt-hour.

[0045] If the corrected clean energy supply amount is less than the instantaneous energy consumption of the server cluster, according to the proportion of clean energy supply amount and thermal power supplement, the clean energy carbon emission coefficient and the thermal power carbon emission coefficient are used as the reference: preferably, the proportion of clean energy supply amount and the thermal power carbon emission coefficient 0.9 kilograms of carbon dioxide equivalent per kilowatt-hour are used as the weight, and the mixed dynamic carbon emission factor is calculated by weighted average; a real-time dynamic carbon emission factor sequence is generated.

[0046] The embodiment integrates a carbon emission factor modeling subsystem in the prediction scheduling module to realize real-time dynamic evaluation of the carbon footprint during server operation, specifically including three links of clean energy proportion judgment, emission factor dynamic calculation, and carbon footprint data chaining.

[0047] Clean energy supply and demand state judgment: Through the preposed Internet of Things collection interface, two core input variables are obtained in real time: 1. The corrected clean energy supply, i.e. the available energy of water and electricity after deducting power transmission loss; 2. The instantaneous energy consumption of the server cluster, i.e. the total power of all servers in the current period, with the unit of kilowatt-hour, which can be obtained by collecting the voltage and current of the server power unit and performing integral processing.

[0048] The embodiment uses conditional judgment logic to judge the relationship between supply and demand in the period: If the corrected clean energy supply is greater than or equal to the instantaneous energy consumption, i.e. the computing power task in this period is completely supported by water and electricity, then the current carbon emission factor is directly set as the fixed carbon emission intensity of water and electricity; if the supply is insufficient, i.e. the available water and electricity is less than the actual energy consumption, it indicates that there is a demand for thermal power energy supplement, and the weighted calculation process is entered to calculate the mixed carbon emission factor.

[0049] Dynamic factor sequence and carbon footprint chain construction: The above-mentioned mixed carbon emission factor will generate a group of time series, called "real-time carbon emission factor sequence", as the basic data of computing power carbon footprint, according to minutes, hours or user-defined period. The value of each time period in the sequence is packaged with the corresponding timestamp, and its digest fingerprint (for example, using SHA256 hash algorithm) is calculated, then combined with the digest of the previous record and encrypted again to generate a chain-structured data node, linked piece by piece, forming an unalterable carbon footprint chain.

[0050] The chain is stored locally and synchronized to the server backend database or blockchain platform regularly for subsequent compliance declaration, user audit or platform transparency report. Each chain data structure contains: timestamp, energy supply state flag, current carbon emission factor value, energy structure proportion (clean proportion and thermal proportion), previous chain hash and current chain hash.

[0051] System operation and deployment suggestions: In actual deployment, the data collection module and the calculation module can be distributedly deployed in the edge server to realize low-latency processing; the carbon factor model supports configuring different regional carbon emission factor reference values; supports multi-energy input scenarios such as water, electricity, photovoltaic and thermal power mix, and can also be extended to wind power, nuclear power and other clean energy structures. All calculation parameters, model logic and threshold judgments are loaded in the form of configuration files, with strong adaptability and engineering implementation.

[0052] By modeling and real-time calculation of dynamic carbon emission factors, the embodiment can accurately reflect the real carbon emission level of computing power tasks under hybrid energy supply conditions, effectively solving the distortion problem caused by the fixed setting of carbon emission factors in existing systems. In addition, real-time carbon emission evaluation of computing power tasks can be based on the current energy supply structure, providing quantitative support for subsequent scheduling decisions, and improving the accuracy and responsiveness of carbon neutral scheduling. The tamper-proof data chain structure further enhances the technical credibility of the system in compliance, audit, and traceability.

[0053] In a preferred scheme, the specific formula for calculating the hybrid carbon emission factor is: ; In the formula, is the hybrid carbon emission factor, is the amount of hydropower supply, is the amount of thermal power supplement, is the total energy consumption.

[0054] In the process of carbon footprint modeling and evaluation, the embodiment adopts a quantitative calculation formula for the hybrid carbon emission factor to further improve the accuracy of carbon emission calculation. The formula introduces the energy supply proportion of hydropower and thermal power in the overall energy consumption into the carbon emission factor model, forming an accurate and operable basis for carbon emission evaluation.

[0055] The specific calculation process is as follows: First, the following three key parameters are obtained through the energy consumption collection module: first, the "hydropower supply amount", which is the actual electric energy obtained by the server system through hydropower supply in the current period, with the unit of kilowatt-hour; second, the "thermal power supplement amount", which refers to the electric energy provided by thermal power to meet the remaining computing power load in the case of insufficient clean energy, also with the unit of kilowatt-hour; third, the "total energy consumption", which is the total electric energy supplied to the server by hydropower and thermal power, also measured in kilowatt-hour.

[0056] After obtaining the above three data, the calculation of the hybrid carbon emission factor is carried out according to the following logic: First, multiply the hydropower supply amount by the hydropower carbon emission coefficient (fixed at 0.01 kilograms of carbon dioxide equivalent per kilowatt-hour).

[0057] Then multiply the thermal power supplement amount by the thermal power carbon emission coefficient (fixed at 0.9 kilograms of carbon dioxide equivalent per kilowatt-hour); Add the results of the above two parts to get the total carbon emission in this period.

[0058] Divide the total carbon emission by the total energy consumption to get the average carbon emission intensity per unit of electric energy, which is the hybrid carbon emission factor.

[0059] For example, if the water and electricity supply is 60 kWh in an hour, and the fire power supplement is 40 kWh, the total energy consumption is 100 kWh. First, multiply 60 by 0.01 to get 0.6, then multiply 40 by 0.9 to get 36, and add the two to get 36.6. Finally, divide 36.6 by 100 to get the mixed carbon emission factor of 0.366 kg of carbon dioxide per kWh.

[0060] This calculation process will be automatically executed in each evaluation period and form a dynamic carbon emission factor sequence to provide high-precision carbon emission basic data for subsequent task scheduling, carbon neutralization evaluation and data chaining.

[0061] By introducing the explicit mixed carbon emission factor formula, the embodiment provides a standardized and quantifiable carbon evaluation mechanism, effectively improving the objectivity and repeatability of carbon emission evaluation. Compared with the traditional static carbon factor estimation method, this method can reflect the dynamic changes of the actual energy structure, accurately reveal the level of clean energy use, and provide accurate carbon emission reference under different power supply structures. This mechanism provides accurate data support for dispatch optimization, carbon emission reduction strategy formulation and compliance audit, and improves the technical advantages of the system in green computing and energy compliance.

[0062] The prediction scheduling module generates tamper-proof carbon footprint tracking chain data based on the corrected clean energy supply data, and includes obtaining the central processor utilization rate, memory occupancy rate and network bandwidth usage rate through the server performance monitoring interface according to the real-time dynamic carbon emission factor sequence, constructing a nonlinear mapping model of hardware resource consumption and power consumption using a polynomial regression algorithm, and obtaining the real-time power consumption allocation value sequence of a single computing power task according to the unique identifier of the calculation job.

[0063] In a preferred embodiment, the nonlinear mapping model of hardware resource consumption and power consumption is constructed according to the following formula: ; In the formula, is the power consumption of the i-th task, A is the central processor utilization rate, B is the memory occupancy rate, D is the network bandwidth usage rate, a is the central processor quadratic power consumption response coefficient, b is the memory linear power consumption weight coefficient, and c is the network bandwidth power consumption contribution coefficient.

[0064] To achieve accurate tracking of carbon emissions at the computing power task level, the embodiment introduces a nonlinear power consumption modeling mechanism based on server resource usage in the prediction scheduling module. By collecting real-time usage of central processors, memory and network resources, and combining regression algorithms to establish the influence of each resource on power consumption, the actual energy consumption of each computing power task is estimated, and the task-level carbon emissions are formed in combination with the carbon emission factor.

[0065] Step 1: Data collection Three types of resource indicators are obtained in real time from the server nodes through a standardized performance collection interface: (1) Central processor utilization (CPU utilization): refers to the proportion of active time of processor executing instructions in a unit of time, with a value range of 0% to 100%. This data is obtained from the performance monitoring module of the server operating system (such as / proc / stat in Linux or WMI interface in Windows).

[0066] (2) Memory occupancy rate: refers to the proportion of allocated and used part of the current physical memory in the total physical memory, with a value range of 0% to 100%. It can be provided by the memory management module of the operating system or the container resource monitoring system.

[0067] (3) Network bandwidth usage rate: represents the proportion of network interface traffic in a unit of time to the theoretical bandwidth upper limit of the interface, usually expressed as a percentage. The data can be obtained through interface state query tools (such as ifstat, ethtool or SNMP protocol).

[0068] Second step: Mapping relationship between resource utilization and power consumption A power consumption calculation model is constructed to convert the above three resource indicators into energy consumption estimates in a unit of time. This model is based on regression analysis, considering that the relationship between central processor utilization and power consumption is a nonlinear quadratic form, while the power consumption response of memory and network is linear.

[0069] The specific modeling method is as follows: First, square the central processor utilization value to obtain its nonlinear enhancement effect on power consumption; then multiply this square value by a preset coefficient a, which represents the incremental contribution of central processor utilization square to power consumption; then multiply the memory occupancy rate by coefficient b, which represents the direct contribution of each unit of memory occupancy to power consumption; similarly, multiply the network bandwidth usage rate by coefficient c, which represents the impact of each unit of network transmission bandwidth on power consumption; finally, add the three product results to obtain the estimated power consumption value of a single computing task at that time point.

[0070] Regression modeling of resource consumption and power consumption relationship, establishing the following estimation structure: First step: Square the central processor utilization (due to the intensification of central processor power consumption under high load); Second step: Set three weight coefficients a, b, and c to weigh the contribution of central processor square value, memory occupancy rate, and bandwidth usage rate to power consumption.

[0071] Third step: Add the three product results to obtain the power consumption value Pi.

[0072] Constraint: a + b + c = 1, ensuring that the three resource channels contribute a total of 100% to the overall power consumption.

[0073] Calculation Example: Known: Central Processor Utilization A = 70%; Memory Occupancy B = 50%; Network Bandwidth Usage D = 30%; Coefficients: a = 0.5, b = 0.3, c = 0.2 (satisfying a + b + c = 1), current period carbon emission factor is 0.35 kg CO2 / kWh.

[0074] The steps are as follows: Central Processor Utilization Square: 70 × 70 = 4900; Central Processor Power Consumption Part: 4900 × 0.5 = 2450 watts; Memory Power Consumption Part: 50 × 0.3 = 15 watts; Network Power Consumption Part: 30 × 0.2 = 6 watts; Total Power Consumption: 2450 + 15 + 6 = 2471 watts; Convert to kilowatts: 2471 ÷ 1000 = 2.471 kilowatts; Convert to kilowatt-hours: 2.471 × (1 ÷ 60) = 0.04118 kilowatt-hours; Carbon Emission Calculation: 0.04118 × 0.35 = 0.01441 kg CO2.

[0075] That is, the carbon emissions of this task in this one-minute period are about 14.41 grams of carbon dioxide equivalent.

[0076] Task Identification and Power Consumption Attribution: In order to bind resource usage data with specific computing power tasks, a unique task identifier is used for association. When each task is allocated by the scheduling system, the system generates a unique identifier for it, such as UUID, container ID or process number, and matches it with the operating system resource monitoring system through scheduling records. According to the identifier, the resource occupation of the task in each time period is identified, and the power consumption of the task in the period is calculated using the aforementioned model.

[0077] This process is continuously executed by minute or finer granularity, forming a "power consumption time series" corresponding to each task, that is, a set of power consumption estimates of the task in each period from start to end of the whole process.

[0078] Fusion of Power Consumption Sequence and Carbon Emission Factor: The power consumption time series is matched with the carbon emission factor time series item by item by time. For example, if the power consumption of a task in a time period is 36.76 watts, and the carbon emission factor of the time period is 0.3 kilograms of carbon dioxide per kilowatt-hour, then 36.76 watts is converted to kilowatt-hours (i.e. 0.03676 kilowatt-hours), and multiplied by 0.3, i.e. the task generates about 0.011 kilograms of carbon dioxide equivalent in the period.

[0079] The final whole life cycle carbon emission data of the task from start to completion can be used in scenarios such as on-chain, monitoring display, emission attribution and carbon audit.

[0080] The embodiment breaks through the defect of rough allocation of traditional node overall energy consumption by constructing a task-level power consumption modeling mechanism based on polynomial regression, and realizes accurate calculation of carbon emissions of a single computing power task. At the same time, the model has good adaptability and real-time performance, can dynamically adjust power consumption estimation with the change of server resource usage state, improve the real reflection ability of carbon emission data, and is beneficial to carbon cost control, green scheduling decision and user carbon billing settlement at the task level, does not depend on additional hardware modification, has high implementability and platform compatibility.

[0081] The prediction and scheduling module generates tamper-proof carbon footprint tracking chain data based on the corrected clean energy supply data, which includes calculating the instantaneous carbon emission of the computing power task level according to the real-time power consumption distribution value sequence of the single computing power task and the dynamic carbon emission factor sequence at the corresponding moment, calculating the total carbon emission in the task execution period by time integration algorithm, and obtaining the carbon footprint quantization data of the computing power task.

[0082] In a preferred scheme, the specific formula for calculating the instantaneous carbon emission of the computing power task level is: ; In the formula, E is the instantaneous carbon emission of the i-th task, is the task power consumption, and F is the carbon emission factor at time t.

[0083] In order to realize the quantitative evaluation of carbon emissions of each computing power task from start to end, the embodiment further introduces a task-level carbon emission calculation mechanism in the prediction scheduling module. Based on the real-time power consumption allocation value sequence and the corresponding time carbon emission factor sequence during the task running, the mechanism establishes a ternary correspondence relationship of "power consumption-carbon factor-carbon emission", combines product operation and time integration, and accurately calculates the carbon emission trajectory of the task. The input data includes two core parts: one is the task power consumption sequence obtained by resource usage rate modeling, with the unit of kilowatt-hour, representing the actual power consumption per minute; the second is the dynamic carbon emission factor sequence calculated based on the proportion of hydropower and thermal power, with the unit of kilogram of carbon dioxide emission intensity per kilowatt-hour. Each task is assigned a unique identifier by the system when scheduling, which is used for cross-module data association and task life cycle identification.

[0084] In the instantaneous carbon emission calculation stage, the actual power consumption value of a task at the current time and the carbon emission factor at the same time are obtained according to a fixed period (such as one minute), and multiplication operation is performed to multiply the two to obtain the instantaneous carbon emission at that time. Here, "task power consumption" reflects the power consumed by the task in that period (unit: kilowatt-hour), and "carbon emission factor" reflects the carbon dioxide emission per unit of power generated in the current power structure (unit: kilogram of carbon dioxide / kilowatt-hour). The product of the two is the carbon emission value of the task at that time point. The calculation is performed at a frequency of minutes or seconds, and a real-time updated "instantaneous carbon emission sequence" is maintained for each task, ensuring high time resolution emission recording capability.

[0085] When the task is completed, the instantaneous carbon emission value during the entire life cycle of the task is automatically time-integrated, and the emission value of each time slice is accumulated in a discretized manner to form the total carbon emission of the task. The integration method is a typical cumulative summation, that is, the "instantaneous carbon emission" at all time points is added to obtain the total carbon footprint during the completion of the task. The calculation process does not involve complex calculus, but uses fixed-period quantitative addition logic, which is convenient for fast calculation and low-power operation. The embodiment also supports visualization of the integral trajectory to analyze the carbon emission trend during the task running and identify high-emission segments.

[0086] For example, if a task runs for 5 minutes, its power consumption for each period is 0.042, 0.045, 0.047, 0.043 and 0.040 kilowatt-hours, and the corresponding carbon emission factors are 0.30, 0.28, 0.33, 0.36 and 0.35 kilograms of carbon dioxide per kilowatt-hour, then the instantaneous carbon emission values are 0.0126, 0.0126, 0.01551, 0.01548 and 0.014 kilograms of carbon dioxide, respectively. Adding these five values together gives the total carbon emissions of the task as 0.07019 kilograms, or 70.19 grams of carbon dioxide. Such products are continuously calculated every minute of the task's execution, and all instantaneous emission values are summed after the task terminates to generate the complete task carbon footprint.

[0087] The above carbon emission data will then be written into the task carbon footprint chain, which uses a hash encryption mechanism to ensure data cannot be tampered with and is traceable. Each task's emission record includes: task identification, power consumption value and carbon factor for each period, instantaneous carbon emissions, and total emissions. The carbon emission report of the task can be output through the API interface, including detailed tables, cumulative charts and block structure summaries, which can be used for computing power scheduling evaluation, user carbon emission settlement, compliance audit or public disclosure. The entire mechanism is based on standard arithmetic logic, with real data sources and clear structure, and has the ability to be quickly deployed and independently run in actual systems.

[0088] The predicted scheduling module generates tamper-proof carbon footprint tracking chain data by correcting the clean energy supply data, which also includes using blockchain distributed ledger technology to record the hash values of the power generation data of the hydropower station, the instantaneous energy consumption data of the server and the carbon emissions of the computing power task, and verifying the data integrity through a smart contract; based on the timestamp and digital signature, the on-chain hash value is mapped with the off-chain original data to generate tamper-proof carbon footprint tracking chain data.

[0089] To ensure the authenticity, integrity and tamper-proof nature of carbon emission evaluation data, the present embodiment introduces a blockchain distributed ledger structure in the predicted scheduling module to build a carbon footprint tracking chain. This chain realizes the bidirectional mapping of on-chain records and off-chain original data by encrypting and summarizing key running data, combining timestamps, task identifiers and digital signatures, ensuring that the carbon emission data of computing power tasks during execution is verifiable and legally binding.

[0090] In actual operation, first, the power generation data in unit time is obtained from the hydropower station control system, the energy consumption value at the corresponding time point is obtained from the server group, and the carbon emission value corresponding to the task in the time period is extracted from the task power consumption model. The three types of data together constitute the "carbon footprint segment data" of the task at that time. The three types of data are spliced in content, and a standard timestamp for data generation is introduced to form a whole data block. Then, a hash algorithm is used to generate a unique data digest value for the data block, that is, the data fingerprint at the time point. The digest is a fixed-length string that cannot be reverse-engineered to obtain the original content, but can uniquely represent the original data content.

[0091] Subsequently, the digital signature key of the task scheduling node or the data source node is called to encrypt and sign the above data digest, task identification and timestamp. The signature is used to prove that the data is generated by a specific identity and prevent third parties from forging or impersonating. The signature result will be used as auxiliary verification information together with the digest value to form a chain data record unit.

[0092] All task carbon footprint digest records are organized in a chain structure in chronological order. Each record not only contains the data digest value, timestamp and signature of the current period, but also saves the digest reference of the previous record, so that the entire chain has the characteristics of non-rollback and non-breakage. If any intermediate data is tampered with, it will cause all subsequent data digest verification to fail, thereby achieving integrity protection. This structure meets the needs of continuity and logical consistency of task-level carbon emission data in multiple time periods.

[0093] In the data query and comparison process, the corresponding original data file is quickly retrieved in the off-chain storage system through the on-chain digest value. For this purpose, an index mapping table is established locally, and each table entry binds the hash digest with the original data segment content, generation time and task identification. When it is necessary to verify whether the on-chain data at a certain time point is real and effective, a new digest value is calculated by re-encrypting the off-chain original data content, and the new digest value is compared with the on-chain record; if they are consistent, it means that the data has not been tampered with; if they are not consistent, it is determined that the data is distorted.

[0094] The above mechanism uses standard cryptography methods and combines the authority system of actual computing nodes to support data evidence, behavior audit, task authentication and other functions. By recording all carbon emission data during task running in a chain form in the block ledger, the present embodiment effectively improves the data credibility and automatic verification capability of the clean energy computing system in green scheduling, compliance disclosure and carbon emission reduction assessment; and can be connected with the national carbon management platform or third-party authentication system to provide standardized and traceable carbon emission underlying data support for the computing power platform.

[0095] By storing the hydropower station power generation data, server instantaneous energy consumption data and carbon emissions of computing power tasks in the form of encrypted hash, and combining the timestamp and digital signature mechanism, the tamper-proof, secure and verifiable and full-process traceable of task carbon footprint data are realized, which significantly improves the credibility and regulatory compliance ability of the system in green computing power scheduling. Compared with the traditional carbon emission recording method, this mechanism can realize the automation of data generation, storage and verification, without relying on centralized third-party verification institutions, reducing the audit cost and human intervention risk; at the same time, this structure supports multi-node synchronous recording and verification, has high fault tolerance and anti-tamper ability, effectively guarantees the transparent and controllable emission behavior of key tasks under the driving of clean energy, and provides key support for building credible, efficient and sustainable low-carbon computing power infrastructure.

[0096] In the preferred scheme, the calculation formula of the central processor quadratic power consumption response coefficient a is: ; In the formula, a is the central processor quadratic power consumption response coefficient, T is the duration length of collecting data and evaluating system state, S is the number of arithmetic logic units participating in calculation during task execution, R is the voltage value maintained by the server central processor core during task execution, G is the average pressure of instruction parallel computing resources during the execution of the current task, and t is the time.

[0097] To realize dynamic response modeling of the server central processor energy consumption behavior, a calculation method of the central processor quadratic power consumption response coefficient is proposed, which realizes the fine quantization of the relationship between the server processing capacity load and energy consumption by introducing the algorithm complexity, computing resource consumption rate and system running pressure quantitative index in the task execution process. In this embodiment, within a set time interval, the number of arithmetic logic units participating in the calculation, the voltage value maintained by the central processor core, the task computing intensity and the system scheduling load and other parameters are collected in real time. Among them, according to the participation degree and algorithm operation complexity of the task at each moment, the response ratio in unit time is calculated, and a dynamic calculation density factor expressed in logarithmic form is introduced to construct the central processor behavior response curve. Subsequently, by integral method, the dynamic behavior is accumulated in the whole data collection period, and is squared to improve the sensitivity of nonlinear change. The calculation result is the square of the integral expression as the numerator, and the square of the sampling period and the current system load intensity as the denominator, forming the final evaluation value of the central processor response coefficient. The result of this coefficient not only considers the cumulative effect of task behavior in the time dimension, but also integrates the coupling relationship between system hardware stability, resource scheduling strategy and task computing characteristics, so as to more truly reflect the energy efficiency performance of the central processor under different intensity tasks.

[0098] The embodiment solves the problems of traditional server energy consumption model, such as insensitive to instantaneous load response, ignoring the difference of task algorithm structure, and unable to depict multi-thread compression characteristics, by a central processing unit square power consumption response coefficient model based on time integral and arithmetic density weighting, and realizes fine evaluation of the computing power system under dynamic energy consumption distribution; the response coefficient is used for real-time participation in power prediction in the system scheduling process, and the accuracy of computing power resource allocation is improved; meanwhile, the response coefficient provides quantifiable computing layer feature factors for carbon emission modeling in the prediction scheduling module, and improves the reliability and adaptability of the task carbon footprint prediction model.

[0099] In the preferred scheme, the calculation formula of the memory linear power consumption weight coefficient b is: ; In the formula, b is the memory linear power consumption weight coefficient, U is the average memory usage of the current task, is the page fault rate change, V is the memory fragmentation rate, and Y is the hot page proportion. c = 1-a-b; In the formula, c is the network bandwidth power consumption contribution coefficient.

[0100] To improve the quantification ability of the influence of task memory behavior on system energy efficiency, the embodiment proposes a calculation method of a memory linear power consumption weight coefficient for representing task memory access intensity. The coefficient considers core memory characteristic indexes such as the average memory usage during the execution of the current task, the fluctuation amplitude of the page fault rate, the fragmentation rate, and the hot page proportion, and realizes the normalized modeling of the memory resource usage behavior by constructing a function expression mode of the combination of the logarithmic growth term of the numerator and the stable term of the denominator. In specific implementation, first, the average memory usage in the task execution period is counted, and the change degree of the page fault rate is extracted in combination with the task page scheduling history, reflecting the memory access intensity. Then, the memory fragmentation rate in the current task execution environment is calculated, that is, the proportion of the assignable but discontinuous memory, and the hot page proportion, that is, the proportion of the frequently accessed pages in the total memory pages, which are used to depict the system memory arrangement efficiency and access focus degree. Finally, the above factors are respectively substituted into the preset function structure, the product of the usage and the change of the page fault rate is logarithmically transformed to enhance the response sensitivity of the change to the power consumption, and the combination of the square of the fragmentation rate and the hot page function forms the denominator, which is used to balance the overall stability of the system. The function result is the memory power consumption weight factor of the current task in the current execution period, which is used for subsequent integration into the total power consumption evaluation model.

[0101] The embodiment improves the expression capability of the task memory access mode in the power consumption model by considering the weight function of the memory usage behavior fluctuation, the spatial fragmentation and the access hot spot distribution characteristics, solves the problems of the memory related factor response lag and the lack of reflection on the hot page and fragmentation in the existing model, and constructs the weight coefficient which not only has the balance between real-time and stability, but also can automatically suppress the interference of abnormal fluctuation on the power consumption evaluation through the denominator structure, so that the credible power consumption estimation result can still be output when the task frequently switches or abnormally runs. In addition, the model structure has strong compatibility, can be embedded in various memory management systems and different scheduling frameworks, has good expansibility and engineering application potential.

[0102] The embodiment collects the power generation of the hydropower station and the server energy consumption data, establishes a dynamic carbon emission factor calculation model, combines the hardware resource consumption index of the server, and realizes accurate calculation of carbon emission at the computing power task level. The whole process data is recorded by using the blockchain technology, so as to ensure the credibility and transparency of carbon footprint tracking; the trend of clean energy supply is predicted by using time series analysis, the computing power task scheduling is optimized, and the balance between energy supply and demand is realized. Finally, the carbon emission is monitored in real time through the visual interface, and the achievement of the carbon neutralization target is evaluated, so as to realize accurate quantification and optimal scheduling of clean energy use of the data center, and provide technical support for low-carbon operation of large-scale computing power service.

[0103] The above embodiment is only a preferred technical solution of the present application, and should not be regarded as a limitation of the present application. The protection scope of the present application should be based on the technical solution recorded in the claims, including the equivalent replacement solution of the technical features recorded in the claims. That is, the equivalent replacement improvement within this range is also within the protection scope of the present application.

Claims

1. A hydropower station integrated computing power intelligent server system, characterized in that, include: The data acquisition and correction module is used to collect real-time power generation data of hydropower stations and instantaneous energy consumption data of server clusters through IoT sensors, and to achieve data timestamp synchronization using the Network Time Protocol to establish a time sequence correspondence between the power generation side and the power consumption side. Based on the power grid line impedance, voltage level and geographical distance parameters, the module calculates the transmission loss coefficient, derives the clean energy transmission efficiency, and obtains the corrected clean energy supply data. The prediction and scheduling module is used to generate immutable carbon footprint tracking chain data from the corrected clean energy supply data; based on the immutable carbon footprint tracking chain data, it uses a long short-term memory network algorithm to predict the clean energy supply fluctuation trend sequence within a preset time period; based on the server's historical energy consumption data and the carbon footprint quantification data of computing power tasks, it trains a random forest model to predict the computing power demand trend sequence; if the predicted clean energy supply fluctuation trend sequence is lower than the computing power demand trend sequence, it optimizes task scheduling according to the carbon emission minimization target and generates dynamic balance configuration data between computing power tasks and clean energy supply. The monitoring and evaluation module is used to output the carbon emissions and clean energy usage ratio at the computing power task level in real time through the application programming interface based on the dynamic balance configuration data of the computing power tasks and clean energy supply. It also uses data visualization technology to generate carbon footprint monitoring dashboard data including line charts and bar charts. If the carbon emissions exceed the preset daily carbon emission threshold, an alarm is triggered and the priority of computing power tasks is adjusted to generate quantitative evaluation data of the carbon neutrality target of computing power services.

2. The hydropower station integrated computing power intelligent server system according to claim 1, characterized in that, Based on parameters such as power grid line impedance, voltage level, and geographical distance, the transmission loss coefficient is calculated, the clean energy transmission efficiency is derived, and the corrected clean energy supply data is obtained, specifically including: The formula for calculating the total resistance of a transmission line is: ； In the formula, The total resistance of the transmission line. The resistance per unit length of cable material at standard temperature. The geographical distance of the power transmission path from the hydropower station to the server cluster; Based on the total resistance and the average current flowing through the line, the power loss consumed as heat energy in the line is calculated using the following formula: ； In the formula, This represents the average current flowing through the transmission line. Based on power loss and the original power generation capacity of the hydropower station, the loss coefficient is calculated, and its complement is the transmission efficiency. The formula is as follows: Transmission loss factor: ; Clean energy transmission efficiency: ; In the formula, This refers to the raw power generation data collected in real time by the hydropower station. Applying transmission efficiency to the raw power generation data of a hydropower station yields the amount of clean energy after correcting for transmission losses, using the following formula: 。 3. The hydropower station integrated computing power intelligent server system according to claim 1, characterized in that, The predictive scheduling module generates tamper-proof carbon footprint tracking chain data, including: Based on the revised clean energy supply data and the instantaneous energy consumption data of the server cluster, a dynamic carbon emission factor calculation model is constructed. If the corrected clean energy supply is greater than or equal to the instantaneous energy consumption of the server cluster, the dynamic carbon emission factor is set to the preset clean energy carbon emission coefficient. If the corrected clean energy supply is less than the instantaneous energy consumption of the server cluster, then based on the ratio of clean energy supply to thermal power supplementation, and using the preset clean energy carbon emission coefficient and thermal power carbon emission coefficient as a benchmark, a mixed dynamic carbon emission factor is calculated by weighted average; and a real-time dynamic carbon emission factor sequence is generated.

4. The hydropower station integrated computing power intelligent server system according to claim 3, characterized in that, The specific formula for calculating the mixed carbon emission factor is as follows: ； In the formula, As a mixed carbon emission factor, For hydropower supply, To supplement thermal power, Total energy consumption.

5. The hydropower station integrated computing power intelligent server system according to claim 1, characterized in that, In the prediction and scheduling module, generating tamper-proof carbon footprint tracking chain data also includes: obtaining CPU utilization, memory usage, and network bandwidth usage through the server performance monitoring interface based on the real-time dynamic carbon emission factor sequence; constructing a nonlinear mapping model of hardware resource consumption and power consumption using a multinomial regression algorithm; associating computing jobs with task unique identifiers; and obtaining a real-time power allocation value sequence for a single computing task.

6. The hydropower station integrated computing power intelligent server system according to claim 5, characterized in that, The specific formula for constructing the nonlinear mapping model between hardware resource consumption and power consumption is as follows: ； In the formula, Let A be the power consumption of the i-th task, B be the CPU utilization, D be the network bandwidth utilization, a be the CPU squared power consumption response coefficient, b be the memory linear power consumption weighting coefficient, and c be the network bandwidth power consumption contribution coefficient.

7. The hydropower station integrated computing power intelligent server system according to claim 6, characterized in that, The formula for calculating the square power consumption response coefficient 'a' of the central processing unit is as follows: ； In the formula, a is the square power consumption response coefficient of the central processing unit, T is the duration of data acquisition and system state evaluation, S is the number of arithmetic logic units involved in the calculation during task execution, R is the voltage value maintained by the server's central processing unit core during task execution, G is the average pressure on instruction parallel computing resources during the current task execution, and t is the time.

8. The hydropower station integrated computing power intelligent server system according to claim 6, characterized in that, The formula for calculating the memory linear power consumption weighting coefficient b is: ； In the formula, b is the linear power consumption weighting coefficient for memory, and U is the average memory usage of the current task. V represents the change in page fault rate, V represents the memory fragmentation rate, and Y represents the proportion of hot pages. c = 1 - a - b; In the formula, c is the network bandwidth power consumption contribution coefficient.

9. The hydropower station integrated computing power intelligent server system according to claim 8, characterized in that, The specific formula for calculating the instantaneous carbon emissions at the computing power task level is as follows: ； In the formula, E represents the instantaneous carbon emissions of the i-th task, and P... i Let F be the power consumption of the task, and F be the carbon emission factor at time t.

10. The hydropower station integrated computing power intelligent server system according to claim 1, characterized in that, The predictive scheduling module, which generates tamper-proof carbon footprint tracking chain data, also includes: Based on the real-time power consumption allocation sequence of a single computing task and the corresponding dynamic carbon emission factor sequence, the instantaneous carbon emissions at the computing task level are calculated by multiplication, and the total carbon emissions within the task execution cycle are accumulated by time integration algorithm to obtain the quantitative data of the computing task's carbon footprint. Based on the carbon footprint quantification data of computing power tasks, the hash values ​​of hydropower station power generation data, server instantaneous energy consumption data, and carbon emissions of computing power tasks are recorded using blockchain distributed ledger technology. The integrity of the data is verified through smart contracts. Based on timestamps and digital signatures, the on-chain hash values ​​are mapped to the off-chain original data to generate an immutable carbon footprint tracking chain data.

Images