Wind-solar-storage integrated edge data center power supply system

Abstract

The invention relates to the technical field of electric energy management, in particular to a wind-solar-storage integrated edge data center power supply system, which comprises an energy power acquisition module, an overlapping degree calculation module, an energy storage scheduling module, a bus routing module and a thermoelectric control module. According to the method, the instantaneous power of wind energy and light energy is collected in real time and subjected to dynamic normalization processing, a filtering algorithm is combined, noise is accurately eliminated, a high-adaptation data center power sequence is generated, and fine identification of load fluctuation and dynamic adjustment of the power supply guarantee rate are achieved through vector analysis and window marking. When the guarantee rate is reduced, an energy storage response compensation scheme is intelligently optimized, instant self-matching supplement of electric energy supply is achieved, an energy supply structure is adjusted according to real-time parameters of a line, the voltage change amplitude is accurately monitored to actively trigger thermal management measures, and the linkage effect of energy consumption control and electric power fluctuation reduction is formed. The matching performance of the power efficiency and the load working condition is remarkably improved, and the risk hidden danger caused by instability of renewable energy sources is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electric energy management, in particular to a wind-solar-storage integrated edge data center power supply system. BACKGROUND

[0002] The technical field of electric energy management involves effective monitoring, dispatching, power distribution, energy storage, and optimization management of power systems, aiming to ensure efficient use of power resources and power supply safety. This field includes key technologies such as operation and maintenance of power systems, power load forecasting, battery energy storage technology, integration and management of renewable energy, and power dispatching optimization. The core task of electric energy management is to make the production, transmission, distribution, and use of electricity more intelligent, more reliable, and lower in cost through the comprehensive use of various technical means, and to flexibly respond to the volatility and uncertainty of differentiated energy sources.

[0003] Among them, the traditional wind-solar-storage integrated edge data center power supply system refers to a power supply mode based on wind energy, solar energy, and energy storage system. By utilizing renewable energy sources such as wind energy and solar energy, combining energy storage devices and coordinated control of the power grid, power is provided for data centers. The traditional wind-solar-storage integrated power supply system combines wind and solar power generation modules with battery energy storage systems, stores power through an intelligent battery management system, and converts the stored power into alternating current that meets the needs of data centers through an inverter. This system relies on centralized power dispatching and faces energy fluctuation problems, and it is difficult to provide stable power supply under the rapid response requirements of edge computing.

[0004] The traditional wind-solar-storage integrated system relies on centralized dispatching and fixed energy storage management, making it difficult to achieve detailed perception and response to instantaneous power changes. When new energy fluctuations and load demands are inconsistent, compensation lags, and the timing of energy storage intervention is not precise, it is easy to cause excessive voltage deviation fluctuations and local overheating problems when encountering frequent load fluctuation conditions, resulting in difficulty in fully guaranteeing power supply stability and equipment safety, as well as a decrease in energy utilization efficiency. Data centers are easily constrained in terms of continuous power supply and energy saving and consumption reduction. SUMMARY

[0005] In order to solve the technical problems of the traditional wind-solar-storage integrated system relying on centralized dispatching and fixed energy storage management, making it difficult to achieve detailed perception and response to instantaneous power changes, compensation lagging when new energy fluctuations and load demands are inconsistent, and the timing of energy storage intervention being imprecise, it is easy to cause excessive voltage deviation fluctuations and local overheating problems when encountering frequent load fluctuation conditions, resulting in difficulty in fully guaranteeing power supply stability and equipment safety, as well as a decrease in energy utilization efficiency. Data centers are easily constrained in terms of continuous power supply and energy saving and consumption reduction, the present application provides a wind-solar-storage integrated edge data center power supply system. The technical solution is as follows: In one aspect, a wind-solar-storage integrated edge data center power supply system is provided, which comprises: An energy power collection module acquires wind energy and light energy node instantaneous power, performs normalization processing, filters out noise by using a Kalman filtering algorithm, generates a data center matching power sequence, and transmits the data center matching power sequence to an overlap calculation module; The overlap calculation module receives the data center matching power sequence, acquires a data center load reference, constructs a node power vector, compares the size relationship between the vector components and the reference, marks the time window exceeding the reference, calculates the window length proportion, obtains a data center power supply guarantee rate, and transmits the data center power supply guarantee rate to an energy storage scheduling module; The energy storage scheduling module receives the data center power supply guarantee rate, detects the energy storage state of charge and power, monitors the load demand, judges the relationship between the guarantee rate and the safety threshold, calculates a compensation scheme by using a particle swarm optimization algorithm when the guarantee rate is lower than the threshold, generates an energy storage compensation instruction, and transmits the energy storage compensation instruction to a bus routing module; The bus routing module calls the energy storage compensation instruction, acquires bus voltage, current and power factor, calls the instruction to set the energy storage output target value, adjusts the proportion of photovoltaic and commercial power supply, monitors the bus voltage fluctuation amplitude and the rated voltage deviation, judges whether it is out of the safety range, generates a thermal management instruction, and transmits the thermal management instruction to a thermoelectric control module.

[0006] As a further scheme of the present application, the data center matching power sequence comprises a load matching index, a power supply response time and an active power fluctuation rate, the data center power supply guarantee rate comprises a time availability, a power redundancy coefficient and a load tracking capability, the energy storage compensation instruction comprises a power gap value, a state of charge threshold and a scheduling priority weight, and the thermal management instruction comprises a cooling system adjustment amplitude, an overheating correction coefficient and a waste heat recovery efficiency.

[0007] As a further scheme of the present application, the energy power collection module comprises: A node power collection submodule collects wind energy and light energy node instantaneous power values and collection time stamps, unifies a data format table structure, synchronously filters power value combinations according to time stamps, fits a time sequence data vector set, and generates node power time sequence values; A normalization conversion submodule calls the node power time sequence values, extracts node power level values, calculates a proportion according to the difference between the current value and the extreme value, maps to a standard interval according to a normalization function, and generates a normalized power sequence; A Kalman filtering submodule calls the normalized power sequence, sets an initial state estimation value and an error covariance, constructs a state and an observation matrix, calculates a correction gain and corrects the power value, updates the covariance and the state vector, and generates a data center matching power sequence.

[0008] As a further scheme of the present application, the node power time sequence value is a sequence composed of multiple wind energy and light energy node instantaneous power data arranged in time stamp order. The normalization function is a mathematical function for numerical conversion according to the proportion of the position of the node power value in the target mapping interval.

[0009] As a further scheme of the present application, the overlap degree calculation module comprises: The power vector construction submodule acquires the data center matching power sequence data, extracts node time sequence power values according to node numbers, constructs node power vectors according to time sequence combination of power values, calls node number power data aggregation processing, and generates node power vector groups. The reference value determination submodule, based on the node power vector groups, combines data center load reference value sequences, aligns matching power components and reference values according to time, compares numerical values and marks time positions greater than the reference values, merges continuous time points into overrun sections, and obtains a node overrun time section set. The guarantee rate measurement submodule calculates the node overrun total duration according to the node overrun time section set, calculates the ratio according to the length of the complete time sequence and converts it into the non-overrun rate, processes the node non-overrun rate data in a unified format, and acquires the data center power supply guarantee rate.

[0010] As a further scheme of the present application, the energy storage scheduling module comprises: The data monitoring submodule acquires the data center power supply guarantee rate, monitors the state of charge value and the charge and discharge power value of the energy storage unit, collects the load demand value and verifies its timeliness and continuity, and obtains the real-time power supply guarantee state. The guarantee rate judgment submodule compares the real-time power supply guarantee state with the power supply safety threshold, calls the load demand value and the state of charge value, calculates the deviation between the current load gap power and the remaining output power of the energy storage, judges the power supply safety interval state, and obtains the guarantee rate safety judgment value. The current load gap power refers to the remaining power difference after subtracting the sum of the current output of the main power supply and the energy storage from the load demand power of the data center at a certain moment, which is used to reflect whether there is a power supply shortage. The remaining output power of the energy storage refers to the maximum compensation power that the energy storage system can continue to provide theoretically under the current state of charge and discharge upper limit constraint, which is used to measure its support ability. The guarantee rate safety judgment value refers to the safety state flag obtained by comparing the current power supply guarantee rate with the set safety threshold, which is used to judge whether it is in the safe power supply interval. The particle swarm optimization calculation submodule calls the safety judgment value and the state of charge value, sets particle coding dimensions including load gap, discharge power and scheduling period, iteratively updates the scheme based on the particle swarm optimization algorithm, generates the energy storage compensation instruction in combination with the scheduling power interval and the load demand curve.

[0011] As a further scheme of the present application, the bus routing module comprises: The energy storage instruction analysis submodule acquires the energy storage compensation instruction, collects HVDC bus voltage values, current values and power factor values through a data communication interface, performs difference calculation on electrical target values in the instruction and current electrical data, judges the matching degree of the power supply state and the energy storage instruction, merges deviation terms to obtain an electrical deviation response quantity; The power setting control submodule matches the energy storage instruction power upper limit and the current current value based on the electrical deviation response quantity, calculates the power difference value and corrects the energy storage output power target value according to the output data of the photovoltaic MPPT module and the AC / DC module, adjusts the photovoltaic and mains power supply ratio, and generates a dual-source power matching value; The voltage deviation judgment submodule extracts bus voltage fluctuation values and sets voltage deviation reference values according to the dual-source power matching value, judges the temperature control trigger condition in combination with the power distribution state, calculates the voltage abnormal section and evaluates the temperature control response demand, and generates a thermal management instruction.

[0012] As a further scheme of the present application, the electrical deviation response quantity is a difference calculation result between electrical target values in the energy storage instruction and current bus voltage, current and power factor data; The dual-source power matching value is a power distribution value calculated according to the electrical deviation response quantity, the energy storage instruction power upper limit and the output power of the photovoltaic and mains; The energy storage output power target value is an energy storage device output power setting value determined in combination with the energy storage instruction power upper limit, the electrical deviation response quantity and the current current data.

[0013] As a further scheme of the present application, the thermoelectric control module receives the thermal management instruction, detects the power module operating temperature, compares the temperature and the over-temperature protection threshold value relationship, reduces the module output participation coefficient when the temperature exceeds the threshold value, starts waste heat recovery and adjusts the cooling output level, and outputs a data center thermoelectric balance control signal; The data center thermoelectric balance control signal comprises a thermal coupling coefficient, a module load factor and a temperature stability index.

[0014] As a further scheme of the present application, the thermoelectric control module comprises: A temperature monitoring submodule calls the thermal management instruction, extracts operating temperatures of photovoltaic MPPT, energy storage DC / DC and mains AC / DC modules, collects temperature sequence values of each module, synchronously organizes data according to timestamps, eliminates missing items and outliers, performs interpolation processing on the data, and generates a temperature dataset; A temperature comparison and adjustment submodule extracts a current temperature from the temperature dataset, compares the current temperature with a module overheating threshold, records a temperature offset amplitude, calculates a dynamic adjustment factor, updates a power proportion in combination with a power output value, and generates an adjusted power participation coefficient. The module overheating threshold refers to a highest temperature upper limit allowed by the module under safe operating conditions, and exceeding the threshold is regarded as an overheating state. The adjusted power participation coefficient is a coefficient value obtained by dynamically correcting a module power output proportion after considering temperature offset and thermal management requirements. A heat recovery scheduling submodule judges whether to start a heat recovery path according to the adjusted power participation coefficient, collects heat recovery power and cooling parameters, calculates a heat release rate and a cooling level in combination with module temperature and cooling load, and outputs a data center thermal power balance control signal.

[0015] The technical scheme provided by the embodiment of the present application has at least the following beneficial effects: By collecting and dynamically normalizing the instantaneous power of wind and light energy in real time, combining a filtering algorithm, accurately eliminating noise and generating a highly adaptive data center power sequence, using vector analysis and window marking to achieve fine identification of load fluctuations, dynamically adjusting the power supply guarantee rate, intelligently optimizing the energy storage response compensation scheme when the guarantee rate decreases, realizing instant self-matching compensation of power supply, adjusting the energy supply structure according to real-time line parameters, and accurately monitoring the voltage change amplitude to actively trigger thermal management measures, the linkage effect of energy consumption control and power fluctuation reduction is formed, the power efficiency and load working condition matching are significantly improved, and the risk hidden danger caused by unstable renewable energy is reduced. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical scheme in the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0017] Figure 1 is a schematic diagram of the system of the present application; Figure 2 is a schematic diagram of the system framework of the present application; Figure 3 is a flowchart of the energy power collection module in the present application; Figure 4 Flow chart of the overlap calculation module in the application; Figure 5 Flow chart of the energy storage scheduling module in the application; Figure 6 Flow chart of the bus routing module in the application; Figure 7 Flow chart of the thermal power control module in the application. DETAILED DESCRIPTION

[0018] The technical solutions in the application will be described below with reference to the drawings.

[0019] In the embodiments of the application, the words such as “example”, “for example” and the like are used to represent an example, illustration or description. Any embodiment or design scheme described as “example” in the application should not be interpreted as more preferred or more advantageous than other embodiments or design schemes. Rather, the word “example” is intended to present a concept in a specific manner. In addition, in the embodiments of the application, the meaning expressed by “and / or” can be both, or can be one of the two.

[0020] In the embodiments of the application, “image” and “picture” can be used interchangeably at times, and it should be pointed out that the meanings expressed are consistent when the distinction is not emphasized. “Of”, “corresponding” and “corresponding” can be used interchangeably at times, and it should be pointed out that the meanings expressed are consistent when the distinction is not emphasized.

[0021] In the embodiments of the application, sometimes the subscript such as W1 can be written in the form of non-subscript such as W1, and the meanings expressed are consistent when the distinction is not emphasized.

[0022] To make the technical problems, technical solutions and advantages to be solved by the application more clear, specific embodiments will be described in detail below with reference to the drawings.

[0023] The embodiments of the application provide a wind-solar-storage integrated edge data center power supply system, as shown in the wind-solar-storage integrated edge data center power supply system schematic diagram, the system comprises: Figures 1-2 An energy power acquisition module acquires wind energy and light energy node instantaneous power, performs normalization processing, filters out noise by using a Kalman filtering algorithm, generates a data center matching power sequence, and transmits the data center matching power sequence to an overlap calculation module. ​The overlap calculation module receives the matching power sequence of the data center, obtains the data center load benchmark, constructs the node power vector, compares the vector components with the benchmark size relationship, marks the time window exceeding the benchmark, calculates the window duration percentage, obtains the data center power supply guarantee rate, and transmits it to the energy storage scheduling module. The energy storage scheduling module receives the power supply guarantee rate of the data center, detects the energy storage charge status and power, monitors load demand, judges the relationship between the guarantee rate and the safety threshold, and uses the particle swarm optimization algorithm to calculate the compensation scheme when the guarantee rate is lower than the threshold, generates energy storage compensation instructions, and transmits them to the bus routing module. The bus circuit module calls the energy storage compensation command to obtain the bus voltage, current and power factor, calls the command to set the energy storage output target value, adjusts the proportion of photovoltaic mains power supply, monitors the bus voltage fluctuation amplitude and the deviation from the rated voltage, determines whether it exceeds the safety range, generates thermal management command and transmits it to the thermoelectric control module. The thermoelectric control module receives thermal management commands, detects the operating temperature of the power module, compares the temperature with the overheat protection threshold, reduces the module output participation coefficient when the temperature exceeds the threshold, starts waste heat recovery and adjusts the cooling output level, and outputs the data center thermoelectric balance control signal. Data center matching power sequence includes load matching index, power supply response time and active power fluctuation rate; data center power supply reliability includes time availability, power redundancy coefficient and load tracking capability; energy storage compensation instructions include power gap value, state of charge threshold and scheduling priority weight; thermal management instructions include cooling system adjustment range, overheat correction coefficient and waste heat recovery efficiency; data center thermal and power balance control signals include thermal coupling coefficient, module load factor and temperature stability index.

[0024] Specifically, such as Figure 2 , 3 As shown, the energy and power acquisition module includes: The node power acquisition submodule collects instantaneous power values ​​and acquisition timestamps of wind and solar nodes, unifies the data format table structure, synchronously filters power value combinations according to timestamps, fits the time series data vector set, and generates node power time series values. The node power acquisition submodule is equipped with power acquisition devices for wind and solar nodes respectively. Wind nodes use a real-time wind speed-wind power conversion table to establish a mapping relationship based on the output power of the wind turbines corresponding to different wind speeds. For example, when the wind speed is 6.0 m / s, the output power mapped according to the table is 1.2 kW. Solar nodes measure the output voltage and current of photovoltaic modules using instantaneous current and voltage detection devices. For example, when the irradiance is 800 W / m², the module output voltage is 34.2 V and the current is 5.3 A, then the instantaneous power is 34.2 V × 5.3 A = 181.26 W. Both nodes use a corresponding timestamp recording module to obtain the data sampling time, such as August 6, 2025, 13:15:08. The instantaneous power value and timestamp of each node are then stored. A unified format was used, with fields defined as "node number," "collection time," and "instantaneous power." Data was stored in a structured table. During the combination process, wind and solar power tables were matched by time field, rounded down to the nearest second, and power values ​​at the same timestamp were filtered. Data with a difference of no more than 1 second were considered samples from the same time. The wind and solar power were then combined into a two-dimensional vector; for example, a wind power of 1.2kW and a solar power of 181.26W would be combined into a vector [1200, 181.26]. Samples were taken every 10 seconds for 60 minutes, resulting in 360 data sets, forming a vector sequence. Based on this, the vectors were concatenated in chronological order to construct a power time-series matrix; for example, the vector at time t would be... , forming a matrix Finally, the time series values ​​of wind and solar node power are generated.

[0025] Table 1: Node Power Sampling Data Table (Three Examples)

[0026] As shown in Table 1, wind power nodes and solar power nodes synchronize data using the same timestamp, and form the power vector for each moment by splicing them together.

[0027] For the "filtering" operation involved in the above execution process, it is necessary to traverse the records in the data of the two nodes, compare the collection time field, round down the timestamp to the "second-level time code" and compare them pairwise to determine whether their absolute time difference does not exceed 1 second. If it does, the power values ​​are combined to form data of the same time. If there are multiple matching cases, the group with the smallest time difference is selected. In the example, if the wind energy collection time is 13:15:08.50 and the solar energy collection time is 13:15:08.90, the time difference between the two is 0.4 seconds, which meets the condition and is matched. The "node number" field is uniformly managed and set with numbers. For example, the wind energy node numbers are set to F001 and F002, and the solar energy node numbers are set to S001 and S002. The sampling interval and duration are preset by the controller and can be set to collect once every 10 seconds, with a total sampling duration of 3600 seconds. Finally, 360 sets of two-dimensional vectors are obtained, forming Time series set.

[0028] The normalization transformation submodule calls the node power time series value, extracts the node power level value, calculates the ratio according to the difference between the current value and the extreme value, maps it to the standard interval according to the normalization function, and generates a normalized power sequence. The normalization conversion submodule calls the node power timing value sequence. For the wind energy component in each power vector With light energy component The values ​​at each time point are extracted to form two independent sequences, namely and The maximum and minimum values ​​are obtained one by one in the two sequences, for example, let the maximum wind power be... minimum value W, maximum optical power W, minimum value W, and then for each sampling point Calculate the difference between the current power value and the extreme value, divide by the range, and then perform linear normalization. The normalization formula is: , ; For example At that time, if W, then: ; like W, then: ; By applying the same method to the 360 ​​power values, a normalized wind and solar power sequence is obtained. each for Ensure that the normalized values ​​are all within Within the range, the original extreme values ​​need to be updated in real time during the normalization process. That is, the maximum and minimum values ​​are recalculated every certain period (e.g., every 60 groups, or every 10 minutes) to avoid the imbalance of the normalization ratio caused by sudden extreme power values. The current maximum and minimum values ​​are recorded in each update period and used for the normalization calculation of the data in the subsequent period. If the range is too small, that is, the difference between the maximum and minimum values ​​is less than the set threshold, such as 10W, the power curve is considered to be in a stable fluctuation state, and the normalization result is directly set to 0.5 to indicate a stable state, avoiding the amplification of minor fluctuations. The entire normalization process does not involve algorithm calls, but only achieves standard mapping through difference calculation and division operations.

[0029] The Kalman filter submodule calls the normalized power sequence, sets the initial state estimate and error covariance, constructs the state and observation matrix, calculates the correction gain and corrects the power value, updates the covariance and state vector, and generates the data center matching power sequence. The Kalman filter submodule is based on the normalized wind and solar power sequence. First, initialize the state estimation vector. With error covariance ,set up: ; This indicates that the initial estimate is at the median level, and the error covariance is the identity matrix. ; This indicates that the initial uncertainty is 1, and then the state transition matrix is ​​constructed. With observation matrix ,in The identity matrix represents the continuity of power state over a short period of time. Also an identity matrix, representing the directly observed system state at each time step. The Kalman prediction and update process begins with the prediction step, where the predicted state vector is... The predicted covariance is Let the process noise covariance be... Next, an update step is performed to calculate the correction gain. Let the observation noise covariance be... Then calculate the residuals. Correct the state vector and update the error covariance. ,For example At that time, set , Calculate the residual Covariance prediction Correction gain The updated status is: ; And so on. to Perform iterative operations to finally generate to This constructs a matched power sequence, representing the filtered system input power estimate. This process involves only matrix addition, subtraction, multiplication, division, and inverse operations, requiring no complex techniques or parameters. and The settings are based on the system's short-term fluctuation characteristics, for example... =0.01 indicates that the maximum allowable variance for each state change step is approximately 0.01. =0.02 indicates that the maximum allowable variance of the observation error is approximately 0.02. This setting is based on the estimated range of sampling noise and sensor error tolerance. It can be evaluated before setting by the standard deviation of the sampling data. If the standard deviation is 0.15, then... This can be set to indicate that the result is the current filtered power sequence. Some observational disturbances have been removed, resulting in a smoother trend, which can be directly used for load matching analysis in edge data centers.

[0030] Specifically, such as Figure 2 , 4 As shown, the overlap calculation module includes: The power vector construction submodule obtains matching power sequence data from the data center, extracts the time series power values ​​of nodes by node number, constructs node power vectors by combining power values ​​according to time order, and aggregates power data by calling node number to generate node power vector groups. The power vector construction submodule obtains the matching power sequence data for the data center. This requires sequentially calling the node numbers, sorting them according to the data center's internal numbering rules. For example, if the node numbers are set to positive integers from 1 to N, and there are four power supply nodes in the data center, numbered 1, 2, 3, and 4, the corresponding matching power sequence data is a two-dimensional array consisting of 360 sampling points sampled once per second, in the form of... , The same process applies to the remaining nodes. In each round of node number traversal, the power time series value corresponding to the current node is extracted from each group of node power sequences. ,in Number the current node. This indicates the current time step number, which is determined during execution based on the time sequence index value. Extract power values ​​sequentially from beginning to end, and combine the 360 ​​power values ​​corresponding to each node into a complete power vector. In this step, the power time series of node 1 is as follows: Table 2: Example table of power time series (node ​​1)

[0031] As shown in Table 2, by extracting the power value of node 1 by time, a vector is constructed from the power data of this node within 360 seconds. After the node power vector is constructed, the same processing is performed on all nodes to form the node power vector. Each item is a vector of length 360. Then, the node numbers are used to aggregate the power vectors. This aggregation process uses the node number as an index to classify and organize the power vector groups of each node, forming a node power vector group matrix. The matrix has dimensions of This represents the power time series of each node, with each column corresponding to a single node. Each row represents the power data of the node at the same time point. During the aggregation operation, for each time point... Extract the first node from the node vector Rows form a set To achieve horizontal comparison of multi-node power at a single point in time, this processing action is reflected in the data structure as a matrix row-by-row extraction process, thereby constructing a complete node power vector group structure. For the aforementioned actions such as "construction," "combination," "extraction," and "aggregation," the specific operation process depends on precise access to the data index position. For example, "extraction" refers to retrieving data from an array by index. Get "Combination" refers to combining multiple Formed by array concatenation "Aggregation" refers to the arrangement of multiple node power vectors in columns at the same time point to form a matrix row. This matrix serves as a group of node power vectors, and the final result is a power vector matrix with 360 rows and 4 columns, which has a synchronization structure across time and nodes.

[0032] The baseline value determination submodule, based on the node power vector group and combined with the data center load baseline value sequence, matches the power components with the baseline value according to time alignment, compares the numerical values ​​and marks the time positions that are greater than the baseline value, and merges consecutive time points into over-limit segments to obtain the node over-limit time segment set. After the baseline value determination submodule is constructed based on the node power vector group, it first processes the power data sequentially according to the node number, traversing each node number in turn. , and its corresponding power vector Data center load benchmark sequence Time alignment is performed, during which the node power values ​​are compared with the reference value point by point to align the time points. ,judge If true, then set the current time point. Points exceeding the limit are marked, and the judgment results are recorded in binary form to form an exceedance mark sequence. ,in Represents a node exist The power value exceeds the reference value at any given time. This indicates that the limit has not been exceeded. The subsequent step is to determine the continuity of the exceeded-limit marker sequence, checking whether multiple "1" markers at consecutive time points constitute an exceeded-limit time period. In practice, this is done using a sliding window. to scanning If multiple consecutive time points (e.g., 3 or more) exist If the continuous time period is considered to constitute a valid over-limit segment, the start and end times will be recorded as follows: and and add it to the node In the set of out-of-limit sections, suppose node 2 has the following sequence of consecutive out-of-limit points: to All meet This forms a segment. If a non-contiguous "1" is encountered, it is treated as an invalid point or discarded. If a node exists in multiple segments throughout the time series, such as , , Wait, and eventually a node will be formed. Set of time limits exceeding the limit During execution, the "judgment" operation checks each pair of... and The direct comparison is completed, where the benchmark value is... It can be set to the average original load of the data center plus a 10% safety redundancy. For example, if the data center is in If the original average load per second is 230W, then set... Benchmark value sequence Generate in this way, execute the logic at each time point to form a complete baseline curve, and obtain the set of time segments exceeding the limit for each node through the above processing.

[0033] The power supply guarantee rate calculation submodule calculates the total over-limit time of nodes based on the set of over-limit time segments, calculates the ratio based on the length of the complete time series and converts it into the non-over-limit ratio, and summarizes the non-over-limit ratio data of nodes after processing it in a unified node format to obtain the power supply guarantee rate of the data center. The guarantee rate calculation submodule obtains the set of overdue time intervals for all nodes. Next, it is necessary to first calculate the total over-limit duration for each node during the entire observation period. The lengths of the concentrated segments of the time limits exceeding the limit are summed up. If the time limit exceeding the limit for node 3 is... , , The corresponding segment lengths are 11 seconds, 9 seconds, and 10 seconds, respectively, with a total overtime of [missing information]. Define this value as Meanwhile, the total data length is 360 seconds, and the calculated percentage of this node that did not exceed the limit is... That is, the non-exceeding-limit ratio of node 3 is 91.67%. This process needs to perform the same operation on each node in turn, changing the non-exceeding-limit ratio value of each node. The data is formatted uniformly to form a sequence of non-exceeding ratios. In the subsequent aggregation stage, the average of this sequence is calculated to determine the average non-exceedance rate of the nodes: ; This allows us to determine the overall power supply reliability of the data center, for example, for... The node system, its ,but This indicates that the data center's power supply reliability during this testing cycle is 91.32%. The aforementioned "calculation" operation is based on the summation of the lengths of the exceeded sections. The section length is obtained by adding one to the difference between the end time and the start time. For example, the length of section [150, 158] is... Seconds, the "ratio" calculation is a division operation between two numbers, and the "summation" operation is performed as a vector average. Add each node one by one and then divide by . The final coverage rate is obtained, and the final result is The value represents the matching stability performance of the current power supply status of the data center under the set benchmark value.

[0034] Specifically, such as Figure 2 , 5 As shown, the energy storage scheduling module includes: The data monitoring submodule obtains the power supply guarantee rate of the data center, monitors the state of charge value and charging and discharging power value of the energy storage unit, collects the load demand value and verifies its timeliness and continuity, and obtains the real-time power supply guarantee status. The data monitoring submodule obtains the data center power supply guarantee rate by first calling the node's non-over-limit ratio sequence and then calculating the overall average guarantee rate value for the current period. This process depends on the vector generated by the previous module. During execution, data is extracted from the data center monitoring system in node order. Values, such as the non-exceeding-limit ratios of nodes 1 to 4 being 0.933, 0.910, 0.895, and 0.945 respectively, then have Perform addition to obtain the sum. Divide by the number of nodes, 4, to get This is the average guarantee rate for the current power supply cycle. Subsequently, the state of charge value of the energy storage unit is retrieved. With the current charging and discharging power value ,in This can be obtained through a real-time battery monitoring module. Assuming the energy storage unit's current state of charge is 62.3% and the charging power is -2.5kW (charging is negative), then... , Next, extract the load demand value. Its source is the data center load monitoring equipment, which reads the power load demand at the current moment, assuming After the kW value is read, its timeliness needs to be verified. The verification process compares the load data generation timestamp with the current system clock time. If the time difference is less than 2 seconds, the data is considered valid. Here, we assume the load value was read at 14:28:10 and the system time is 14:28:11, in which case the data is valid. Simultaneously, the continuity of the load value within the acquisition period is verified by comparing the load values ​​over the past 5 consecutive seconds. The adjacent differences were 0.1, 0.1, 0.2, and 0.1 respectively, with a maximum fluctuation of 0.2kW, which is less than the set continuity threshold of 0.5kW. Therefore, it was determined to be in a continuous and stable state, and the monitoring period ended. The data collected at the second level forms a power supply assurance status ternary set. Its value is The triplet is related to the load requirement value. Together, they constitute the real-time power supply guarantee status for the current cycle.

[0035] Table 3: Key parameter values ​​during a monitoring period of 220 seconds

[0036] As shown in Table 3, all key monitoring indicators have been collected during the monitoring period, the load demand is valid, the data is continuous and stable, and the power supply status indicators are complete.

[0037] The guarantee rate judgment submodule compares the real-time power supply guarantee status with the power supply safety threshold, calls the load demand value and the state of charge value, calculates the deviation between the current load shortfall power and the remaining output power of the energy storage, judges the power supply safety range status, and obtains the guarantee rate safety judgment value. The failure rate assessment submodule obtains the real-time power supply guarantee status. First, the system's preset power supply safety threshold is invoked. The current power supply guarantee rate The current power supply stability is directly compared with this threshold to determine whether it is within a safe range. The settings should refer to the data center operation strategy and be set to 90%, that is... ,when At that time, it is considered that the power supply status meets the guarantee requirements. If so, it is considered that there is a power supply risk, and further analysis of the load gap and energy storage output capacity is needed. Assuming the current... ,because Therefore, it is determined that the power supply does not meet the safety threshold, and then the current load shortfall power is calculated. It is the difference between the load power demand and the energy storage output power, i.e. ,like , kW, then kW indicates that there is still a power supply gap of 4.5kW. Further calculations are needed to determine the remaining output deviation of energy storage. This value is based on the current state of charge. With minimum charge threshold To determine its sustainable discharge capability, if set... ,current The allowable discharge ratio is Assuming the maximum power of the energy storage system is 20kW, the sustainable output power corresponding to a 7.5% charge level is calculated using a linear relationship as follows: Then the deviation value This indicates that there is an additional 3.0kW power shortfall that cannot be compensated by the energy storage system. Therefore, the system is judged to be in an unsafe power supply range. The above judgment process forms the power supply security status indicator. If the system is secure, then If it is not safe, then In this example, the final This indicates that the current power supply guarantee is not met.

[0038] The particle swarm optimization calculation submodule calls the guarantee rate safety judgment value and the state of charge value, sets the particle encoding dimension including load gap, discharge power and scheduling cycle, it iteratively updates the scheme based on the particle swarm optimization algorithm, and generates energy storage compensation instructions by combining the scheduling power range and the load demand curve. Particle Swarm Optimization Computation Submodule Reading With current state of charge ,like Then, the optimized scheduling strategy is executed to generate compensation instructions, defining the particle encoding dimension as including the current load gap power. Target discharge power of the schedule and scheduling cycle Three variables constitute the encoding vector initialize it to This indicates that the current particle state has a 4.5kW gap, a 3.2kW discharge power, and a 30-second scheduling period. During the optimization iteration process, different scheduling discharge power values ​​are tried one particle at a time. With time Combinations are selected from the candidate intervals, and compared with the actual load gap. A difference assessment is performed. If the difference is the smallest, the current combination is recorded as the optimal scheduling state. In the example, if the scheduling power is 3.7kW and the period is 35 seconds, the remaining gap is 0.1kW, which is better than the original combination. As one of the optimal points of the iteration, the objective function value is calculated for the particles. This function can be set as the notch squared error, i.e. The particle with the smallest error is selected as the current optimal position. After 20 consecutive iterations, the solution with the smallest error is selected as the scheduling compensation output result. Finally, the energy storage system scheduling command is generated to control it to output 3.7kW discharge power in the next 35 seconds to compensate for the current load gap. If the state of charge is insufficient or the minimum state of charge is reached within the scheduling cycle, the scheduling power is automatically limited to the maximum output power to ensure that the scheduling strategy can be implemented.

[0039] Specifically, such as Figure 2 , 6 As shown, the busbar consists of modules including: The energy storage command parsing submodule obtains the energy storage compensation command, collects the HVDC bus voltage, current and power factor values ​​through the data communication interface, calculates the difference between the electrical target value in the command and the current electrical data, determines the matching degree between the power supply status and the energy storage command, merges the deviation items, and obtains the electrical deviation response quantity. After obtaining the energy storage compensation command, the energy storage command parsing submodule sequentially retrieves the real-time voltage, current, and power factor data of the HVDC bus connected to the energy storage device from the data communication interface. The retrieval action corresponds to the reading operation of the specified sensor port signal in the acquisition device. For example, when retrieving the bus voltage signal, it needs to read the data buffer content of sensor number V101 and convert it to standard voltage units in real time. Assuming the current voltage is 720V, current is 26A, and power factor is 0.94, the corresponding timestamp is T=12:05:03. Then, this set of real-time data is compared and analyzed with the target output power, current limit, and target bus voltage value in the compensation command. Assuming the current energy storage command's target power is 20kW, target current is 28A, and target voltage is 750V, the difference between the current energy storage electrical state and the target state needs to be calculated to form a difference index. The voltage deviation is the target voltage of 750V minus the real-time voltage of 720V, resulting in 30V; the current deviation is 28A minus 26A, resulting in 2A; and the power deviation is calculated using the power calculation formula. The current actual power is calculated to be... W, or 17.59kW, deviates from the target power of 20kW by 2.41kW. These deviations are then aggregated to form a unified electrical deviation response value, which is set as the weighted comprehensive error value. The calculation method uses a weighted linear combination, where the voltage weight is set to 0.4, the current weight to 0.3, and the power weight to 0.3. The corresponding comprehensive error is: ; This response value serves as a quantitative evaluation indicator of the matching degree between the energy storage execution state and the command state. Its acquisition process can specifically reflect the degree of deviation of the current operating state of the energy storage system, thereby providing a precise adjustment basis for the subsequent scheduling module. Based on the above process, Table 4 lists the command target value, the collected electrical value, and the deviation calculation results within a certain scheduling cycle: Table 4: Comparison of Energy Storage Electrical Commands and Acquired Data

[0040] As shown in Table 4, the deviations between the collected data and the target command were calculated to obtain results of 30V, 2A, and 2.41kW, respectively. Based on the set weights, a comprehensive response value of 13.323 was calculated, which was used as the input value for the next stage power setting and control submodule. The advantage of the comprehensive error response value is that by introducing a weighted error aggregation method for the three key electrical parameters, a unified quantitative calculation of multi-dimensional deviation data was achieved, which facilitates direct matching and judgment by the system in subsequent actions. The weighting of the participating items was determined with reference to the sensitivity weight of parameters in the stability of the electrical system. Voltage was the primary control target, assigned a value of 0.4, while current and power were set to 0.3 and 0.3 respectively, as secondary targets. Assuming that voltage fluctuations exceeding 25V in the system would affect the stability of the bus power, the deviation benchmark was set to 25V. The current deviation was 30V, which is in the abnormal range. The current error benchmark was ±1.5A, and the current error of 2A is also within the limit. The power error limit was set to 2kW, and the current value of 2.41kW is slightly higher than the critical value. Therefore, the overall comprehensive response value of 13.323 was determined to be in the moderate deviation range. This result indicates that there is a certain degree of mismatch between the current energy storage system response state and the command target, requiring further power correction processing.

[0041] The power setting and control submodule matches the upper limit of the energy storage command power with the current value based on the electrical deviation response. It calculates the power difference and corrects the target value of the energy storage output power according to the output data of the photovoltaic MPPT module and the AC / DC module of the mains power, adjusts the ratio of photovoltaic and mains power supply, and generates a dual-source power matching value. The power setting and control submodule interprets the electrical deviation response value. Then, the target power setpoint is first extracted from the energy storage compensation command. Simultaneously, the actual output current of the current energy storage system is called through the electrical interface module. HVDC bus voltage and power factor The actual power is recalculated to obtain the current power output value. Then, the difference between the current output power and the target power is determined. This difference serves as a reference input for the current power regulation correction. The system determines the value based on the magnitude of the current difference and the power regulation sensitivity parameter setting range. If the power deviation range is set to 0–1kW for mild, 1–3kW for moderate, and 3–5kW for severe, then the current value of 2.41kW falls into the moderate deviation range, requiring a correction algorithm for phased compensation. The system calls the latest output power data from the photovoltaic MPPT module and the AC / DC mains module, respectively. , The sum of these two is 17.6kW. Given that the current actual output of the energy storage is 17.59kW, this indicates that most of the power in the existing power supply structure is supplied by external input. It is necessary to determine whether there is redundant power compensation capability. If it is determined that the current combined MPPT and AC output is close to the target set power, then the maximum allowable upward adjustment value is calculated from the allowable limits of the energy storage system. Assuming the current maximum power setting of the energy storage module is 24kW, and the current output is 17.59kW, the remaining power adjustment space is... The system sets a correction ratio for the power adjustment level based on the electrical deviation response value. The correction ratio for moderate deviation is set to 45%, therefore the calculated corrected power is... After adjusting the current output power by this amount, the target output power is corrected to... Since the value exceeds the initial target value of 20kW, the system adjusts it back to the set maximum value of 20kW through the limit function and generates a target output command. The final output of the energy storage power setting and control submodule is: increase the energy storage power value to 20.0kW, while keeping the ratio of photovoltaic to grid power input unchanged, and the power supply ratio will not be further optimized or adjusted in the future.

[0042] The voltage deviation determination submodule extracts the bus voltage fluctuation value and the set voltage deviation benchmark based on the dual-source power matching value, determines the temperature control triggering condition in combination with the power distribution status, calculates the voltage abnormal section and evaluates the temperature control response requirements, and generates thermal management instructions. After the voltage deviation determination submodule starts, it first calls the energy storage output power command generated in the previous module and the dual-source power supply matching result to extract the bus voltage fluctuation data within the corresponding time period. Assuming the control period is between T=12:05:03 and T=12:06:03, the system continuously acquires the bus voltage data sequence within this time period through the bus voltage acquisition device. ; Then, based on the set voltage deviation reference value Calculate the voltage difference at each moment, in sequence as follows: The absolute value is taken and the volatility is assessed in the form of an average difference. The average volatility difference is: ; The voltage deviation determination grading intervals are set as follows: A slight deviation is 0–15V, a moderate deviation is 15–30V, and a severe deviation is greater than 30V. Therefore, the current 25.89V falls within the moderate deviation range, which is consistent with the power matching results in the dual-source power supply structure. The system determines that the current bus voltage fluctuation is mainly caused by the instantaneous current output fluctuation of the AC mains power. It then assesses whether the system temperature control trigger condition is met. The bus voltage deviation trigger temperature control condition is set to an average deviation exceeding 25V. The current value is 25.89V, which meets the trigger condition. At the same time, the current temperature monitored by the AC mains module is 65℃. The trigger threshold is set to 70℃, which has not yet reached the trigger threshold. However, the voltage fluctuation has reached the temperature control strategy requirements. Therefore, the system issues an instruction to the thermal management module to generate a phased thermal management instruction label, indicating that it is currently in a state where dynamic temperature rise tracking is required. The system also starts the bus voltage real-time monitoring subroutine for continuous updates of the subsequent temperature control strategy.

[0043] Specifically, such as Figure 2 , 7 As shown, the thermoelectric control module includes: The temperature monitoring submodule calls thermal management commands to extract the operating temperatures of the photovoltaic MPPT, energy storage DC / DC, and mains AC / DC modules. It collects the temperature sequence values ​​of each module, synchronizes and organizes the data according to timestamps, removes missing items and outliers, and then performs interpolation processing on the data to generate a temperature dataset. The temperature monitoring submodule, based on the thermal management instructions generated by the preceding thermal management module, performs temperature information acquisition tasks for three types of modules: photovoltaic MPPT, energy storage DC / DC, and mains AC / DC. First, it calls the module identification field in the thermal management instructions and performs a status confirmation operation on the specified module. It reads the running status code from the module status register to confirm that the current module is in working state and the temperature acquisition function enable flag is "1". Then, it reads the raw temperature data sequence recorded by the module's embedded temperature sensor, where the MPPT module's temperature record for the current hour is... The temperature sequence of the DC / DC module is as follows: ; The AC / DC module is: ; The above data was recorded with a sampling period of 5 minutes, totaling five time points. The timestamp synchronization mechanism was then invoked to extract the system's real-time clock and compare it sequentially with the timestamp tags of each module. If any null records or data gaps due to communication errors were found, an anomaly detection process was executed, sequentially determining whether the data was numerical and whether it exceeded a reasonable temperature range. If the above conditions are not met, the data is marked as an outlier. Linear interpolation is then used to repair the missing or outlier data. For example, if the third sampling point of the DC / DC module has no record due to a sensor malfunction, its correction value is the average of the two preceding and following points. After data repair is completed, the three sets of data are aligned according to a unified timestamp to form a standard temperature dataset, which is then output in JSON format for downstream module calls, as shown below: Table 5: Standard Data Set for Temperature Monitoring

[0044] As shown in Table 5, the module temperature data is distributed within the normal operating range set by the temperature control strategy. The corrected data is uniformly incorporated into the subsequent power calculation process. The dataset output by the temperature monitoring submodule meets the input standard for real-time power adjustment. The data marked with "*" in the table are interpolation correction values.

[0045] The results demonstrate that by performing operations such as status verification of the target module, sensor retrieval, temperature data acquisition, outlier removal, and interpolation completion by the temperature monitoring submodule, a structured standard temperature dataset can be obtained, in which the module temperature is maintained within the specified range. This provides temperature reference input data for subsequent power regulation strategy calculations, forming key data support for combined heat and power management.

[0046] The temperature comparison and adjustment submodule extracts the current temperature from the temperature dataset, compares it with the module overheating threshold, records the temperature deviation, calculates the dynamic adjustment factor, updates the power ratio in combination with the power output value, and generates the adjusted power participation coefficient. After receiving the standard temperature dataset, the temperature comparison and adjustment submodule first extracts the current temperature value of the module, which is then used by the photovoltaic MPPT module. Energy storage DC / DC module AC / DC module Then, it calls the module overheat threshold set in the system parameter library, which are respectively , , Perform a term-by-term difference calculation on the current temperature and its corresponding threshold to obtain the temperature offset magnitudes as follows: , , Next, based on the temperature sequence values ​​collected by each module, its temperature fluctuation standard deviation is calculated, where the photovoltaic MPPT module sequence is as follows: The standard deviation is calculated as follows: ; The conclusion is The other two modules are respectively , Then, a normalized offset is calculated for each temperature sequence, which involves dividing each temperature offset by the corresponding normalization factor: ; If the first temperature of the AC / DC module is 64.0℃, the normalized denominator is... The numerator of the difference is Normalized offset is The temperature data is then normalized and summed based on the offset, and then divided by the sample size n=5 to obtain the average normalized offset. The system then retrieves the current real-time power data based on the module's current power output value and maximum allowable power output value. Calculate the adjusted power participation factor using the formula: ; in, Representing the The adjusted power participation factor of the module. Representing the Current power output value of the module. Representing the The module's maximum allowable power output value, Representing the Module power adjustment tolerance, Representing the Module in Temperature sequence values ​​at each time point, Representing the The module's overheating temperature threshold, Representing the The standard deviation of the module temperature fluctuation range The total number of samples representing temperature series values. Indicates the first The first module temperature sample sequence Summation of terms.

[0047] Substituting the parameters, we get: ; Let the sum of the five normalized offsets be 1.23, then: ; Similarly, the same calculations were performed on the other two modules, resulting in a participation coefficient of 0.592 for the DC / DC module and 0.565 for the AC / DC module. The three participation coefficients were then normalized and assigned to the corresponding channels of the system power scheduling controller as the basis for power allocation in the next cycle, ultimately generating three adjusted power ratio control commands.

[0048] The heat recovery scheduling submodule determines whether to start the heat recovery path based on the adjusted power participation coefficient, collects heat recovery power and cooling parameters, calculates heat release rate and cooling level by combining module temperature and cooling load, and outputs data center heat and power balance control signal. After obtaining the adjusted power participation factor, the heat recovery scheduling submodule first performs a residual heat value judgment operation, calculating the heat power release of the three modules under the current operating state. Assuming the photovoltaic MPPT module has a current electrical efficiency of 92% and an input power of 5.7kW, the heat power is... The energy storage DC / DC module has an efficiency of 91%, an input power of 6.0kW, and a thermal power of [missing information]. The AC / DC module has an efficiency of 89%, an input power of 6.9kW, and a thermal power of [missing information]. Then, the heat release of the modules is summed to obtain the total heat release. The system checks if the current cooling output exceeds a preset threshold of 1.5kW. If so, the current value exceeds the threshold, requiring the activation of the heat recovery path. It calls the current status flag field of the cooling loop to determine if it is in a working state. If not, the heat recovery path is set to active. Simultaneously, the cooling module load capacity is calculated as 2.5kW, and the cooling level is determined. In the cooling capacity level classification, the range of light cooling level is set as 0–0.4, medium as 0.4–0.75, and heavy as 0.75–1.0. The current value of 0.702 is in the range of medium cooling level. The system calls the PID control strategy to execute the cooling fan speed adjustment. The current default fan speed is set to 30%, which is increased to 55% according to the medium level. The thermal management control byte is updated to generate the thermal balance control signal for the current cycle. Synchronous control commands are sent to the cooling loop, power scheduler, and temperature monitoring module respectively to complete the full-link status update of heat recovery and cooling response for this cycle.

[0049] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A wind-solar-storage integrated edge data center power supply system, characterized in that, The system includes: The energy power acquisition module acquires the instantaneous power of wind and solar nodes, performs normalization processing, uses the Kalman filter algorithm to filter out noise, generates a data center matching power sequence, and transmits it to the overlap calculation module. The overlap calculation module receives the matching power sequence of the data center, obtains the data center load benchmark, constructs the node power vector, compares the vector components with the benchmark size relationship, marks the time window exceeding the benchmark, calculates the window duration percentage, obtains the data center power supply guarantee rate, and transmits it to the energy storage scheduling module. The energy storage scheduling module receives the power supply guarantee rate of the data center, detects the energy storage charge status and power, monitors load demand, determines the relationship between the guarantee rate and the safety threshold, and when the guarantee rate is lower than the threshold, it uses the particle swarm optimization algorithm to calculate the compensation scheme, generates an energy storage compensation command, and transmits it to the bus routing module. The bus routing module calls the energy storage compensation command to obtain the bus voltage, current, and power factor, calls the command to set the energy storage output target value, adjusts the proportion of photovoltaic mains power supply, monitors the bus voltage fluctuation amplitude and the deviation from the rated voltage, determines whether it exceeds the safety range, generates a thermal management command, and transmits it to the thermoelectric control module.

2. The integrated wind-solar-storage edge data center power supply system according to claim 1, characterized in that, The data center matching power sequence includes load matching indicators, power supply response time and active power fluctuation rate; the data center power supply guarantee rate includes time availability, power redundancy coefficient and load tracking capability; the energy storage compensation command includes power gap value, state of charge threshold and scheduling priority weight; and the thermal management command includes cooling system adjustment range, overheat correction coefficient and waste heat recovery efficiency.

3. The integrated wind-solar-storage edge data center power supply system according to claim 1, characterized in that, The energy acquisition module includes: The node power acquisition submodule collects instantaneous power values ​​and acquisition timestamps of wind and solar nodes, unifies the data format table structure, synchronously filters power value combinations according to timestamps, fits the time series data vector set, and generates node power time series values. The normalization transformation submodule calls the node power time series value, extracts the node power level value, calculates the ratio according to the difference between the current value and the extreme value, maps it to the standard interval according to the normalization function, and generates a normalized power sequence. The Kalman filter submodule calls the normalized power sequence, sets the initial state estimate and error covariance, constructs the state and observation matrix, calculates the correction gain and corrects the power value, updates the covariance and state vector, and generates the data center matching power sequence.

4. The wind-solar-storage integrated edge data center power supply system according to claim 3, characterized in that, The node power time series value is a sequence of instantaneous power data of multiple wind and solar energy nodes arranged in timestamp order; The normalization function is a mathematical function that performs numerical transformation based on the positional proportion of node power values ​​within the target mapping interval.

5. The wind-solar-storage integrated edge data center power supply system according to claim 3, characterized in that, The overlap calculation module includes: The power vector construction submodule obtains the matching power sequence data of the data center, extracts the node time series power value according to the node number, combines the power values ​​according to the time order to construct the node power vector, and calls the node number to aggregate the power data to generate a node power vector group. The benchmark value determination submodule, based on the node power vector group and combined with the data center load benchmark value sequence, matches the power components with the benchmark value according to time alignment, compares the numerical values ​​and marks the time positions that are greater than the benchmark value, and merges consecutive time points into over-limit segments to obtain the node over-limit time segment set. The power supply guarantee rate calculation submodule calculates the total overdue time of nodes based on the set of overdue time segments, calculates the ratio based on the length of the complete time series and converts it into the non-overdue ratio, processes it according to a unified node format and summarizes the non-overdue ratio data of nodes to obtain the power supply guarantee rate of the data center.

6. The wind-solar-storage integrated edge data center power supply system according to claim 5, characterized in that, The energy storage scheduling module includes: The data monitoring submodule acquires the power supply guarantee rate of the data center, monitors the state of charge value and charging and discharging power value of the energy storage unit, collects the load demand value and verifies its timeliness and continuity, and obtains the real-time power supply guarantee status. The guarantee rate judgment submodule compares the real-time power supply guarantee status with the power supply safety threshold, calls the load demand value and the state of charge value, calculates the deviation between the current load shortfall power and the remaining output power of the energy storage, judges the power supply safety range status, and obtains the guarantee rate safety judgment value. The particle swarm optimization calculation submodule calls the security judgment value of the guarantee rate and the state of charge value, sets the particle encoding dimension including load gap, discharge power and scheduling cycle, and generates energy storage compensation instructions based on the iterative update scheme of the particle swarm optimization algorithm and the scheduling power range and load demand curve.

7. The wind-solar-storage integrated edge data center power supply system according to claim 6, characterized in that, The busbar consists of modules including: The energy storage command parsing submodule obtains the energy storage compensation command, collects the HVDC bus voltage, current and power factor values ​​through the data communication interface, calculates the difference between the electrical target value in the command and the current electrical data, determines the matching degree between the power supply status and the energy storage command, merges the deviation items, and obtains the electrical deviation response quantity. The power setting and control submodule, based on the electrical deviation response, matches the upper limit of the energy storage command power with the current current value, calculates the power difference and corrects the target value of energy storage output power according to the output data of the photovoltaic MPPT module and the AC / DC module of the mains power, adjusts the ratio of photovoltaic and mains power supply, and generates a dual-source power matching value. The voltage deviation determination submodule extracts the bus voltage fluctuation value and the set voltage deviation benchmark based on the dual-source power matching value, determines the temperature control triggering condition in combination with the power distribution state, calculates the voltage abnormal section and evaluates the temperature control response requirements, and generates thermal management instructions.

8. The wind-solar-storage integrated edge data center power supply system according to claim 7, characterized in that, The electrical deviation response is the calculated result of the difference between the electrical target value in the energy storage command and the current bus voltage, current and power factor data; The dual-source power matching value is a power allocation value calculated based on the electrical deviation response, the upper limit of the energy storage command power, and the output power of photovoltaic and mains power. The target value for energy storage output power is the set value for the output power of the energy storage device determined by combining the upper limit of the energy storage command power, the electrical deviation response, and the current current data.

9. The integrated wind-solar-storage edge data center power supply system according to claim 1, characterized in that, The thermoelectric control module receives the thermal management command, detects the operating temperature of the power module, compares the temperature with the overheat protection threshold, reduces the module output participation coefficient when the temperature exceeds the threshold, starts waste heat recovery and adjusts the cooling output level, and outputs the data center thermoelectric balance control signal. The data center thermoelectric balance control signal includes thermal coupling coefficient, module load factor, and temperature stability index.

10. The wind-solar-storage integrated edge data center power supply system according to claim 9, characterized in that, The thermoelectric control module includes: The temperature monitoring submodule calls the thermal management command to extract the operating temperatures of the photovoltaic MPPT, energy storage DC / DC, and mains AC / DC modules, collects the temperature sequence values ​​of each module, synchronizes and organizes the data according to the timestamp, removes missing items and outliers, and then performs interpolation processing on the data to generate a temperature dataset. The temperature comparison and adjustment submodule extracts the current temperature from the temperature dataset, compares it with the module overheating threshold, records the temperature deviation, calculates the dynamic adjustment factor, updates the power ratio in combination with the power output value, and generates the adjusted power participation coefficient. The heat recovery scheduling submodule determines whether to start the heat recovery path based on the adjusted power participation coefficient, collects heat recovery power and cooling parameters, calculates the heat release rate and cooling level by combining the module temperature and cooling load, and outputs the data center heat and power balance control signal.

Images