A park energy storage optimization system and method based on multi-source intelligent dispatching

Abstract

The application discloses a kind of park energy storage optimization system and method based on multi-source intelligent scheduling, it is related to park energy storage optimization technical field, including: the type information of building in park and energy storage related information are collected, the ordering mode of building to be matched is planned, the ordering sequence of current park to be matched object is generated, after generating the ordering sequence of current park to be matched object, building pair matching is carried out, matched building pair and failed building pair are generated, different energy storage data prediction mode is planned for matched building pair and failed building pair respectively, the charge and discharge optimization is carried out to park energy storage equipment, improve the data matching efficiency, save load data matching time, and then improve the efficiency of charge and discharge power regulation to building in park, conducive to efficient park energy storage optimization.

Description

Technical Field

[0001] This invention relates to the field of park energy storage optimization technology, specifically a park energy storage optimization system and method based on multi-source intelligent scheduling. Background Technology

[0002] Park energy storage is a core component for achieving park-level energy synergy. By integrating distributed photovoltaics, adjustable loads, and energy storage to form a self-balancing energy ecosystem unit, it has been regarded as the core path to zero-carbon parks. Park energy storage optimization can be carried out in a differentiated manner from core dimensions such as load coordination scheduling, system capacity configuration, and infrastructure integration, which can effectively improve power supply reliability.

[0003] Regulating the charging and discharging power of energy storage devices is one of the core means to achieve optimized scheduling of energy storage in industrial parks. For multi-building parks, coordinating the charging and discharging power of some buildings can effectively reduce the workload of regulation. However, this requires matching with buildings that can be coordinated to adjust the charging and discharging power. In existing technologies, the load data of all buildings are usually matched one by one in a random order to find mutually matching buildings and then coordinate the adjustment. This method has the problem of low matching efficiency, which cannot effectively improve the efficiency of charging and discharging power regulation and is not conducive to high-efficiency optimization of energy storage in industrial parks. Summary of the Invention

[0004] This invention provides the following technical solution: a method for optimizing park energy storage based on multi-source intelligent scheduling, comprising the following steps:

[0005] S1: Collect information on the type of buildings and energy storage-related information within the park. The energy storage-related information includes load data and photovoltaic data of all buildings within the park.

[0006] S2: Plan the sorting method for the buildings to be matched, and generate a sorting sequence of objects to be matched in the current park: Randomly number all buildings in the current park according to their type. Buildings of the same type have the same number. Generate the building number sequence of the current park as {1,2,...,n}, where n represents the total number of different types of buildings in the current park. Count the number of buildings with different numbers in the current park, and select the building with the most numbers as the first-order building. The number of the first-order building is the first-order number. Collect the set of the number of matches between the buildings with the same type as the remaining n-1 building types and the buildings with the same type as the first-order building, which is F={F1,F2,...,F... n-1F1 indicates that a total of F1 buildings of the same type as the first-order building have been matched with the remaining buildings of the first type. Matching between buildings refers to matching their load data. The query set F checks if there are any equal values: if not, it checks for any sorting interference, arranges the remaining n-1 building numbers in descending order of values ​​in set F, and then adds the first-order number to the sorting queue, generating the sorted sequence of objects to be matched in the current park as {A1, A2, ..., A...}. n}, A1 represents the first ordinal number, A2 represents the building number with the highest probability of successfully matching the first ordinal building among the remaining n-1 types of buildings, i.e., the second ordinal number. The larger the value in F, the greater the probability of the two types of buildings successfully matching. If it exists, it is determined that there are sorting interference items, the priority order of sorting interference items is planned, and the sorting sequence of objects to be matched in the current park is generated after the planning is completed.

[0007] S3: After generating the sorted sequence of objects to be matched in the current park, perform building pair matching to generate matching building pairs and unmatched buildings. The matching process is as follows: match all first-order buildings with the remaining buildings one by one according to the sorted sequence. Do not match all first-order buildings with each other, and do not match all buildings with the same position. If there is a building that matches a random first-order building, then the two matching buildings are considered a matching building pair. If a random first-order building does not find a matching building, then the corresponding first-order building is considered an unmatched building. Remove the matching building pairs and unmatched buildings from the objects to be matched. After completing the matching of all first-order buildings, match all second-order buildings with the remaining buildings one by one according to the sorted sequence. After completing the matching of all second-order buildings, proceed to the matching of third-order buildings, until only the nth-order building remains in the objects to be matched. All remaining nth-order buildings are considered unmatched buildings.

[0008] This invention considers the potential strong negative correlation between load data of two different types of buildings. First, it linearly sums the load data of the two buildings with this relationship. After linear summation, the random noise and extreme fluctuations of a single building's load partially cancel each other out during the summation process, making the load curve corresponding to the load data smoother. This allows the LSTM model to more easily capture long-term trends rather than local disturbances, significantly improving the stability of load data prediction. Second, when matching buildings, considering the diversity of building types within the park, buildings are first classified and numbered according to different types. Then, historical data analysis is used to determine the building types previously matched by different building types, and the number of matches is counted. The building type with the most buildings in the park is prioritized for matching. Building numbers are arranged in descending order of the number of matches, planning the matching order of buildings in the current park. This increases the probability that the building with the highest proportion of buildings will match its counterpart as early as possible, effectively improving load data matching efficiency and saving load data matching time. This, in turn, improves the efficiency of adjusting the charging and discharging power of buildings in the park, facilitating high-efficiency park energy storage optimization.

[0009] S4: Plan different energy storage data prediction methods for matched building pairs and unmatched buildings, and optimize the charging and discharging of energy storage equipment in the park.

[0010] Preferably, collaborative data prediction and charge / discharge optimization are performed for matched building pairs: The load data of a randomly selected matched building pair are linearly summed, and the photovoltaic data of the corresponding matched building pair are linearly summed. The linearly summed load data and photovoltaic data are input into an LSTM model. The LSTM model is used to predict the load data and photovoltaic data at future time points. The LSTM model outputs the load data and photovoltaic data at future time points. The time interval between load data at different future time points is T, and the time interval between photovoltaic data at different future time points is equal to T. The future load data, photovoltaic data, and SOC are input into an MPC model, and... The MPC model outputs the optimal charge and discharge power sequence of the energy storage device {Pess(1),Pess(2),...,Pess(j)}, where j represents the number of parameters in the optimal charge and discharge power sequence output by the MPC model, and Pess(1) represents the predicted optimal charge and discharge power in the time period [t,t+T], where t represents the current time. If Pess(1)>0, the discharge power of the energy storage device of the corresponding matching building pair will be increased in the time period [t,t+T]. If Pess(1)=0, the charge and discharge power of the energy storage device of the corresponding matching building pair will not be adjusted. If Pess(1)<0, the charging power of the energy storage device of the corresponding matching building pair will be increased in the time period [t,t+T].

[0011] Preferably, data prediction and charge-discharge optimization are performed separately for each building with failed matching: the load data and photovoltaic data of a randomly selected building with failed matching are input into the LSTM model, and the LSTM model outputs the load data and photovoltaic data at future times. The load data, photovoltaic data, and SOC at future times are input into the MPC model, and the MPC model outputs the optimal charge-discharge power sequence of the energy storage device. The charge-discharge power of the energy storage device of the corresponding building with failed matching is adjusted according to the optimal charge-discharge power sequence, and the adjustment method is the same as that of the matching building pair;

[0012] The MPC model is a control strategy based on rolling optimization and feedback correction, which can integrate load prediction data, photovoltaic prediction data, and real-time SOC, and output the optimal charge-discharge power sequence through rolling optimization to achieve the coordination of source-network-load-storage.

[0013] Preferably, the judgment method for the mutual matching of the load data is as follows: the load powers of two buildings are collected in the way that the initial collection time is the same and the collection time interval is the same, generating two load power sequences W1 and W2. The lengths of the two load power sequences are equal and the timestamps are aligned with each other. According to the following formula, it is judged whether the load data of the two buildings match each other:

[0014] ;

[0015] where, P represents the negative correlation degree between the loads of the two buildings, Cov(W1, W2) represents the covariance between the load power sequences W1 and W2, represents the standard deviation of the sequence W1, represents the standard deviation of the sequence W2. The negative correlation degree threshold is set as R1, where R1 is a negative number. If P < R1, it is judged that the load data of the two buildings match each other, that is, the load data of the two buildings are strongly negatively correlated; otherwise, it is judged that the load data of the two buildings do not match.

[0016] Preferably, if there is a sorting interference item, the planning method for the priority order of the sorting interference item is as follows: k buildings with the same number of previous matches as the building of the first order are selected. The building with the same type as the building of the first order is used as the reference building. For each type of building, there are x buildings that are matched with the reference building one by one, and there are a total of x * k historical matching building pairs, where * is the multiplication sign. The information of x historical matching building pairs corresponding to a randomly selected building is collected: the load power sequences of the two buildings belonging to the i-th building pair are respectively C(m) and D(m), and a cross-correlation function H is constructed: , where m represents the integer index, y represents the time shift parameter, and M represents the length of the two load power sequences. According to the formula calculate the time shift parameter y when the cross-correlation function H reaches the maximum value（max,i） The time-shift parameter set Y={y} is obtained. （max,1） ,y （max,2） ,...,y （max,i） ,...,y （max,x）}, y （max,x） Let Y represent the time-shift parameter of the x-th building pair corresponding to a random building. The average value of the parameters in set Y is used to obtain the comprehensive time-shift parameter between the random building and the benchmark building. The comprehensive time-shift parameters between k buildings and the benchmark building are calculated and compared. The numbers of the k buildings are arranged in ascending order of comprehensive time-shift parameter, with the number of the building with the smallest comprehensive time-shift parameter in the first priority position. The building numbers are arranged in descending order of the values ​​in set F. The arranged numbers of the k buildings are added to the queue to generate the sorted sequence of objects to be matched in the current park.

[0017] During the data matching process, this invention also considers that the number of matches between different types of buildings and the first-order building may be equal, which would interfere with the sorting. Therefore, by further constructing the cross-correlation function between buildings and analyzing the time shift parameters, this invention plans how to sort the building numbers with equal matching numbers in the event of such interference, ensuring the smooth sorting of building matching under interference conditions, further improving data matching efficiency and saving data matching time.

[0018] A multi-source intelligent scheduling-based park energy storage optimization system includes a park information acquisition module, a building matching module, a building matching module, and an energy storage optimization module. The park information acquisition module collects information on the type of buildings in the park, load data of all buildings in the park, and photovoltaic data. The building matching module plans the sorting method of the buildings to be matched and generates a sorting sequence of the buildings to be matched in the current park. The building matching module performs building pair matching and generates matched building pairs and unmatched buildings. The energy storage optimization module plans different energy storage data prediction methods for matched building pairs and unmatched buildings to optimize the charging and discharging of the park's energy storage equipment.

[0019] Preferably, the building sorting module includes a sorting unit with interfering items and a sorting unit without interfering items;

[0020] The interference item sorting unit is used to plan the priority order of sorting interference items when it is determined that there are sorting interference items, and then add the planned sorting interference items to the queuing queue to generate the sorting sequence of objects to be matched in the current park.

[0021] The interference-free sorting unit is used to directly arrange the building numbers according to the number of matching items between buildings when it is determined that there are no interference items, thereby generating a sorting sequence of objects to be matched in the current park.

[0022] Preferably, the energy storage optimization module includes a collaborative energy storage optimization unit and a separate energy storage optimization unit;

[0023] The collaborative energy storage optimization unit is used to perform collaborative charging and discharging regulation of the energy storage devices of the matching building pairs after data prediction of the matching building pairs.

[0024] The individual energy storage optimization unit is used to perform individual charge and discharge regulation of the energy storage devices of each building that failed to match data prediction.

[0025] Compared with the prior art, the beneficial effects of the present invention are:

[0026] This invention superimposes load data of mutually matching buildings in the park and superimposes photovoltaic data to predict load and photovoltaic data. Then, the predicted data is input into the MPC model to predict the optimal charging and discharging power. Based on the prediction results, the park's energy storage is optimized, transforming passive monitoring and regulation into active prediction and optimization, thereby reducing the probability of abnormalities in the park's energy storage.

[0027] Considering the potential strong negative correlation between load data of two different types of buildings, the load data of the two buildings with this relationship are first linearly summed. After linear summation, the random noise and extreme fluctuations of the load of a single building partially cancel each other out during the summation process, which can make the load curve corresponding to the load data tend to be smoother. Then, load data prediction is performed, which makes the LSTM model more likely to capture long-term trends rather than local disturbances, significantly improving the stability of load data prediction. Secondly, when matching buildings, considering the diversity of building types in the park, the buildings are first classified and numbered according to different types. Then, the building types matched by different types of buildings in the past are analyzed through historical data, and the number of matches is counted. The building type with the most buildings in the park is ranked first to be matched. The building numbers are arranged in descending order of the number of matches, and the matching order of buildings in the current park is planned. This increases the probability that the building with the highest number of buildings will be matched with its matching building as early as possible, which effectively improves the load data matching efficiency, saves the load data matching time, and thus improves the efficiency of charging and discharging power adjustment of buildings in the park, which is conducive to high-efficiency park energy storage optimization.

[0028] During the data matching process, this invention also considers that the number of matches between different types of buildings and the first-order building may be equal, which would interfere with the sorting. Therefore, by further constructing the cross-correlation function between buildings and analyzing the time shift parameters, this invention plans how to sort the building numbers with equal matching numbers in the event of such interference, ensuring the smooth sorting of building matching under interference conditions, further improving data matching efficiency and saving data matching time. Attached Figure Description

[0029] Figure 1 This is a flowchart illustrating a method for optimizing park energy storage based on multi-source intelligent scheduling according to the present invention. Detailed Implementation

[0030] like Figure 1 As shown, a method for optimizing park energy storage based on multi-source intelligent scheduling is provided, including:

[0031] S1: Collect information on building type and energy storage related information within the park: Collect information on building type, load data and photovoltaic data of all buildings within the park. Photovoltaic data refers to photovoltaic power.

[0032] S2: Plan the sorting method for the buildings to be matched, and generate a sorting sequence of objects to be matched in the current park: Randomly number all buildings in the current park according to their type. Buildings of the same type have the same number. Generate the building number sequence of the current park as {1,2,...,n}, where n represents the total number of different types of buildings in the current park. Count the number of buildings with different numbers in the current park, and select the building with the most numbers as the first-order building. The number of the first-order building is the first-order number. Collect the set of the number of matches between the buildings with the same type as the remaining n-1 building types and the buildings with the same type as the first-order building, which is F={F1,F2,...,F... n-1 F1 indicates that a total of F1 buildings of the same type as the first-order building have been matched with the remaining buildings of the first type. Matching between buildings refers to matching their load data. The query set F checks if there are any equal values: if not, it checks for any sorting interference, arranges the remaining n-1 building numbers in descending order of values ​​in set F, and then adds the first-order number to the sorting queue, generating the sorted sequence of objects to be matched in the current park as {A1, A2, ..., A...}. n}, A1 represents the first-order number, and A2 represents the building number with the highest probability of successfully matching the first-order building among the remaining n-1 types of buildings; if it exists, it is determined that there are sorting interference items, the priority order of sorting interference items is planned, and the sorting sequence of objects to be matched in the current park is generated after the planning is completed.

[0033] The method for judging the matching of load data is as follows: collect the load powers of two buildings in the way that the initial collection time is the same and the collection time interval is the same, generate two load power sequences W1 and W2. The lengths of the two load power sequences are equal and the timestamps are aligned with each other. Judge whether the load data of the two buildings match according to the following formula:

[0034] ;

[0035] where P represents the negative correlation degree between the loads of the two buildings, Cov(W1, W2) represents the covariance between the load power sequences W1 and W2, represents the standard deviation of the sequence W1, represents the standard deviation of the sequence W2. Set the negative correlation degree threshold as R1, where R1 is a negative number. If P < R1, judge that the load data of the two buildings match; otherwise, judge that the load data of the two buildings do not match;

[0036] If there are sorting interference items, the planning method for the priority order of the sorting interference items is as follows: screen out k buildings with the same number of previous matches as the building of the first order. Take the building with the same type as the building of the first order as the reference building. For each building, there are x buildings that are matched with the reference building one by one, and there are a total of x * k historical matching building pairs. Collect the information of x historical matching building pairs corresponding to a randomly selected building: obtain the load power sequences of the two buildings belonging to the i-th building pair as C(m) and D(m) respectively, and construct the cross-correlation function H: , m represents the integer index, y represents the time shift parameter, M represents the length of the two load power sequences. According to the formula calculate the time shift parameter y when the cross-correlation function H reaches the maximum value （max,i） , obtain the time shift parameter set Y = {y （max,1） , y （max,2） ,..., y （max,i） ,..., y （max,x）}, y （max,x） represents the time shift parameter of the x-th building pair corresponding to a randomly selected building. Calculate the average value of the parameters in the set Y to obtain the comprehensive time shift parameter between a randomly selected building and the reference building. Calculate and compare the comprehensive time shift parameters between the k buildings and the reference building, and arrange the numbers of the k buildings in ascending order of the comprehensive time shift parameters. The number of the building with the smallest comprehensive time shift parameter is in the first priority order among the k numbers;

[0037] Arrange the numbers of the buildings in descending order of the values in the set F, add the arranged numbers of the k buildings to the queuing queue, and generate the sorting sequence of the to-be-matched objects in the current park;

[0038] Example 1: If there are 5 different types of buildings in the current park, with building numbers {1,2,3,4,5} and the number of buildings of the 5 different types respectively {4,2,3,1,6}, the building with the most buildings is the 5th type. The 5th type is taken as the first-order building. The set of the number of buildings that match the 1st, 2nd, 3rd, and 4th types and the 5th type is collected as F={15,5,12,3}, where 15 represents the number of buildings that match the 1st type and the 5th type. If no equal value is found in the set F, the numbers of the 1st, 2nd, 3rd, and 4th types of buildings are arranged in descending order of the values ​​in F. Then the first-order number is added to the sorting queue, generating the sorting sequence of the objects to be matched in the current park as {5,1,3,2,4}.

[0039] Example 2: If there are 5 different types of buildings in the current park, with building numbers {1,2,3,4,5} and the number of buildings of the 5 different types are {4,2,3,1,6} respectively, the building with the most buildings is the 5th type. Taking the 5th type as the first-order building, collect the set of previous matches between buildings of the same type as the 1st, 2nd, 3rd, and 4th types and the 5th type, which is F={15,5,12,12}. If there are equal values ​​in set F, it is determined that there is a sorting interference item. The 3rd and 4th types of buildings are selected. Both the 3rd and 4th types of buildings have 12 previous matches with the 5th type, x=12. Obtain the corresponding number of buildings of the 3rd type... The load power sequence of a building pair is obtained by constructing a cross-correlation function to obtain the time-shift parameters of 12 building pairs corresponding to the third type of building. The average time-shift parameters of the 12 building pairs are then used to obtain the comprehensive time-shift parameter 'a' between the third type of building and the benchmark building. The comprehensive time-shift parameter 'b' between the fourth type of building and the benchmark building is obtained in the same way. Comparing 'a' and 'b', we find that 'a' > 'b'. The numbers of the third and fourth types of buildings are arranged in ascending order of comprehensive time-shift parameters to obtain {4,3}. The building numbers are then arranged in descending order of values ​​in set F. {4,3} is added to the queuing queue, generating the sorted sequence of objects to be matched in the current park as: {5,1,4,3,2}.

[0040] S3: After generating the sorted sequence of objects to be matched in the current park, perform building pair matching to generate matching building pairs and unmatched buildings: Match all first-order buildings with the remaining buildings one by one according to the sorted sequence. If there is a building that matches a random first-order building, the two matching buildings are considered a matching building pair; if a random first-order building does not match a building, the corresponding first-order building is considered an unmatched building. Remove the matching building pairs and unmatched buildings from the objects to be matched. After completing the matching of all first-order buildings, match all second-order buildings with the remaining buildings one by one according to the sorted sequence. After completing the matching of all second-order buildings, match the third-order buildings, until only the nth-order building remains in the objects to be matched. All remaining nth-order buildings are considered unmatched buildings. Since all remaining nth-order buildings are of the same type, buildings of the same type are not matched with each other. Therefore, all remaining nth-order buildings are considered unmatched buildings.

[0041] S4: Different energy storage data prediction methods are planned for matched building pairs and unmatched buildings, and charging and discharging optimization is performed on the energy storage equipment in the park: Coordinated data prediction and charging / discharging optimization are performed for matched building pairs: The load data of a randomly matched building pair is linearly summed, and the photovoltaic data of the corresponding matched building pair is linearly summed. The linearly summed load data and photovoltaic data are input into an LSTM model. The LSTM model is used to predict the load data and photovoltaic data at future times. The LSTM model outputs the load data and photovoltaic data at future times. The time interval between load data at different future times is T, and the time interval between photovoltaic data at different future times is equal to T. The load data, photovoltaic data, and SOC at future times are input into the MPC model. SOC represents the state of charge. The MPC model needs to ensure that the energy storage equipment is within a safe range. In the internal operation, SOC is the direct basis for preventing battery overcharging and over-discharging. Therefore, inputting SOC into the MPC model can avoid safety risks such as battery damage or thermal runaway caused by deep charging and discharging of energy storage devices. The MPC model outputs the optimal charging and discharging power sequence of energy storage devices {Pess(1),Pess(2),...,Pess(j)}, where j represents the number of parameters in the optimal charging and discharging power sequence output by the MPC model, and Pess(1) represents the predicted optimal charging and discharging power in the time period [t,t+T], where t represents the current time. If Pess(1)>0, the discharge power of the energy storage device of the corresponding matching building pair will be increased in the time period [t,t+T]. If Pess(1)=0, the charging and discharging power of the energy storage device of the corresponding matching building pair will not be adjusted. If Pess(1)<0, the charging power of the energy storage device of the corresponding matching building pair will be increased in the time period [t,t+T].

[0042] For each building that fails to match, data prediction and charge / discharge optimization are performed separately: the load data and photovoltaic data of a randomly selected building that fails to match are input into the LSTM model, and the LSTM model outputs the load data and photovoltaic data for future time moments. The load data, photovoltaic data, and SOC for future time moments are input into the MPC model, and the MPC model outputs the optimal charge / discharge power sequence of the energy storage device. The charge / discharge power of the energy storage device for the corresponding building that fails to match is adjusted according to the optimal charge / discharge power sequence. The adjustment method is the same as the adjustment method for the matching building pair.

[0043] Example 3: If the output of the MPC model is Pess(1)=5 and Pess(1)>0, the discharge power of the energy storage device of the corresponding matching building pair will be increased to 5kW during the time period [t,t+T].

Claims

1. A method for optimizing park energy storage based on multi-source intelligent scheduling, characterized in that: The method includes: S1: Collect information on the type of buildings and energy storage-related information within the park; S2: Plan the sorting method of the buildings to be matched and generate the sorting sequence of the objects to be matched in the current park; S3: After generating the sorted sequence of objects to be matched in the current park, perform building pair matching to generate matching building pairs and buildings that failed to match. S4: Plan different energy storage data prediction methods for matching building pairs and buildings that fail to match, and optimize the charging and discharging of energy storage equipment in the park; S2 includes: randomly numbering all buildings in the current park according to their type, with buildings of the same type having the same number, generating a building number sequence of the current park as {1,2,...,n}, where n represents the total number of n different types of buildings in the current park; counting the number of buildings with different numbers in the current park; selecting the building with the most numbers as the first-order building, and assigning it the first-order number; and collecting the set of numbers matching the buildings of the same type as the remaining n-1 building types and the buildings of the same type as the first-order building, which is F={F1,F2,...,F...}. n-1 F1 indicates that a total of F1 buildings of the same type as the first-order building have been matched with the remaining buildings of the first type. Matching between buildings refers to matching their load data. The query set F checks if there are any equal values: if not, it checks for any sorting interference, arranges the remaining n-1 building numbers in descending order of values ​​in set F, and then adds the first-order number to the sorting queue, generating the sorted sequence of objects to be matched in the current park as {A1, A2, ..., A...}. n }, A1 represents the first-order number, and A2 represents the building number with the highest probability of successfully matching the first-order building among the remaining n-1 types of buildings; if it exists, it is determined that there are sorting interference items, the priority order of sorting interference items is planned, and the sorting sequence of objects to be matched in the current park is generated after the planning is completed.

2. The method for optimizing park energy storage based on multi-source intelligent scheduling according to claim 1, characterized in that: The method for determining whether the load data matches is as follows: Load power data for two buildings are collected at the same initial collection time and with the same collection time interval, generating two load power sequences W1 and W2. The two load power sequences have equal lengths and their timestamps are aligned. The following formula is used to determine whether the load data of the two buildings match: ； Among them, P represents the negative correlation degree between two building loads, Cov(W1, W2) represents the covariance between the load power sequences W1 and W2, represents the standard deviation of the sequence W1, represents the standard deviation of the sequence W2. Set the negative correlation degree threshold as R1, where R1 is a negative number. If P < R1, it is judged that the load data of the two buildings match each other; otherwise, it is judged that the load data of the two buildings do not match.

3. The method for optimizing park energy storage based on multi-source intelligent scheduling according to claim 1, characterized in that: If there are ranking interference items, the priority order of these interference items is planned as follows: Select k types of buildings with the same number of previous matches as the first-order building type. Use these buildings as the baseline buildings. Each type of building has x buildings that match the baseline buildings one-to-one, resulting in x*k historical matching building pairs. Collect information on x historical matching building pairs corresponding to a random type of building. Obtain the load power sequences of the two buildings belonging to the i-th building pair, C(m) and D(m), respectively. Construct the cross-correlation function H: Where m represents the integer index, y represents the time shift parameter, and M represents the length of the two load power sequences, according to the formula... Calculate the time-shift parameter y that maximizes the cross-correlation function H. （max,i） The time-shift parameter set Y={y （max,1） ,y （max,2） ,...,y （max,i） ,...,y （max,x） }, y （max,x） Let Y represent the time shift parameter of the x-th building pair corresponding to a random building. The average value of the parameters in set Y is used to obtain the comprehensive time shift parameter between the random building and the benchmark building. The comprehensive time shift parameters between k buildings and the benchmark building are calculated and compared. The numbers of the k buildings are arranged in ascending order of comprehensive time shift parameter. The number of the building with the smallest comprehensive time shift parameter is placed in the first priority order among the k numbers. Arrange the building numbers in descending order of the values ​​in set F, add the arranged k building numbers to the queue, and generate the sorted sequence of objects to be matched in the current park.

4. The park energy storage optimization method based on multi-source intelligent scheduling according to claim 3, characterized in that: S3 includes: matching all first-order buildings with the remaining buildings one by one according to the sorting sequence; if there is a building that matches a random first-order building, the two matching buildings are considered a matching building pair; if a random first-order building does not find a matching building, the corresponding first-order building is considered a failed match building, and the matching building pair and the failed match building are removed from the matching objects. After completing the matching of all first-order buildings, matching all second-order buildings with the remaining buildings one by one according to the sorting sequence is performed. After completing the matching of all second-order buildings, the matching of third-order buildings is performed, until only the nth-order building remains in the matching objects, and all the remaining nth-order buildings are considered failed match buildings.

5. The park energy storage optimization method based on multi-source intelligent scheduling according to claim 4, characterized in that: S4 includes: performing collaborative data prediction and charge / discharge optimization for matched building pairs: linearly summing the load data of a random matched building pair, linearly summing the photovoltaic data of the corresponding matched building pair, inputting the linearly summed load data and photovoltaic data into an LSTM model, using the LSTM model to predict the load data and photovoltaic data at future times, and outputting the load data and photovoltaic data at future times. The time interval between load data at different future times is T, and the time interval between photovoltaic data at different future times is equal to T. The load data, photovoltaic data, and SOC at future times are input into an MPC model, and the MPC model outputs the energy storage data. The optimal charge / discharge power sequence of the device is {Pess(1),Pess(2),...,Pess(j)}, where j represents the number of parameters in the optimal charge / discharge power sequence output by the MPC model, and Pess(1) represents the predicted optimal charge / discharge power in the time period [t,t+T], where t represents the current time. If Pess(1)>0, the discharge power of the energy storage device of the corresponding matching building pair will be increased in the time period [t,t+T]. If Pess(1)=0, the charge / discharge power of the energy storage device of the corresponding matching building pair will not be adjusted. If Pess(1)<0, the charging power of the energy storage device of the corresponding matching building pair will be increased in the time period [t,t+T].

6. The park energy storage optimization method based on multi-source intelligent scheduling according to claim 5, characterized in that: For each building that fails to match, data prediction and charge / discharge optimization are performed individually: the load data and photovoltaic data of a randomly selected building that fails to match are input into the LSTM model, and the LSTM model outputs the load data and photovoltaic data for future time moments. The load data, photovoltaic data, and SOC for future time moments are input into the MPC model, and the MPC model outputs the optimal charge / discharge power sequence of the energy storage device. The charge / discharge power of the energy storage device for the corresponding building that fails to match is adjusted according to the optimal charge / discharge power sequence, and the adjustment method is the same as that for the matching building pair.

7. A park energy storage optimization system based on multi-source intelligent scheduling, characterized in that: The system includes: The system includes a park information collection module, a building sorting module, a building matching module, and an energy storage optimization module. The park information collection module collects information on the type of buildings in the park, load data of all buildings in the park, and photovoltaic data. The sorting module for buildings to be matched plans the sorting method for the buildings to be matched and generates a sorting sequence of objects to be matched in the current park. The building matching module performs building pair matching, generating matched building pairs and buildings that failed to match. The energy storage optimization module plans different energy storage data prediction methods for matched building pairs and unmatched buildings, and optimizes the charging and discharging of the park's energy storage equipment. All buildings in the current park are randomly numbered according to their type, with buildings of the same type having the same number. This generates a building number sequence {1, 2, ..., n}, where n represents the number of different building types in the current park. The number of buildings with different numbers is counted, and the building with the most numbers is selected as the first-order building. The number of the first-order building is then assigned as its first-order number. The set of matching numbers between buildings of the same type as the remaining n-1 building types and buildings of the same type as the first-order building is collected, which is F = {F1, F2, ..., F...}. n-1 F1 indicates that a total of F1 buildings of the same type as the first-order building have been matched with the remaining buildings of the first type. Matching between buildings refers to matching their load data. The query set F checks if there are any equal values: if not, it checks for any sorting interference, arranges the remaining n-1 building numbers in descending order of values ​​in set F, and then adds the first-order number to the sorting queue, generating the sorted sequence of objects to be matched in the current park as {A1, A2, ..., A...}. n }, A1 represents the first-order number, and A2 represents the building number with the highest probability of successfully matching the first-order building among the remaining n-1 types of buildings; if it exists, it is determined that there are sorting interference items, the priority order of sorting interference items is planned, and the sorting sequence of objects to be matched in the current park is generated after the planning is completed.

8. A park energy storage optimization system based on multi-source intelligent scheduling according to claim 7, characterized in that: The building sorting module to be matched includes a sorting unit with interference items and a sorting unit without interference items. The interference item sorting unit is used to plan the priority order of sorting interference items when it is determined that there are sorting interference items, and then add the planned sorting interference items to the queuing queue to generate the sorting sequence of objects to be matched in the current park. The interference-free sorting unit is used to directly arrange the building numbers according to the number of matching items between buildings when it is determined that there are no interference items, thereby generating a sorting sequence of objects to be matched in the current park.

9. A park energy storage optimization system based on multi-source intelligent scheduling according to claim 8, characterized in that: The energy storage optimization module includes a collaborative energy storage optimization unit and a separate energy storage optimization unit; The collaborative energy storage optimization unit is used to perform collaborative charging and discharging regulation of the energy storage devices of the matching building pairs after data prediction of the matching building pairs. The individual energy storage optimization unit is used to perform individual charge and discharge regulation of the energy storage devices of each building that failed to match data prediction.

Images