A Photovoltaic-Storage Cooperative Optimization Scheduling Method and System Based on Multidimensional Features

Abstract

This invention relates to the field of power system operation and control technology, and particularly to a method and system for coordinated optimization scheduling of photovoltaic (PV) and energy storage systems based on multi-dimensional features. The method includes: real-time acquisition of the ideal grid-connected power of each PV and energy storage system at the next moment; triggering coordinated scheduling when the total output exceeds the distribution network overload threshold; extracting power fluctuation, regulation margin characteristics, and output power sequences based on historical operating data; identifying strong positive and negative correlations between devices; and obtaining synchronous and complementary coordinated features. This invention constructs a particle swarm optimization model, adaptively adjusts the individual learning factors of the algorithm based on coordinated features, constructs a fitness function combined with regulation margin, iteratively solves for the optimal grid-connected power under grid security constraints, generates scheduling commands, and issues them to the local controllers of each PV and energy storage system, achieving safe and coordinated optimized operation of the cluster. This invention, through multi-feature adaptive coordinated scheduling of PV and energy storage, avoids the grid security risks caused by multi-machine synchronous grid connection and improves the scheduling security of PV and energy storage clusters.

Description

Technical Field

[0001] This invention relates to the field of power system operation and control technology. In particular, it relates to a photovoltaic-storage collaborative optimization scheduling method and system based on multi-dimensional features. Background Technology

[0002] With the rapid development of new energy technologies, the coordinated and optimized scheduling of photovoltaic and energy storage systems has become a key means to improve the distribution network's ability to absorb distributed energy. By rationally configuring and scheduling photovoltaic and energy storage systems, the volatility of new energy output can be effectively mitigated, and power supply reliability can be improved.

[0003] A Chinese invention patent application with publication number CN114036451A discloses an energy storage control method and system for a grid-connected photovoltaic-storage-charging device. This application mainly collects local operating data, meteorological data, and historical power grid data to autonomously decide the charging and discharging sequence of energy storage, thereby realizing the spatiotemporal transfer of electrical energy and playing a basic role in peak shaving and valley filling to a certain extent.

[0004] However, the aforementioned existing technologies primarily employ independent, autonomous local control modes for each photovoltaic-storage system, lacking a cluster-level collaborative scheduling mechanism. In practical applications, especially during peak power output periods with ample sunlight, this independent control mode easily leads to multiple photovoltaic-storage systems simultaneously connecting to the grid for power sales. This synchronization can cause reverse power overload in the distribution network, subsequently inducing risks such as voltage exceeding limits, increased line losses, and relay protection malfunctions, severely impacting the safe and stable operation of the distribution network. Summary of the Invention

[0005] To address the problem that existing technologies, due to a lack of cluster coordination, can easily lead to multiple photovoltaic and energy storage systems simultaneously selling electricity during peak hours, causing distribution network overload and voltage exceeding risks, thus threatening grid safety, this invention provides solutions in the following aspects.

[0006] In the first aspect, the photovoltaic-storage collaborative optimization scheduling method based on multi-dimensional features includes: real-time acquisition of the ideal grid-connected power of multiple photovoltaic-storage systems under the distribution network; when the sum of the ideal grid-connected power of multiple photovoltaic-storage systems at the next moment exceeds a preset distribution network overload threshold, determining that there is a power overload risk and triggering collaborative optimization scheduling, and acquiring historical operating data of each photovoltaic-storage system to be allocated; based on the historical operating data, extracting the grid-connected power fluctuation characteristics, adjustment margin characteristics, and historical output power sequences of each photovoltaic-storage system to be allocated; defining strong positive and strong negative correlations between each photovoltaic-storage system based on the historical output power sequences, extracting a first collaborative feature characterizing the system cluster synchronization characteristics based on the strong positive correlation, and extracting a second collaborative feature characterizing the output complementarity characteristics based on the strong negative correlation; constructing a particle swarm optimization model, using the grid-connected power output of each photovoltaic-storage system at the next moment as the particle dimension, and constructing a fitness function based on the adjustment margin characteristics; calculating the adjustment suppression coefficient using the first and second collaborative features, and performing dimensional adaptive adjustment of the individual learning factors of the particle swarm algorithm based on the adjustment suppression coefficient, and based on the updated individual learning factors and The fitness function iteratively optimizes the particles. Under the premise of satisfying the total power safety constraint of the distribution network, the globally optimal particle position of the optimization result is taken as the optimal output grid-connected power of each photovoltaic and energy storage system at the next moment. The optimal output grid-connected power is used to generate scheduling instructions and sent to the local controllers of each photovoltaic and energy storage system. The first cooperative feature specifically includes: constructing a synchronous association set based on the strong positive correlation between each photovoltaic and energy storage system to be allocated; wherein, any two photovoltaic and energy storage systems to be allocated within the same synchronous association set have a direct or indirect strong positive correlation, and photovoltaic and energy storage systems to be allocated between different synchronous association sets do not have the strong positive correlation; calculating the sum of the adjustment margin features of all photovoltaic and energy storage systems in the synchronous association set to obtain the total adjustment margin of the synchronous association set, and taking the proportion of the adjustment margin feature of each photovoltaic and energy storage system as the first cooperative feature; the second cooperative feature specifically includes: taking any photovoltaic and energy storage system to be allocated as the target system, counting the total number of all correlations of the target system; calculating the proportion of the number of strong negative correlations of the target system to the total number of all correlations of the target system as the second cooperative feature of the photovoltaic and energy storage system to be allocated.

[0007] Preferably, after determining the existence of power overload risk and triggering collaborative optimization scheduling, and before obtaining the historical operating data of each photovoltaic-storage system to be allocated, the method further includes: The photovoltaic and energy storage systems in the distribution network whose ideal grid-connected power is less than or equal to 0 at the next moment are eliminated, and the eliminated photovoltaic and energy storage systems are taken as photovoltaic and energy storage systems to be allocated.

[0008] Preferably, the extraction of grid-connected power fluctuation characteristics of a single photovoltaic-storage system specifically includes: Taking any photovoltaic-storage system to be allocated as the target system, the historical grid-connected power sequence of the target system is traversed, and the grid-connected power change between adjacent sampling times is calculated sequentially; weighting coefficients are configured according to the sign difference of the grid-connected power product at adjacent times, and the weighted operation of each grid-connected power change is performed and the arithmetic mean is calculated; the arithmetic mean is mapped to obtain the grid-connected power fluctuation characteristics of the target system.

[0009] Preferably, the step of calculating the adjustment margin characteristic includes: Taking any photovoltaic-storage system to be allocated as the target system, obtain the upper and lower limits of the energy storage state of charge (SBC) of the target system, as well as the current SBC, energy storage health status, and grid-connected power fluctuation characteristics. Based on the upper and lower limits of the SBC and the current SBC of the target system, calculate the ratio of the remaining charging capacity to the total adjustable energy storage capacity. Using the ratio, the grid-connected power fluctuation characteristics, and the energy storage health status, calculate the adjustment margin characteristics of the target system at the current moment.

[0010] Preferably, the strong positive correlation and the strong negative correlation include: Traverse the historical grid-connected power sequence of each photovoltaic and energy storage system to be allocated. For the grid-connected power at any moment in the historical grid-connected power sequence, if the grid-connected power is greater than 0, it is determined to be positive output power and the original value is retained. Otherwise, if it is less than or equal to 0, the grid-connected power value at the corresponding moment is set to zero, thus obtaining the historical output power sequence of each photovoltaic and energy storage system to be allocated. The photovoltaic and energy storage systems within the photovoltaic and energy storage system cluster to be allocated are combined in pairs, and the correlation values ​​between the historical output power sequences of the two photovoltaic and energy storage systems in each pair are calculated based on cosine similarity. Based on the sign of the correlation values, all combinations are divided into a positive correlation set or a negative correlation set. All correlation values ​​in the positive correlation set are sorted in ascending order, and the value at the corresponding position of the preset quantile is selected as the positive correlation threshold. All correlation values ​​in the negative correlation set are sorted in descending order, and the value at the corresponding position of the preset quantile is selected as the negative correlation threshold. If the correlation value is greater than the positive correlation threshold, a strong positive correlation is determined between the two photovoltaic-storage systems in the corresponding photovoltaic-storage system combination; otherwise, if the correlation value is less than or equal to the positive correlation threshold, no analysis is performed. If the correlation value is less than the negative correlation threshold, a strong negative correlation is determined between the two photovoltaic-storage systems in the corresponding photovoltaic-storage system combination; otherwise, if the correlation value is greater than or equal to the negative correlation threshold, no analysis is performed.

[0011] Preferably, the calculation of the modulation inhibition coefficient includes: Taking any photovoltaic-storage system to be allocated as the target system, determine whether there is a strong positive correlation or a strong negative correlation between the target systems; If it exists, the average of the sum of the first and second cooperative features of the target system is calculated as the regulation inhibition coefficient; If it does not exist, the target system is determined to be an isolated system, and the regulation margin characteristic of the target system minus 1 is taken as the regulation inhibition coefficient.

[0012] Preferably, the construction of the fitness function based on the adjustment margin feature includes: Taking any one of the photovoltaic-storage systems in the cluster to be allocated as the target system, the adjustment margin characteristics of the target system are exponentially mapped, and the mapping result is used as the adjustment cost coefficient. Calculate the difference between the ideal grid-connected power of the target system at the next time step and the actual grid-connected power output at the next time step, and use the ratio of the difference to the ideal grid-connected power at the next time step as the degree of deviation of the target system. The product of the adjustment cost coefficient and the degree of deviation is used as the individual scheduling cost of the target system. The individual scheduling costs of all optical storage systems are accumulated to obtain the overall scheduling cost, which is then used as the fitness function of a particle.

[0013] Secondly, a multi-dimensional feature-based photovoltaic-storage collaborative optimization scheduling system includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned multi-dimensional feature-based photovoltaic-storage collaborative optimization scheduling method is implemented.

[0014] The present invention has the following effects: 1. This invention mines historical operating data of photovoltaic and energy storage systems to quantify the individual characteristics of power output fluctuations and adjustment margins. By combining cosine similarity and adaptive quantile thresholds, it accurately identifies the synchronous and complementary coupling characteristics between devices, and triggers cluster collaborative scheduling in a targeted manner. This suppresses the impact of multiple devices synchronizing and connecting to the grid from the source, effectively solves the problem of power flow runaway caused by traditional single-machine control, and significantly improves the operational stability and security of high-proportion photovoltaic and energy storage connected to the distribution network.

[0015] 2. This invention calculates the regulation suppression coefficient by considering grid connection fluctuation characteristics, regulation margin characteristics, and synchronization and complementary coordination characteristics. It performs dimensional adaptive optimization of the individual learning factors of the particle swarm algorithm, enabling differentiated parameter correction and output adjustment based on the operating conditions, fluctuation characteristics, and cluster coupling relationships of different photovoltaic and energy storage devices. This effectively overcomes the shortcomings of traditional optimization algorithms, such as fixed parameters and poor adaptability, and significantly improves the accuracy and scenario adaptability of photovoltaic and energy storage cluster collaborative scheduling.

[0016] 3. This invention constructs a scheduling cost fitness function with adjustment margin as its core, comprehensively considering factors such as energy storage regulation losses and output deviation costs. Under the premise of strictly meeting the total power safety constraints of the distribution network and equipment output limits, it iteratively seeks the optimal grid-connected power output scheme with the goal of minimizing the global scheduling cost. It not only strengthens the safe operation boundary of the power grid through cluster collaborative regulation and avoids overload risks, but also prioritizes the use of high-margin, low-loss, and low-fluctuation photovoltaic and energy storage equipment for regulation, effectively reducing energy storage operation losses and the overall scheduling cost of the cluster. This achieves a multi-dimensional optimal balance of grid safety, equipment controllability, and economic efficiency, significantly improving the comprehensive benefits of distributed photovoltaic and energy storage cluster grid-connected operation. Attached Figure Description

[0017] Figure 1 This is a flowchart of steps S1-S4 in the optical-storage collaborative optimization scheduling method based on multi-dimensional features according to an embodiment of the present invention.

[0018] Figure 2 This is a structural block diagram of the optical-storage collaborative optimization scheduling system based on multi-dimensional features according to an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0020] Reference Figure 1 The photovoltaic-storage collaborative optimization scheduling method based on multi-dimensional features includes steps S1-S4, as detailed below: S1: Real-time acquisition of grid-connected power of multiple photovoltaic and energy storage systems under the distribution network. When the sum of the grid-connected power of multiple photovoltaic and energy storage systems exceeds the preset distribution network overload threshold, it determines that there is a power overload risk and triggers collaborative optimization scheduling to acquire the historical operating data of each photovoltaic and energy storage system to be allocated.

[0021] The system acquires the operating parameters of the distribution network and the ideal grid-connected power of the photovoltaic and energy storage systems. It obtains the maximum allowable total overload power threshold of the distribution network under the current operating conditions. Simultaneously, it identifies all distributed photovoltaic and energy storage systems connected to the distribution network and, using the current time as a reference, acquires the ideal grid-connected power output by the local controller of each system at the next moment, exemplarily every 30 seconds, according to a preset acquisition frequency. The ideal grid-connected power is calculated and generated by the built-in autonomous charging and discharging control strategy of each photovoltaic and energy storage system. The specific implementation process of this strategy can be found in Chinese invention patent application CN114036451A, entitled "Energy Storage Control Method and System for Grid-Connected Photovoltaic and Energy Storage Charging Device."

[0022] First, the maximum total overload power threshold allowed for safe operation of the distribution network under the current operating conditions is obtained in advance; based on the current sampling time... Using this as the reference time, data is collected from the local controllers of all distributed photovoltaic and energy storage systems within the distribution network. Ideal grid-connected power at all times. In the unified physical conventions for power measurement in distribution networks, the active power delivered by the photovoltaic-storage system to the distribution network is defined as the positive direction of grid-connected power; based on this positive / negative determination rule, [the following is selected]... All photovoltaic (PV) and energy storage (ESS) units with an ideal grid-connected power value greater than 0 at any given time are selected and collectively form a cluster of PV and ESS systems to be allocated. For each PV and ESS system within this cluster... The ideal grid-connected power is calculated cumulatively to obtain the desired grid-connected power of the cluster to be allocated.

[0023] The calculated desired grid-connected power of the photovoltaic and energy storage clusters is compared with the aforementioned total overload power threshold of the distribution network. If the desired grid-connected power exceeds the total overload power threshold, it indicates that if all photovoltaic and energy storage systems fully implement their local autonomous charging and discharging strategies, the total active power output synchronously transmitted to the distribution network will exceed the safe carrying capacity of the distribution network lines. This could easily lead to safety hazards such as reverse power overload, node voltage exceeding limits, a significant increase in line active power loss, and relay protection malfunctions. Therefore, it is determined that the distribution network faces a power overload risk, and the multi-photovoltaic-energy storage cluster collaborative optimization scheduling process of this invention is triggered. If the desired grid-connected power is less than or equal to the total overload power threshold of the distribution network, it means that each photovoltaic and energy storage system will not exceed the power safety constraints of the distribution network when operating according to its local autonomous control strategy. The distribution network is in a safe operating range, and collaborative optimization scheduling does not need to be initiated. All photovoltaic and energy storage systems can directly operate according to their own generated power output. The ideal grid-connected power output for charging and discharging is always achieved.

[0024] When a coordinated optimization scheduling is triggered, the current time is used as the last sampling time. Following a preset sampling window (exemplarily including the past 100 sampling times), the historical grid-connected power sequence, state of charge (SOC), and state of health (SOH) of each photovoltaic-storage system within the cluster to be allocated are collected. The historical grid-connected power sequence characterizes the recent power output fluctuations of each system; SOC represents the ratio of current remaining power to the rated energy storage capacity, used to assess the battery's charge / discharge margin; and SOH characterizes the battery's health level, with a value range of [value missing]. A higher value indicates a better battery health condition, used to assess the battery's safety boundaries and response capabilities in subsequent scheduling. Simultaneously, the grid-connected power at each moment in the historical grid-connected power sequence is screened for positive output effectiveness: specifically, it is determined whether the grid-connected power at that moment is greater than 0. If it is greater than 0, it indicates that the photovoltaic-storage system is actually supplying power to the grid at that moment, and is considered positive output power and retained. If it is less than or equal to 0, it indicates that the photovoltaic-storage system is not outputting power, is in standby, or is absorbing power from the grid at that moment, and is considered non-positive output and eliminated. Finally, based on the above screening results, the photovoltaic-storage systems with retained positive output power data are ultimately determined as the photovoltaic-storage system clusters to be allocated for subsequent collaborative scheduling optimization.

[0025] S2: Based on historical operating data, extract the grid-connected power fluctuation characteristics and adjustment margin characteristics of each photovoltaic-storage system to be allocated; define strong positive and strong negative correlations between each photovoltaic-storage system based on historical output power sequences; extract the first collaborative feature characterizing the system cluster synchronization characteristics based on the strong positive correlation, and extract the second collaborative feature characterizing the output complementarity characteristics based on the strong negative correlation.

[0026] For any photovoltaic-storage unit within the photovoltaic-storage system cluster to be allocated, the target system is selected, and the complete historical grid-connected power sampling sequence of that target system is retrieved. The sequence contains continuous... The grid-connected power values ​​corresponding to each time-series sampling point are obtained. The historical grid-connected power sequence is traversed sequentially according to the sampling time sequence, and the grid-connected power difference corresponding to each group of adjacent sampling times in the sequence is solved in turn to obtain the grid-connected power change amplitude for each time period.

[0027] To differentiate the degree of danger posed by different fluctuation conditions to system operation, weighting coefficients are configured differently based on the sign of the product of grid-connected power values ​​at two adjacent sampling times: if the product of grid-connected power values ​​at adjacent times is less than 0, it means that the grid-connected power transmission direction has switched between positive and negative during that period, which is a severe and high-risk fluctuation condition, and this group of changes is given a higher weight; if the product of grid-connected power values ​​at adjacent times is greater than or equal to 0, it means that the power transmission direction fluctuates slightly during that period, without any reversal of power flow direction, and this group of changes is given a lower weight.

[0028] The absolute values ​​of the grid-connected power change amplitude at each adjacent time point are weighted using corresponding weighting coefficients. The sum of all weighted power change amplitudes is then divided by the total number of effective time periods to obtain the weighted arithmetic mean of the power change. Finally, the arithmetic mean is normalized using the hyperbolic tangent function to constrain the calculation result to... Within the range, the grid-connected power fluctuation characteristics that can quantify the severity of the power output fluctuation of the target system are obtained.

[0029] Specifically, the grid-connected power fluctuation characteristics satisfy the following relationship: ; In the formula, This represents the grid-connected power fluctuation characteristics of the target system, with a value range of... ; Represents the historical grid-connected power sequence of the target system. Grid-connected power at each sampling time; Represents the historical grid-connected power sequence of the target system. The grid-connected power at each sampling time; This represents the total number of sampling times in the historical grid-connected power sequence; Indicates the first The weighting coefficients at each sampling time, that is, when When the grid-connected power undergoes a switch between positive and negative flow directions, it can easily cause severe impacts on the power flow and voltage of the distribution network, which is considered a high-risk fluctuation condition. ;when The power transmission direction did not reverse; there were only small power fluctuations in one direction, resulting in a lower risk of disturbance. .

[0030] Select one photovoltaic-storage unit from the photovoltaic-storage system cluster to be allocated as the target system, and retrieve the upper limit of the energy storage state of charge (SBC) of the target energy storage device as specified by the factory calibration and operational constraints. Energy storage state of charge limit Simultaneously collect current data. The target system's real-time monitored state of charge and battery health status, along with the grid-connected power fluctuation characteristics obtained in previous steps, are considered. Based on the upper and lower limits of the energy storage state of charge (SOC) and the current real-time SOC values, the ratio of the remaining rechargeable capacity of the energy storage to the overall adjustable capacity of the energy storage is calculated. This ratio directly represents the current charging output potential of the energy storage. Then, the ratio of the remaining rechargeable capacity, the grid-connected power fluctuation characteristics (representing the level of output disturbance), and the energy storage health status (representing the degree of battery degradation) are coupled and calculated together. The grid-connected power fluctuation characteristics are the complement of the grid-connected power fluctuation characteristics. Finally, the current... The adjustment margin characteristics of the target system at any given time can comprehensively quantify the schedulable space of the target system, the adjustment operating cost, and the level of safety risks during the power adjustment process.

[0031] Specifically, the adjustment margin characteristic satisfies the following relationship: ; In the formula, Indicates the target system at the th The adjustment margin characteristics at any given time; Indicates the upper limit of the energy storage state of charge; Indicates the lower limit of the energy storage's state of charge; This indicates the grid connection fluctuation characteristics of the target system; Indicates the target system at the th The battery health status at any time.

[0032] fractional terms Used to quantify the relative proportion of the current remaining rechargeable capacity of energy storage within the overall adjustable capacity range of energy storage; in the current Real-time state of charge of energy storage The smaller the value, the greater the maximum energy storage distance for full charging. Remaining charging capacity The larger the value, the stronger the potential for adjusting the grid-connected power of photovoltaics by absorbing excess photovoltaic output and reducing the power of photovoltaic-storage grid connection; where... As a power fluctuation tolerance correction coefficient, it is used to characterize the system's ability to withstand superimposed power regulation operations. If the grid-connected power fluctuation characteristics... The larger the value, the more severe the historical power output fluctuations of the system, and the smaller the corresponding power fluctuation tolerance coefficient. In this case, applying additional power regulation commands will superimpose new power disturbances on top of the existing severe fluctuations, increasing the operational risks of power flow oscillations and voltage instability in the distribution network, ultimately reducing the overall regulation margin characteristics; energy storage health status. It directly reflects the degree of degradation and continuous regulation tolerance of the energy storage battery. Higher values ​​indicate better battery performance and lower long-term regulation losses. The combined relationship among these three indicators reveals that a larger remaining rechargeable energy storage capacity and better grid-connected power fluctuation characteristics... Smaller, better battery health The higher the value, the better the adjustment margin characteristic obtained. The larger the value, the more power downsizing and scheduling space the target system has, the lower energy storage regulation loss cost, and the smaller the additional operational risks when carrying out power regulation.

[0033] After quantifying the individual operational characteristics of a single photovoltaic-storage system, such as grid-connected power fluctuations and regulation margins, only the adjustable potential and operational disturbance level of a single photovoltaic-storage unit can be characterized. This cannot reflect the coordinated coupling patterns of output timing among multiple photovoltaic-storage clusters. Furthermore, the root cause of reverse power overload risk in the distribution network lies in the synchronous high-power forward power transmission from multiple photovoltaic-storage systems. Different devices exhibit synchronous output or peak-shifting complementarity coupling relationships, and these cluster-level correlation characteristics directly influence the optimization direction and adjustment priority of coordinated scheduling. Therefore, this invention, based on the individual characteristics of a single system, further explores the temporal output correlation between multiple photovoltaic-storage systems. This is achieved by constructing standardized historical output power sequences, introducing a cosine similarity metric to quantify the output correlation between pairs of devices, and using the quantile method to adaptively divide the strong and weak correlation boundaries, defining strong positive and strong negative correlations between devices within the cluster. The complete process is as follows: First, the complete historical grid-connected power time series corresponding to all photovoltaic and energy storage systems to be allocated are retrieved. For the grid-connected power values ​​at each sampling time within each series, uniform preprocessing is performed: if the grid-connected power value at a certain time is greater than 0, it means that the photovoltaic and energy storage system is positively transmitting active power to the distribution network at that time, and this value is directly retained without modification; if the grid-connected power value at a certain time is less than or equal to 0, it means that the photovoltaic and energy storage system is not transmitting power externally at that time, and will not cause reverse power overload of the distribution network. Therefore, the power value at that time is uniformly set to zero. After the above uniform processing, a unique historical output power series for each photovoltaic and energy storage system to be allocated is generated. Only the positive output data that affects the reverse overload of the distribution network is retained, and irrelevant interference items caused by charging and power receiving conditions are removed.

[0034] After sequence preprocessing, all photovoltaic and energy storage units within the photovoltaic-energy storage system cluster to be allocated are paired up in full. For any pairing, the cosine similarity algorithm is used to calculate the temporal correlation of the corresponding historical output power sequences of the two pairs. The output value range is [insert range here]. The correlation value can accurately characterize the degree of synchronization and complementarity of the historical positive output time sequence of two photovoltaic and energy storage systems.

[0035] Based on the positive and negative attributes of the correlation values ​​of each pairing, all pairing combinations are divided into two categories: positively correlated combinations and negatively correlated combinations. For the positively correlated sets, all correlation values ​​within the set are arranged in ascending order from smallest to largest, and the correlation values ​​corresponding to preset quantile points are extracted as the positive correlation judgment threshold. For example, the correlation values ​​corresponding to 60% of the preset quantile points are used as the positive correlation judgment threshold, and other quantiles such as 50% and 70% can be adjusted according to the number of photovoltaic storage devices connected to the feeder and the historical output dispersion. For the negatively correlated sets, all correlation values ​​within the set are arranged in descending order from largest to smallest, and the correlation values ​​corresponding to the same preset quantile points are selected as the negative correlation judgment threshold. The boundary between strong and weak correlations is adaptively divided using the quantile method to adapt to the differentiated output distribution characteristics of photovoltaic storage clusters in different distribution networks.

[0036] The following criteria are used to determine the correlation between paired systems: If the correlation value of a paired system is greater than the positive correlation threshold, it indicates that the historical positive output time series of the two photovoltaic and energy storage systems are highly synchronized, and they are likely to generate high-power power simultaneously, causing distribution network overload. Therefore, a strong positive correlation is determined between the two systems. If the correlation value is less than or equal to the positive correlation threshold, the synchronous output characteristics of the two systems are weak, and they are not included in the category of strong positive correlation. If the correlation value of a paired system is less than the aforementioned negative correlation threshold, it indicates that the historical positive output time series of the two photovoltaic and energy storage systems have peak-shifting and complementary characteristics. When one system generates high-power power, the other system is mostly in a low-output state. Therefore, a strong negative correlation is determined between the two systems. If the correlation value is greater than or equal to the negative correlation threshold, the complementary output characteristics of the two systems are not obvious, and they are not included in the category of strong negative correlation.

[0037] Based on the strong positive correlations already determined between the photovoltaic and energy storage systems to be allocated, cluster subgroups are divided to construct several independent synchronous correlation sets. The division logic is as follows: any two photovoltaic and energy storage systems to be allocated within the same synchronous correlation set can be directly connected or indirectly transmitted through one or more strong positive correlation links, and all equipment within the set has the operating characteristics of synchronous high-power positive power transmission; there are no direct or indirect strong positive correlation transmission links between photovoltaic and energy storage systems belonging to different synchronous correlation sets, and the output synchronization of different sub-clusters is independent of each other.

[0038] For each completed synchronization association set, the adjustment margin feature values ​​corresponding to all optical storage systems within the set are summed, and the summation result is the total adjustment margin of the synchronization association set. The total adjustment margin can characterize the upper limit of the potential for power reduction of the entire synchronization sub-cluster. The proportion of the adjustment margin feature of a single target system in the total adjustment margin of the synchronization association set to which the target system belongs is calculated. Based on the proportion mapping, the first collaborative feature used to characterize the adjustment contribution capability of a single device in the synchronization cluster is obtained. This feature can be used to distinguish the power reduction priority of each optical storage system in the synchronization cluster.

[0039] The total number of pairwise pairings between the target system and all other optical storage units within the target system cluster is calculated. This total number represents the total scale of equipment in the current scheduling cluster that can form an output coupling relationship with the target system. Next, the number of all strongly negatively correlated pairings corresponding to the target system is calculated. Strongly negatively correlated pairings refer to other optical storage units whose historical positive output exhibits a staggered and complementary pattern with the target system. The second collaborative characteristic corresponding to the target system is obtained by dividing the number of strongly negatively correlated pairings of the target system by the total number of pairings. The characteristic value range is fixed. If the value of the second collaborative feature is larger, it indicates that the target system and more devices in the cluster have a coupling relationship of complementary output and staggered operation. When the cluster synchronously sends power and causes the distribution network to overload, the system is more likely to be in a low-output condition. During the collaborative scheduling process, the power reduction of the system can be appropriately reduced.

[0040] After sequentially completing the quantitative solutions for individual characteristics of single systems, the first collaborative characteristics of cluster synchronization, and the second collaborative characteristics of cluster complementarity, a comprehensive understanding has been achieved of the multi-dimensional operational attributes of each photovoltaic-storage system to be allocated, including its own adjustable potential, its adjustment contribution capability in the synchronous sub-cluster, and its degree of peak-shaving complementarity with other equipment. However, the above multi-dimensional characteristics can only quantitatively represent the adjustment priority of equipment and cannot directly output the optimal power allocation scheme for each photovoltaic-storage system that takes into account both the power security of the distribution network and the overall adjustment cost. Therefore, this invention introduces an improved particle swarm optimization algorithm to solve for the optimal power. Based on the fusion calculation of the first and second collaborative characteristics, the adjustment suppression coefficient is obtained, thereby realizing the adaptive correction of the algorithm parameters for each equipment. Combined with the scheduling cost fitness function constructed with the adjustment margin as the core, the algorithm iteratively optimizes within the total power security constraint boundary of the distribution network to obtain the optimal grid-connected power output of each photovoltaic-storage system at the next moment. The specific execution steps are as follows: S3: Construct a particle swarm optimization model. Using the particle dimension of the grid-connected power output of each photovoltaic and energy storage system at the next time step, construct a fitness function based on the adjustment margin feature. Calculate the adjustment suppression coefficient using the first and second cooperative features. Based on the adjustment suppression coefficient, perform dimensional adaptive adjustment of the individual learning factor of the particle swarm algorithm. Based on the updated individual learning factor and fitness function, iteratively optimize the particles. Under the premise of satisfying the total power safety constraint of the distribution network, take the globally optimal particle position of the optimization result as the optimal output grid-connected power of each photovoltaic and energy storage system at the next time step.

[0041] First, for any target system within the photovoltaic-storage system cluster to be allocated, a natural exponential negative mapping operation is performed on the adjustment margin characteristics of the target system. The value obtained after mapping is defined as the adjustment cost coefficient. The higher the adjustment margin characteristic value, the more adjustable the energy storage device has and the lower the adjustment loss. The smaller the corresponding adjustment cost coefficient after exponential mapping, the more likely the algorithm will prioritize allocating power reduction tasks to this type of device.

[0042] Secondly, quantify the deviation of a single photovoltaic-storage system's output from the original autonomous strategy: determine the target system... The difference between the ideal grid-connected power and the output grid-connected power at any given time is used to determine the output deviation of a single device. The larger the deviation value, the greater the change in the original output scheme of the system by the dispatch command, and the higher the economic and loss costs brought about by the adjustment of energy storage charging and discharging.

[0043] The individual scheduling cost of a single target system is obtained by multiplying the adjustment cost coefficient of the single target system by the output deviation. All optical storage systems in the set to be allocated are traversed and the individual scheduling cost of each device is accumulated. The summation is used to obtain the overall scheduling cost of the cluster. The overall scheduling cost of the cluster is set as the fitness function of the particle swarm optimization algorithm. The algorithm takes minimizing this fitness value as the core optimization objective.

[0044] Specifically, the fitness function satisfies the following relationship: ; In the formula, This represents the fitness of the particle; This represents the total number of systems in the photovoltaic-storage system cluster to be allocated; Indicates the first The regulation margin characteristics of an individual photovoltaic energy storage system; Indicates the first A photovoltaic storage system in Ideal grid-connected power at any given time; Indicates the first A photovoltaic storage system in Output grid-connected power at any given time.

[0045] For any target system within the photovoltaic-storage system cluster to be allocated, classification and discrimination are carried out based on the results of strong positive correlation and strong negative correlation: it is determined whether the target system has a strong positive correlation or a strong negative correlation with any other photovoltaic-storage unit in the cluster, and the adjustment and suppression coefficients used for algorithm parameter adaptation are calculated under two different operating conditions.

[0046] If the target system has at least one pair of strong positive correlation or strong negative correlation, it means that the system has output synchronization or complementary coupling characteristics with other devices in the cluster. The arithmetic mean of the sum of the values ​​of the first cooperative feature and the second cooperative feature obtained by solving the target system is taken as the regulation and suppression coefficient of the target system. The regulation and suppression coefficient comprehensively reflects the regulation weight and peak-shifting operation attribute of the system in the synchronous cluster. The larger the coefficient, the less suitable the system is to significantly reduce the grid-connected power.

[0047] If the target system has neither any strong positive correlation pairing nor any strong negative correlation pairing, it is defined as an isolated system. The output timing of such equipment has no obvious synchronization or complementary pattern with the other equipment in the cluster. The result of subtracting the target system's own adjustment margin characteristic from the value of 1 is directly used as the adjustment suppression coefficient. The higher the adjustment margin characteristic of the isolated system and the stronger its own adjustment potential, the smaller the corresponding adjustment suppression coefficient value, which means that the power reduction operation of the equipment can be given priority during the scheduling process.

[0048] Based on the adaptively updated individual learning factors for each optical energy storage system dimension, the preset unified social learning factor, and the linearly decreasing inertial weights, a particle loop iterative optimization process is initiated. The complete iterative update logic is as follows: For each particle in the population, the particle velocity vector is first updated using an adaptive dimensional-level individual learning factor. Then, the particle's position in each dimension is corrected based on the updated velocity. Each dimension position corresponds to the grid-connected power to be optimized for a single photovoltaic-storage system at the next moment. After the particle position is updated, a dual constraint check is performed simultaneously: first, the upper and lower limits of the charging and discharging power of a single energy storage device; second, the total output of the distribution network cluster does not exceed the maximum allowable transmission power threshold of the distribution network and is controlled below the distribution network overload threshold. If the power of any dimension of the particle exceeds the limit, or the total output of the cluster exceeds the distribution network safety threshold, the power of the out-of-bounds dimension is truncated to ensure that the output allocation scheme corresponding to the particle is engineering feasible.

[0049] After constraint verification is completed, the corrected particle position is substituted into the overall scheduling cost fitness function constructed above to solve for the current particle fitness value; the current fitness of the particle is compared with the fitness corresponding to the best position of the particle in history. If the current fitness is better, the best position of the particle is updated; after traversing all particles to complete the individual best update, the best fitness of all particles is compared again, and the global best particle position of the population is refreshed synchronously.

[0050] The entire iterative process of continuously updating speed, updating position, verifying constraints, calculating fitness, and updating individual / global optimum is executed in a loop until a preset iteration termination condition is met. The termination condition can be exemplarily set as reaching the maximum number of iterations or the global optimum fitness showing no significant decrease over multiple consecutive generations. After iteration convergence, the global optimum particle position vector that satisfies the total power safety constraint of the distribution network is extracted. Each dimension of this vector corresponds to the optimal output grid-connected power of a photovoltaic-storage system to be allocated at the next moment.

[0051] S4: Generate scheduling instructions for the optimal output grid-connected power and send them to the local controllers of each photovoltaic and energy storage system.

[0052] The globally optimal particle position vector obtained after iterative optimization convergence is extracted. The optimal output grid-connected power in each dimension of the vector is standardized and encapsulated to generate standardized collaborative scheduling instructions adapted to the communication protocol of the local controllers of photovoltaic and energy storage systems. The scheduling instructions store the target grid-connected output value to be executed by each photovoltaic and energy storage system to be allocated at the next moment according to the device number. The encapsulated scheduling instructions are sent point-to-point to the local controllers associated with each photovoltaic and energy storage system to be allocated through the distribution network communication network. After receiving the scheduling instructions, each local controller abandons the ideal grid-connected power generated by its own energy storage autonomous charging and discharging control strategy and performs energy storage charging and discharging adjustment operations according to the optimal output grid-connected power specified in the scheduling instructions. This keeps the total grid-connected output of the cluster within the distribution network overload threshold and eliminates the safety risk of distribution network power overload.

[0053] This invention also provides a photoelectric storage collaborative optimization scheduling system based on multi-dimensional features. For example... Figure 2 As shown, the system includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the optical-storage collaborative optimization scheduling method based on multi-dimensional features according to the first aspect of the present invention. The system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface. Their configuration and functions are known in the art and will not be described further here.

[0054] It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of this invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A photoelectric-storage collaborative optimization scheduling method based on multi-dimensional features, characterized in that, include: The ideal grid-connected power of multiple photovoltaic and energy storage systems under the distribution network is obtained in real time. When the sum of the ideal grid-connected power of multiple photovoltaic and energy storage systems at the next moment is greater than the preset distribution network overload threshold, it is determined that there is a power overload risk and triggers collaborative optimization scheduling, and obtains the historical operation data of each photovoltaic and energy storage system to be allocated. Based on the historical operating data, the grid-connected power fluctuation characteristics, adjustment margin characteristics, and historical output power sequences of each photovoltaic and energy storage system to be allocated are extracted; based on the historical output power sequences, strong positive and strong negative correlations between each photovoltaic and energy storage system are defined; based on the strong positive correlation, the first collaborative feature characterizing the system cluster synchronization characteristics is extracted, and based on the strong negative correlation, the second collaborative feature characterizing the power output complementarity characteristics is extracted. A particle swarm optimization model is constructed, with the output grid-connected power of each photovoltaic and energy storage system at the next time step as the particle dimension. A fitness function is constructed based on the adjustment margin feature. The first and second cooperative features are used to calculate the adjustment suppression coefficient. Based on the adjustment suppression coefficient, the individual learning factor of the particle swarm algorithm is adaptively adjusted at the dimensional level. Based on the updated individual learning factor and fitness function, the particles are iteratively optimized. Under the premise of satisfying the total power safety constraint of the distribution network, the globally optimal particle position of the optimization result is taken as the optimal output grid-connected power of each photovoltaic and energy storage system at the next time step. The optimal output grid-connected power is used to generate scheduling instructions and sent to the local controllers of each photovoltaic and energy storage system. The first collaborative feature specifically includes: Based on the strong positive correlation between each optical storage system to be allocated, a synchronous correlation set is constructed; wherein, any two optical storage systems to be allocated within the same synchronous correlation set have a direct or indirect strong positive correlation, and optical storage systems to be allocated in different synchronous correlation sets do not have the strong positive correlation. The sum of the regulation margin characteristics of all photovoltaic and energy storage systems in the synchronous association set is calculated to obtain the total regulation margin of the synchronous association set. The proportion of the regulation margin characteristic of each photovoltaic and energy storage system is taken as the first cooperative characteristic. The second collaborative feature specifically includes: Taking any photovoltaic-storage system to be allocated as the target system, the total number of all relationships of the target system is counted; the ratio of the number of strong negative relationships of the target system to the total number of all relationships of the target system is calculated as the second synergistic feature of the photovoltaic-storage system to be allocated.

2. The photoelectric-storage collaborative optimization scheduling method based on multi-dimensional features according to claim 1, characterized in that, After determining the existence of power overload risk and triggering collaborative optimization scheduling, and before obtaining the historical operating data of each photovoltaic-storage system to be allocated, the process also includes: The photovoltaic and energy storage systems in the distribution network whose ideal grid-connected power is less than or equal to 0 at the next moment are eliminated, and the eliminated photovoltaic and energy storage systems are taken as photovoltaic and energy storage systems to be allocated.

3. The photoelectric-storage collaborative optimization scheduling method based on multi-dimensional features according to claim 1, characterized in that, The extraction of grid-connected power fluctuation characteristics of a single photovoltaic-storage system specifically includes: Taking any photovoltaic-storage system to be allocated as the target system, the historical grid-connected power sequence of the target system is traversed, and the grid-connected power change between adjacent sampling times is calculated sequentially; weighting coefficients are configured according to the sign difference of the grid-connected power product at adjacent times, and the weighted operation of each grid-connected power change is performed and the arithmetic mean is calculated; the arithmetic mean is mapped to obtain the grid-connected power fluctuation characteristics of the target system.

4. The optical-storage collaborative optimization scheduling method based on multi-dimensional features according to claim 1, characterized in that, The step of calculating the adjustment margin feature includes: Taking any photovoltaic-storage system to be allocated as the target system, obtain the upper and lower limits of the energy storage state of charge (SBC) of the target system, as well as the current SBC, energy storage health status, and grid-connected power fluctuation characteristics. Based on the upper and lower limits of the SBC and the current SBC of the target system, calculate the ratio of the remaining charging capacity to the total adjustable energy storage capacity. Using the ratio, the grid-connected power fluctuation characteristics, and the energy storage health status, calculate the adjustment margin characteristics of the target system at the current moment.

5. The optical-storage collaborative optimization scheduling method based on multi-dimensional features according to claim 1, characterized in that, The strong positive associations and strong negative associations include: Traverse the historical grid-connected power sequence of each photovoltaic and energy storage system to be allocated. For the grid-connected power at any moment in the historical grid-connected power sequence, if the grid-connected power is greater than 0, it is determined to be positive output power and the original value is retained. Otherwise, if it is less than or equal to 0, the grid-connected power value at the corresponding moment is set to zero, thus obtaining the historical output power sequence of each photovoltaic and energy storage system to be allocated. The photovoltaic and energy storage systems within the photovoltaic and energy storage system cluster to be allocated are combined in pairs, and the correlation values ​​between the historical output power sequences of the two photovoltaic and energy storage systems in each pair are calculated based on cosine similarity. Based on the sign of the correlation values, all combinations are divided into a positive correlation set or a negative correlation set. All correlation values ​​in the positive correlation set are sorted in ascending order, and the value at the corresponding position of the preset quantile is selected as the positive correlation threshold. All correlation values ​​in the negative correlation set are sorted in descending order, and the value at the corresponding position of the preset quantile is selected as the negative correlation threshold. If the correlation value is greater than the positive correlation threshold, a strong positive correlation is determined between the two photovoltaic-storage systems in the corresponding photovoltaic-storage system combination; otherwise, if the correlation value is less than or equal to the positive correlation threshold, no analysis is performed. If the correlation value is less than the negative correlation threshold, a strong negative correlation is determined between the two photovoltaic-storage systems in the corresponding photovoltaic-storage system combination; otherwise, if the correlation value is greater than or equal to the negative correlation threshold, no analysis is performed.

6. The optical-storage collaborative optimization scheduling method based on multi-dimensional features according to claim 1, characterized in that, The calculation of the modulation inhibition coefficient includes: Taking any photovoltaic-storage system to be allocated as the target system, determine whether there is a strong positive correlation or a strong negative correlation between the target systems; If it exists, the average of the sum of the first and second cooperative features of the target system is calculated as the regulation inhibition coefficient; If it does not exist, the target system is determined to be an isolated system, and the regulation margin characteristic of the target system minus 1 is taken as the regulation inhibition coefficient.

7. The optical-storage collaborative optimization scheduling method based on multi-dimensional features according to claim 1, characterized in that, The fitness function constructed based on the adjustment margin feature includes: Taking any one of the photovoltaic-storage systems in the cluster to be allocated as the target system, the adjustment margin characteristics of the target system are exponentially mapped, and the mapping result is used as the adjustment cost coefficient. Calculate the difference between the ideal grid-connected power of the target system at the next time step and the actual grid-connected power output at the next time step, and use the ratio of the difference to the ideal grid-connected power at the next time step as the degree of deviation of the target system. The product of the adjustment cost coefficient and the degree of deviation is used as the individual scheduling cost of the target system. The individual scheduling costs of all optical storage systems are accumulated to obtain the overall scheduling cost, which is then used as the fitness function of a particle.

8. A photovoltaic-storage collaborative optimization scheduling system based on multi-dimensional features, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement the optical-storage collaborative optimization scheduling method based on multi-dimensional features according to any one of claims 1-7.

Images