Network resource cooperative scheduling method and system for photovoltaic power generation

Abstract

The invention discloses a network resource collaborative scheduling method and system for photovoltaic power generation, and mainly relates to the technical field of resource scheduling. Comprising the following steps: constructing a working memory snapshot mechanism, collecting working memory snapshots of a photovoltaic power generation system, an energy storage system, an uninterruptible power supply and a network load according to a preset period, and obtaining real-time site operation memory snapshots; constructing a long-term collaborative scheduling memory bank; determining a matched cooperative scheduling memory; and performing network resource cooperative scheduling prediction on the real-time site operation memory snapshot based on a matched power supply strategy in the matched cooperative scheduling memory, and if a prediction result meets a requirement, performing scheduling based on the matched power supply strategy. The method has the beneficial effects that the technical problem of low resource scheduling reliability caused by the limitation that photovoltaic power generation only focuses on energy supply and lacks linkage with network equipment in the prior art is solved, and the technical effects of improving resource scheduling accuracy and response timeliness are achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of resource scheduling, in particular to a network resource cooperative scheduling method and system for photovoltaic power generation. BACKGROUND

[0002] With the rapid deployment of 5G networks, large-scale data centers and edge computing nodes, the energy consumption of communication sites continues to grow, leading to increased operation and maintenance costs and intensified energy supply and demand contradictions. In recent years, photovoltaic power generation and supporting energy storage systems have been gradually applied to communication base stations to reduce electricity purchase costs and improve green energy utilization. However, existing photovoltaic base stations mainly use photovoltaic as an independent energy supplement unit, and their operation process relies on fixed strategies or simple threshold logic, which cannot adjust resources in real time according to changes in network traffic. At the same time, the load of communication network equipment is highly dynamic, and the amount of concurrent traffic, link utilization, and device power consumption fluctuate rapidly over time, making it difficult to accurately match photovoltaic output and network load.

[0003] Since existing photovoltaic or energy storage control strategies are usually executed according to fixed rules, without establishing a linkage mechanism between "photovoltaic power generation output-network device load", it is not possible to dynamically adjust power supply decisions according to changes in traffic load, resulting in low photovoltaic utilization and power supply strategy lag. And current base station energy scheduling mostly relies on the state of the current moment in isolation, without using historical site operation patterns and successful strategies, so it is not possible to reuse historical effective power supply decisions under similar weather, time period or load composition. At the same time, the switching of traditional UPS and energy storage systems is mostly based on hardware thresholds such as voltage and current, and does not take into account network SLA, traffic priority or the importance of critical devices, resulting in the inability to ensure the business continuity of critical network devices when there is insufficient light or sudden load.

[0004] The existing technology has the limitation that photovoltaic power generation only focuses on energy supply, lacks linkage with network equipment, and causes the technical problem of low reliability of resource scheduling. SUMMARY

[0005] The present application provides a network resource cooperative scheduling method and system for photovoltaic power generation, which is used to solve the technical problem of low reliability of resource scheduling caused by the limitation that photovoltaic power generation only focuses on energy supply and lacks linkage with network equipment in the prior art.

[0006] In view of the above problems, the present application provides a network resource cooperative scheduling method and system for photovoltaic power generation.

[0007] In a first aspect of the present application, a network resource collaborative scheduling method for photovoltaic power generation is provided, which comprises: constructing a working memory snapshot mechanism, collecting working memory snapshots of a photovoltaic power generation system, an energy storage system, an uninterruptible power supply and a network load according to a preset period to obtain a real-time site operation memory snapshot; constructing a long-term collaborative scheduling memory library based on a historical site operation memory snapshot set and a corresponding historical power supply strategy set, wherein the long-term collaborative scheduling memory library comprises a plurality of collaborative scheduling memories, and each collaborative scheduling memory has a scene tag; matching and identifying the long-term collaborative scheduling memory library according to the real-time site operation memory snapshot in combination with the scene tag to determine a matching collaborative scheduling memory; and performing network resource collaborative scheduling prediction on the real-time site operation memory snapshot based on a matching power supply strategy in the matching collaborative scheduling memory, and performing scheduling based on the matching power supply strategy if the prediction result meets the requirements.

[0008] In a possible implementation, the working memory snapshot mechanism is constructed, the working memory snapshots of the photovoltaic power generation system, the energy storage system, the uninterruptible power supply and the network load are collected according to a preset period to obtain a real-time site operation memory snapshot, which comprises: aligning the photovoltaic power generation system, the energy storage system, the uninterruptible power supply and the network load according to the same clock reference; calling state parameters of the photovoltaic power generation system, the energy storage system, the uninterruptible power supply and the network load in the working memory snapshot mechanism, collecting data of the photovoltaic power generation system, the energy storage system, the uninterruptible power supply and the network load according to the preset period to obtain the real-time site operation memory snapshot; wherein the real-time site operation memory snapshot comprises a photovoltaic side operation situation, an energy storage side operation situation, a power supply side operation situation and a network side operation situation.

[0009] In a possible implementation, the long-term collaborative scheduling memory library is constructed based on the historical site operation memory snapshot set and the corresponding historical power supply strategy set, wherein the long-term collaborative scheduling memory library comprises a plurality of collaborative scheduling memories, and each collaborative scheduling memory has a scene tag, which comprises: clustering and dividing based on the historical site operation memory snapshot set to obtain a plurality of clustered historical site operation memory snapshot sets; identifying scene tags according to the plurality of clustered historical site operation memory snapshot sets to determine a plurality of scene tags; mapping and clustering the historical power supply strategy set according to the plurality of clustered historical site operation memory snapshot sets to obtain a plurality of clustered historical power supply strategy sets; identifying collaborative scheduling memories from the plurality of clustered historical power supply strategy sets to obtain a plurality of collaborative scheduling memories; and identifying the plurality of collaborative scheduling memories using the plurality of scene tags to construct the long-term collaborative scheduling memory library.

[0010] In a possible implementation, the scene label identification is performed on the set of clustered historical site operation memory snapshots, and a plurality of scene labels are determined, including: scene feature extraction is respectively performed on the set of clustered historical site operation memory snapshots to obtain a plurality of scene feature sets; two-by-two scene feature non-repeated interaction enhancement is respectively performed on the plurality of scene feature sets to obtain a plurality of interaction-enhanced scene feature sets; and mean value processing is performed on the plurality of interaction-enhanced scene feature sets to obtain the plurality of scene labels.

[0011] In a possible implementation, the cooperative scheduling memory identification is performed on the set of clustered historical power supply strategies, and a plurality of cooperative scheduling memories are obtained, including: a historical power supply strategy with the highest similarity to other historical power supply strategies in the set of clustered historical power supply strategies is extracted as a plurality of first cooperative scheduling memories; historical power supply strategies in the set of clustered historical power supply strategies that have a similarity to the plurality of first cooperative scheduling memories within a preset similarity threshold are summarized to construct a plurality of first cooperative scheduling memory spaces; the number of historical power supply strategies in the plurality of first cooperative scheduling memory spaces is counted to obtain a plurality of first memory quantities; and the plurality of first cooperative scheduling memories are iteratively updated in combination with the plurality of first memory quantities, with the plurality of first cooperative scheduling memory spaces as memory identification constraints, to obtain the plurality of cooperative scheduling memories.

[0012] In a possible implementation, the plurality of first cooperative scheduling memories are iteratively updated in combination with the plurality of first memory quantities, with the plurality of first cooperative scheduling memory spaces as memory identification constraints, to obtain the plurality of cooperative scheduling memories, including: a historical power supply strategy is randomly extracted from the plurality of first cooperative scheduling memory spaces as a plurality of to-be-analyzed historical power supply strategies, with the plurality of first cooperative scheduling memory spaces as memory identification constraints; the number of historical power supply strategies in the set of clustered historical power supply strategies that have a similarity to the plurality of to-be-analyzed historical power supply strategies within a preset similarity threshold is counted to obtain a plurality of second memory quantities; and it is determined whether the plurality of second memory quantities are greater than or equal to the plurality of first memory quantities, if yes, the plurality of to-be-analyzed historical power supply strategies are taken as a plurality of second cooperative scheduling memories, and the plurality of second cooperative scheduling memories are iteratively updated in combination with the plurality of second memory quantities, with the plurality of first cooperative scheduling memory spaces as memory identification constraints, until a preset update number of times is met, to obtain the plurality of cooperative scheduling memories.

[0013] In a possible implementation, if no, a historical power supply strategy is again randomly extracted from the plurality of first cooperative scheduling memory spaces as a plurality of to-be-analyzed historical power supply strategies, in combination with the plurality of first memory quantities, until a preset re-selection number of times is met, to obtain the plurality of cooperative scheduling memories.

[0014] In a possible implementation, if the prediction result does not meet the requirement, the matching power supply strategy is randomly adjusted in a random direction for multiple times to obtain multiple adjusted power supply strategies; the real-time station operation memory snapshot is predicted for network resource collaborative scheduling based on the multiple adjusted power supply strategies to obtain multiple adjusted prediction results; it is judged whether there is an adjusted prediction result meeting the requirement in the multiple adjusted prediction results, if yes, an adjustment direction is obtained, the matching power supply strategy is adjusted and screened according to a preset amplitude to obtain a target power supply strategy, and scheduling is performed based on the target power supply strategy.

[0015] In a possible implementation, the adjusting and screening of the matching power supply strategy according to the preset amplitude to obtain the target power supply strategy includes: taking the optimal solution in the multiple adjusted prediction results as the adjustment direction; adjusting the matching power supply strategy according to the preset amplitude to obtain multiple direction-adjusted matching power supply strategies; predicting the real-time station operation memory snapshot for network resource collaborative scheduling based on the multiple direction-adjusted matching power supply strategies to obtain multiple direction-adjusted prediction results; and when there is a direction-adjusted prediction result better than the optimal solution in the multiple adjusted prediction results in the multiple direction-adjusted prediction results, taking the direction-adjusted matching power supply strategy corresponding to the optimal solution in the multiple direction-adjusted prediction results as the target power supply strategy.

[0016] In a second aspect, the present application provides a network resource collaborative scheduling system for photovoltaic power generation, the system comprising: A memory snapshot collection module is configured to construct a working memory snapshot mechanism, collect working memory snapshots of a photovoltaic power generation system, an energy storage system, an uninterruptible power supply and a network load according to a preset period to obtain a real-time station operation memory snapshot; a memory bank construction module is configured to construct a long-term collaborative scheduling memory bank based on a historical station operation memory snapshot set and a corresponding historical power supply strategy set, wherein the long-term collaborative scheduling memory bank comprises a plurality of collaborative scheduling memories, and each collaborative scheduling memory has a scene tag; a matching collaborative scheduling memory determination module is configured to match and identify the long-term collaborative scheduling memory bank according to the real-time station operation memory snapshot and in combination with the scene tag to determine a matching collaborative scheduling memory; and a collaborative scheduling module is configured to predict the real-time station operation memory snapshot for network resource collaborative scheduling based on a matching power supply strategy in the matching collaborative scheduling memory, and perform scheduling based on the matching power supply strategy if the prediction result meets the requirement.

[0017] The one or more technical solutions provided in the present application have at least the following technical effects or advantages: The application acquires real-time station operation memory snapshot by constructing a working memory snapshot mechanism, collecting working memory snapshots of a photovoltaic power generation system, an energy storage system, an uninterruptible power supply and a network load according to a preset period; a long-term collaborative scheduling memory library is constructed based on a historical station operation memory snapshot set and a corresponding historical power supply strategy set, wherein the long-term collaborative scheduling memory library includes a plurality of collaborative scheduling memories, and each collaborative scheduling memory has a scene tag; the long-term collaborative scheduling memory library is matched and identified according to the real-time station operation memory snapshot combined with the scene tag, and a matched collaborative scheduling memory is determined; the real-time station operation memory snapshot is predicted for network resource collaborative scheduling based on a matched power supply strategy in the matched collaborative scheduling memory, and if the prediction result meets the requirement, the matched power supply strategy is used for scheduling. The technical effect of realizing intelligent linkage between photovoltaic output, energy storage capacity, UPS guarantee and network load, automatically matching the optimal historical strategy in multiple scenes and performing real-time correction, significantly improving power supply stability, photovoltaic utilization rate and business SLA achievement rate, and reducing overall energy consumption cost is achieved. BRIEF DESCRIPTION OF DRAWINGS

[0018] FIG. 1 is a flowchart of a network resource collaborative scheduling method for photovoltaic power generation provided by an embodiment of the application. Figure 1 FIG. 2 is a structural diagram of a network resource collaborative scheduling system for photovoltaic power generation provided by an embodiment of the application.

[0019] FIG. 3 is a schematic diagram of a network resource collaborative scheduling system for photovoltaic power generation provided by an embodiment of the application. Figure 2 FIG. 4 is a structural diagram of a network resource collaborative scheduling system for photovoltaic power generation provided by an embodiment of the application.

[0020] Reference signs in the drawings: Memory snapshot acquisition module 11, memory library construction module 12, matched collaborative scheduling memory determination module 13, collaborative scheduling module 14. DETAILED DESCRIPTION

[0021] The application will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the application and not to limit the scope of the application. Furthermore, it should be understood that those skilled in the art can make various modifications or changes to the application after reading the content taught by the application, and these equivalent forms also fall within the scope defined by the appended claims of the application. It should be noted that the terms "include" and "have" are intended to cover non-exclusive inclusion, for example, a process, method, system, product or server including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or modules not clearly listed or inherent to these processes, methods, products or devices.

[0022] Embodiment one, as shown in FIG. 1, the application provides a network resource collaborative scheduling method for photovoltaic power generation, wherein the method comprises: Figure 1 ​Step S100: Construct a working memory snapshot mechanism, collect working memory snapshots of the photovoltaic power generation system, energy storage system, uninterruptible power supply and network load according to a preset period, and obtain real-time site operation memory snapshots; Further, the working memory snapshot mechanism is constructed, the working memory snapshots of the photovoltaic power generation system, energy storage system, uninterruptible power supply and network load are collected according to a preset period, and real-time site operation memory snapshots are obtained. The step S100 of the embodiment of the application further includes: Aligning the photovoltaic power generation system, energy storage system, uninterruptible power supply and network load according to the same clock reference; Retrieving the state parameters of the photovoltaic power generation system, energy storage system, uninterruptible power supply and network load in the working memory snapshot mechanism, collecting data of the photovoltaic power generation system, energy storage system, uninterruptible power supply and network load according to a preset period, and obtaining real-time site operation memory snapshots; The real-time site operation memory snapshot includes photovoltaic side operation situation, energy storage side operation situation, power supply side operation situation and network side operation situation.

[0023] It should be noted that the working memory snapshot mechanism is used to periodically intercept the current operation state of the site, which can be understood as a full amount of operation image of the photovoltaic power generation system, energy storage system, uninterruptible power supply and network load at a certain moment, and is used to uniformly describe a plurality of real-time state parameters of the photovoltaic side, energy storage side, UPS power supply side and network load side. The preset period is a fixed sampling interval set according to the business characteristics and the response speed of the equipment, for example, 1 minute, 5 minutes or 10 minutes, which is used to unify the data collection rhythm of different subsystems. The same clock reference is used to keep the time of each device consistent through the precision time protocol or network time protocol, so as to ensure that the data collected by different devices are aligned in time, and solve the problem of different data synchronization caused by different sampling periods or clock drift of the photovoltaic system, energy storage system, UPS and network equipment.

[0024] Preferably, the real-time site operation memory snapshot includes photovoltaic side operation situation, energy storage side operation situation, power supply side operation situation and network side operation situation. The photovoltaic side operation situation includes photovoltaic power generation power, irradiance, inverter efficiency, etc.; the energy storage side operation situation includes SOC, charging and discharging power, charging mode, etc.; the power supply side operation situation refers to the output power of the UPS, backup time, bypass state, etc.; and the network side operation situation involves CPU utilization, link bandwidth occupancy, business throughput and device power consumption of the network equipment.

[0025] For example, at 10:00 at a certain site, the photovoltaic system generates 3.2 kW of power, the inverter temperature is 42°C; the energy storage system is in discharge mode, the SOC is 58%, and the current output power is 1.0 kW; the UPS is in online mode, and the output load rate is about 20%; in terms of network equipment, the link utilization rate of the switch is 35%, the CPU utilization rate of the server is 40%, and the instantaneous business load of the entire site corresponds to a total equipment power consumption of 950 W.

[0026] The above data is uniformly packaged into a real-time site operation memory snapshot, recorded as: time 10:00, photovoltaic = 3.2 kW, energy storage = 1.0 kW (SOC 58%), UPS = online 20% load, network load = 35% bandwidth, CPU = 40%. The technical effect of providing data support for subsequent scene recognition, memory matching and scheduling decision is achieved.

[0027] Step S200: based on the set of historical site operation memory snapshots and the set of corresponding historical power supply strategies, a long-term collaborative scheduling memory library is constructed, wherein the long-term collaborative scheduling memory library includes a plurality of collaborative scheduling memories, and each collaborative scheduling memory has a scene tag; Further, based on the set of historical site operation memory snapshots and the set of corresponding historical power supply strategies, a long-term collaborative scheduling memory library is constructed, wherein the long-term collaborative scheduling memory library includes a plurality of collaborative scheduling memories, and each collaborative scheduling memory has a scene tag, and the step S200 of the embodiment of the application further includes: Based on the set of historical site operation memory snapshots, clustering division is performed to obtain a plurality of clustered historical site operation memory snapshot sets; According to the plurality of clustered historical site operation memory snapshot sets, scene tag identification is performed to determine a plurality of scene tags; According to the plurality of clustered historical site operation memory snapshot sets, the set of historical power supply strategies is mapped and clustered to obtain a plurality of clustered historical power supply strategy sets; The plurality of clustered historical power supply strategy sets are subjected to collaborative scheduling memory identification to obtain a plurality of collaborative scheduling memories; The plurality of scene tags are used to identify the plurality of collaborative scheduling memories to construct the long-term collaborative scheduling memory library.

[0028] Further, according to the plurality of clustered historical site operation memory snapshot sets, scene tag identification is performed to determine a plurality of scene tags, and the step S200 of the embodiment of the application further includes: The plurality of clustered historical site operation memory snapshot sets are respectively subjected to scene feature extraction to obtain a plurality of scene feature sets; The plurality of scene feature sets are respectively subjected to intra-set two-by-two scene feature non-repeated interaction enhancement to obtain a plurality of interaction-enhanced scene feature sets; The plurality of interaction-enhanced scene feature sets are subjected to mean value processing to obtain the plurality of scene labels.

[0029] In one possible embodiment, the historical site operation memory snapshot set refers to a large number of site operation snapshots generated and stored periodically in step S100, which reflect the real operation state of the site under different time periods, different weather, and different load levels. The historical power supply strategy set refers to the strategies actually executed by the system during the generation of these snapshots, such as a photovoltaic priority strategy, a storage discharging strategy, a UPS switching strategy, a device degradation strategy, and the like. Preferably, the historical power supply strategy is generally divided into the following three levels: when the photovoltaic power generation is sufficient, the network equipment is preferentially powered and the energy storage system is charged; when the illumination is insufficient, the energy storage system is supplemented; and in the extreme case, the UPS is switched to guarantee, the data synchronization and instruction issuing are realized through the optical fiber transmission network, and the enterprise R&D equipment UPS, the optical fiber attenuator, and the computer network equipment are directly associated with the business. Each historical power supply strategy is refined on the basis of the above three levels, such as when the photovoltaic power generation reaches the power generation amount set by the person skilled in the art, the network equipment is preferentially powered and the energy storage system is charged.

[0030] All the historical site operation memory snapshot sets are input into a clustering algorithm, such as a K-means, DBSCAN, or hierarchical clustering algorithm. The clustering algorithm performs similarity calculation according to the feature vectors of various parameters in the historical site operation memory snapshot, such as photovoltaic output, energy storage SOC, UPS state, device power consumption, network load, and the like, and automatically classifies the snapshots with similar features into the same class.

[0031] For example, in the tens of thousands of snapshots collected within 30 days, the clustering algorithm can divide them into 6 main categories: such as a sunny afternoon high-illumination scene, an overcast all-day low-illumination scene, a night low-business load scene, a late-peak high-business pressure scene, an energy storage high-SOC charging window scene, and a UPS temporary intervention scene. Each clustering historical site operation memory snapshot set contains a snapshot set with highly consistent operation characteristics.

[0032] In one possible embodiment, key features representing the operation state of each clustering historical site operation memory snapshot set are extracted from the set, thereby obtaining the plurality of scene feature sets. Exemplarily, the scene feature set includes an illumination intensity interval, a photovoltaic output level, an energy storage SOC distribution, whether the UPS is involved, a network load curve form, and the like, thereby abstracting the complex multi-dimensional snapshot data into element expressions that can be used for scene judgment.

[0033] The feature items in each scene feature set are enhanced in pairs within the set. In the cluster of high light in sunny afternoon, the light intensity and high photovoltaic output are combined with the light intensity and large energy storage charging to strengthen the composite features that truly reflect the characteristics of the scene, so that the scene label is more accurate.

[0034] The average values of the enhanced interaction-enhanced scene feature sets are processed to extract the central feature expression of the scene, such as the average irradiance of 850 W / m², the average photovoltaic output of 3.1 kW, the average energy storage SOC of 47%, and the average network load of 35%. These average features are semantically generated into scene labels.

[0035] Specifically, a first scene feature set is randomly extracted from a plurality of scene feature sets. Then, the first scene feature set is combined in a non-repeating manner to obtain a plurality of scene feature combinations. Then, similarity calculation is performed on two scene features in each scene feature combination, and the calculation results are normalized and added to an initially empty matrix to construct an interaction-enhanced matrix. The two scene features are respectively convoluted and enhanced by the interaction-enhanced matrix to obtain two interaction-enhanced scene features. Based on the same principle, the plurality of scene feature combinations in the first scene feature set are interaction-enhanced to obtain a first interaction-enhanced scene feature set. The plurality of scene feature sets are processed in the same way as the first scene feature set to obtain a plurality of interaction-enhanced scene feature sets.

[0036] The collaborative scheduling memory recognition is to identify the most representative and reusable power supply strategy structure from each clustered historical power supply strategy set, including analyzing the common strategy mode, typical strategy combination and successful performance of the strategy in the cluster historical strategy, so as to extract a collaborative scheduling memory. That is, the collaborative scheduling memory is a structured knowledge unit that contains what strategy can obtain good results in a certain scene. Further, the collaborative scheduling memory extracted above is bound with the scene label generated in the previous step to form a one-to-one corresponding structure of scene-strategy mode in the long-term collaborative scheduling memory library.

[0037] Further, the plurality of clustered historical power supply strategy sets are subjected to collaborative scheduling memory recognition to obtain a plurality of collaborative scheduling memories. The embodiment of the application step S200 further comprises: respectively extracting the historical power supply strategy with the highest similarity to other historical power supply strategies in the plurality of clustered historical power supply strategy sets as a plurality of first collaborative scheduling memories; aggregate the historical power supply strategies in the plurality of clustered historical power supply strategy sets that respectively have a similarity to the plurality of first collaborative scheduling memories within a preset similarity threshold, to construct a plurality of first collaborative scheduling memory spaces; count the number of historical power supply strategies in the plurality of first collaborative scheduling memory spaces, to obtain a plurality of first memory quantities; perform iterative updating on the plurality of first collaborative scheduling memories in combination with the plurality of first memory quantities, with the plurality of first collaborative scheduling memory spaces as memory identification constraints, to obtain a plurality of collaborative scheduling memories.

[0038] Further, perform iterative updating on the plurality of first collaborative scheduling memories in combination with the plurality of first memory quantities, with the plurality of first collaborative scheduling memory spaces as memory identification constraints, to obtain a plurality of collaborative scheduling memories. The step S200 of the embodiment of the present application further includes: randomly extract one historical power supply strategy from the plurality of first collaborative scheduling memory spaces as a plurality of historical power supply strategies to be analyzed, with the plurality of first collaborative scheduling memory spaces as memory identification constraints; count the number of historical power supply strategies in the plurality of clustered historical power supply strategy sets that respectively have a similarity to the plurality of historical power supply strategies to be analyzed within a preset similarity threshold, to obtain a plurality of second memory quantities; determine whether the plurality of second memory quantities are greater than or equal to the plurality of first memory quantities. If yes, take the plurality of historical power supply strategies to be analyzed as a plurality of second collaborative scheduling memories, continue to perform iterative updating on the plurality of second collaborative scheduling memories in combination with the plurality of second memory quantities, with the plurality of first collaborative scheduling memory spaces as memory identification constraints, until a preset number of updates is met, to obtain a plurality of collaborative scheduling memories.

[0039] Further, if no, again randomly extract one historical power supply strategy from the plurality of first collaborative scheduling memory spaces as a plurality of historical power supply strategies to be analyzed, with the plurality of first collaborative scheduling memory spaces as memory identification constraints, and perform iterative updating in combination with the plurality of first memory quantities, until a preset number of reselections is met, to obtain a plurality of collaborative scheduling memories.

[0040] In one embodiment, the similarity between all strategies in each clustered historical power supply strategy set is calculated. For example, in a certain evening peak high-load cluster, there are 90 historical strategies in total, and it is found through calculation that the average similarity of a certain strategy is the highest, and therefore the strategy is selected as the first collaborative scheduling memory. According to a preset similarity threshold, such as 0.85, all historical strategies with a similarity greater than or equal to 0.85 to the first collaborative scheduling memory are screened out to form a first collaborative scheduling memory space, and the number of strategies in the space is counted, for example, 65, which is the first memory quantity. Thus, a plurality of first collaborative scheduling memories, a plurality of first collaborative scheduling memory spaces, and a plurality of first memory quantities corresponding thereto are obtained.

[0041] Further, an iterative updating phase is entered, a strategy is randomly extracted from the plurality of first collaborative scheduling memory spaces as a plurality of historical power supply strategies to be analyzed, the similarity of the strategy to the corresponding entire clustering historical strategy set is calculated, and the number of strategies with a similarity greater than or equal to a threshold of 0.85, that is, the plurality of second memory quantities, is counted. If the plurality of second memory quantities is greater than or equal to the plurality of first memory quantities, it indicates that the plurality of historical power supply strategies to be analyzed are more representative than the plurality of first collaborative scheduling memories in terms of strategy structure. At this time, the plurality of historical power supply strategies to be analyzed are used as new center strategies, the plurality of second collaborative scheduling memories, and the process is repeated to further find better center strategies. This process will continue to be executed for a plurality of update rounds until a preset number of updates, such as 10 times, is met.

[0042] If the second memory quantity of the historical power supply strategy to be analyzed does not meet the condition in a certain extraction process, the strategy is randomly extracted from the first collaborative scheduling memory space for analysis according to the re-selection number requirement, so as to avoid deviation from some abnormal points. Finally, the collaborative scheduling memory output in this step is the most representative typical strategy in the cluster.

[0043] Step S300: According to the real-time site running memory snapshot, the long-term collaborative scheduling memory library is matched and identified in combination with the scene label, and a matching collaborative scheduling memory is determined. In one embodiment, the real-time site running memory snapshot is a comprehensive state snapshot at the current time generated in step S100 according to a preset period, which records key indicators such as photovoltaic output, energy storage SOC, UPS state, network service load, etc. For example, the photovoltaic output is 2.8kW, the energy storage SOC is 52%, the UPS is in bypass standby, and the network link load is 38%. These real-time parameters are converted into a feature vector, and a similarity comparison is performed with the scene label in the long-term collaborative scheduling memory library. The collaborative scheduling memory with the highest similarity is used as the matching collaborative scheduling memory.

[0044] Through automatic alignment of real-time scenes and historical experience, the consistency of the scheduling strategy is ensured, the response speed of the scheduling is greatly improved, and the technical effect of avoiding re-solving complex optimization is achieved.

[0045] Step S400: Based on the matching power supply strategy in the matching collaborative scheduling memory, network resource collaborative scheduling prediction is performed on the real-time site running memory snapshot. If the prediction result meets the requirement, the matching power supply strategy is used for scheduling.

[0046] Further, if the prediction result does not meet the requirement, the matching power supply strategy is randomly adjusted in a random direction for multiple times to obtain a plurality of adjusted power supply strategies. perform network resource collaborative scheduling prediction on the real-time site operation memory snapshot based on the plurality of adjustment power supply strategies, to obtain a plurality of adjustment prediction results; determine whether there is an adjustment prediction result meeting the requirement in the plurality of adjustment prediction results, if yes, obtain an adjustment direction, adjust and screen the matching power supply strategy according to a preset amplitude, obtain a target power supply strategy, and perform scheduling based on the target power supply strategy.

[0047] Further, an adjustment direction is obtained, the matching power supply strategy is adjusted and screened according to a preset amplitude, a target power supply strategy is obtained, and the method further comprises: the optimal solution in the plurality of adjustment prediction results is taken as the adjustment direction; the matching power supply strategy is adjusted according to a preset amplitude to the adjustment direction, to obtain a plurality of direction adjustment matching power supply strategies; perform network resource collaborative scheduling prediction on the real-time site operation memory snapshot based on the plurality of direction adjustment matching power supply strategies, to obtain a plurality of direction adjustment prediction results; when there is a direction adjustment prediction result better than the optimal solution in the plurality of adjustment prediction results in the plurality of direction adjustment prediction results, the direction adjustment matching power supply strategy corresponding to the optimal solution in the plurality of direction adjustment prediction results is taken as the target power supply strategy.

[0048] In one embodiment, according to the real-time site operation memory snapshot and the candidate power supply strategy, the energy flow direction, load support ability memory and resource scheduling achievement of the site after executing the strategy are predicted and verified. Preferably, a plurality of sample matching power supply strategies and a plurality of sample real-time site operation memory snapshots and corresponding a plurality of sample scheduling results are obtained as training data, a framework constructed based on a feedforward neural network is supervised and trained until the training converges, to obtain a trained collaborative scheduling predictor. The network resource collaborative scheduling predictor is used to perform network resource collaborative scheduling prediction on the matching power supply strategy and the real-time site operation memory snapshot, to obtain a prediction result. The prediction result is compared with the requirement set by the person skilled in the art, if the prediction result meets the requirement, scheduling is performed based on the matching power supply strategy.

[0049] If the prediction result does not meet the requirement, for example, if photovoltaic fluctuation causes energy storage to be unable to be charged in time, resulting in power supply risk in the subsequent business peak period, the matching power supply strategy needs to be fine-tuned. For example, the matching strategy is photovoltaic power supply 2.5kW, energy storage charging 0.5kW, and network maintaining full speed. In the first round of random adjustment, the following candidate strategies are generated: energy storage is changed to discharge 0.3kW, UPS pre-input threshold is increased by 5%, network low-priority business load limiting is 10%, and photovoltaic output fluctuation smoothing parameter is adjusted to be high.

[0050] Similarly, the real-time station operation memory snapshot is predicted for network resource collaborative scheduling by using the collaborative scheduling predictor to obtain multiple adjustment prediction results. It is determined whether there is an adjustment prediction result meeting the requirement in the multiple adjustment prediction results. If yes, the optimal solution in the multiple adjustment prediction results is taken as the adjustment direction, and then a preset amplitude of adjustment is performed on the matching power supply strategy based on the adjustment direction. For example, multiple direction adjustment strategies are generated with an amplitude of 10%, such as load limiting by 10%, 15%, and 20%, and scheduling prediction is performed again. Assuming that the prediction effect of load limiting by 15% is significantly better than the previous optimal solution in this round, the load limiting by 15% is determined as the target power supply strategy.

[0051] Thus, the technical effect of avoiding the adaptability problem caused by directly using the historical strategy through prediction verification and ensuring that the strategy can adapt to the micro-environmental changes by combining random disturbance and directional fine-tuning is achieved.

[0052] In the second embodiment, based on the same inventive concept as the network resource collaborative scheduling method for photovoltaic power generation in the foregoing embodiments, as shown in FIG. 11, the present application provides a network resource collaborative scheduling system for photovoltaic power generation. The system and method embodiments in the present application are based on the same inventive concept. The system includes: Figure 2 The memory snapshot collection module 11 is configured to construct a working memory snapshot mechanism, collect working memory snapshots of the photovoltaic power generation system, the energy storage system, the uninterruptible power supply, and the network load according to a preset period, and obtain real-time station operation memory snapshots. The memory bank construction module 12 is configured to construct a long-term collaborative scheduling memory bank based on a historical station operation memory snapshot set and a corresponding historical power supply strategy set. The long-term collaborative scheduling memory bank includes multiple collaborative scheduling memories, and each collaborative scheduling memory has a scene tag. The matching collaborative scheduling memory determination module 13 is configured to perform matching identification on the long-term collaborative scheduling memory bank according to the real-time station operation memory snapshot and in combination with the scene tag, and determine a matching collaborative scheduling memory. The collaborative scheduling module 14 is configured to perform network resource collaborative scheduling prediction on the real-time station operation memory snapshot based on a matching power supply strategy in the matching collaborative scheduling memory. If the prediction result meets the requirement, the matching power supply strategy is used for scheduling.

[0053] Further, the memory snapshot collection module 11 is configured to perform the following steps: The photovoltaic power generation system, the energy storage system, the uninterruptible power supply, and the network load are aligned according to the same clock reference. ​The state parameters of the photovoltaic power generation system, the energy storage system, the uninterruptible power supply and the network load in the working memory snapshot mechanism are called, and the data of the photovoltaic power generation system, the energy storage system, the uninterruptible power supply and the network load are collected according to a preset period to obtain a real-time station operation memory snapshot; The real-time station operation memory snapshot includes a photovoltaic side operation situation, an energy storage side operation situation, a power supply side operation situation and a network side operation situation.

[0054] Further, the memory library construction module 12 is configured to perform the following steps: Based on the clustering of the set of historical station operation memory snapshots, a plurality of clustering sets of historical station operation memory snapshots are obtained; According to the plurality of clustering sets of historical station operation memory snapshots, scene label identification is performed to determine a plurality of scene labels; According to the plurality of clustering sets of historical station operation memory snapshots, the set of historical power supply strategies is mapped and clustered to obtain a plurality of clustering sets of historical power supply strategies; The plurality of clustering sets of historical power supply strategies are subjected to cooperative scheduling memory identification to obtain a plurality of cooperative scheduling memories; The plurality of scene labels are used to identify the plurality of cooperative scheduling memories to construct the long-term cooperative scheduling memory library.

[0055] Further, the memory library construction module 12 is configured to perform the following steps: The plurality of clustering sets of historical station operation memory snapshots are subjected to scene feature extraction to obtain a plurality of scene feature sets; The plurality of scene feature sets are subjected to non-repeated interaction enhancement between two scene features in the set to obtain a plurality of interaction-enhanced scene feature sets; The plurality of interaction-enhanced scene feature sets are subjected to mean value processing to obtain the plurality of scene labels.

[0056] Further, the memory library construction module 12 is configured to perform the following steps: The historical power supply strategies with the highest similarity to other historical power supply strategies in the plurality of clustering sets of historical power supply strategies are extracted as a plurality of first cooperative scheduling memories; The historical power supply strategies in the plurality of clustering sets of historical power supply strategies that have a similarity within a preset similarity threshold to the plurality of first cooperative scheduling memories are summarized to construct a plurality of first cooperative scheduling memory spaces; The number of historical power supply strategies in the plurality of first cooperative scheduling memory spaces is counted to obtain a plurality of first memory quantities; With the plurality of first collaborative scheduling memory spaces as memory identification constraints, the plurality of first collaborative scheduling memories are iteratively updated in combination with the plurality of first memory quantities, to obtain a plurality of collaborative scheduling memories.

[0057] Further, the memory library construction module 12 is configured to perform the following steps: With the plurality of first collaborative scheduling memory spaces as memory identification constraints, a historical power supply strategy is randomly extracted in the plurality of first collaborative scheduling memory spaces respectively as a plurality of historical power supply strategies to be analyzed; The number of historical power supply strategies in the plurality of clustered historical power supply strategy sets that have a similarity to the plurality of historical power supply strategies to be analyzed within a preset similarity threshold is counted, to obtain a plurality of second memory quantities; It is determined whether the plurality of second memory quantities are greater than or equal to the plurality of first memory quantities. If yes, the plurality of historical power supply strategies to be analyzed are taken as a plurality of second collaborative scheduling memories, and the plurality of second collaborative scheduling memories are iteratively updated in combination with the plurality of second memory quantities, with the plurality of first collaborative scheduling memory spaces as memory identification constraints, until a preset number of updates is met, to obtain a plurality of collaborative scheduling memories.

[0058] Further, if no, a historical power supply strategy is randomly extracted again in the plurality of first collaborative scheduling memory spaces respectively with the plurality of first collaborative scheduling memory spaces as memory identification constraints, and iteratively updated in combination with the plurality of first memory quantities, until a preset number of reselections is met, to obtain a plurality of collaborative scheduling memories.

[0059] Further, the system is further configured to perform the following functions: If the prediction result does not meet the requirement, the matching power supply strategy is randomly adjusted in a random direction for multiple times to obtain a plurality of adjusted power supply strategies; Based on the plurality of adjusted power supply strategies, network resource collaborative scheduling prediction is performed on the real-time station operation memory snapshot to obtain a plurality of adjusted prediction results; It is determined whether there is an adjusted prediction result that meets the requirement in the plurality of adjusted prediction results. If yes, an adjustment direction is obtained, the matching power supply strategy is adjusted and screened according to a preset amplitude to obtain a target power supply strategy, and the target power supply strategy is used for scheduling.

[0060] Further, the system is further configured to perform the following functions: The optimal solution in the plurality of adjusted prediction results is taken as the adjustment direction; The matching power supply strategy is adjusted in the adjustment direction according to a preset amplitude to obtain a plurality of direction-adjusted matching power supply strategies; The real-time station operation memory snapshot is predicted for network resource cooperative scheduling based on the direction adjustment matching power supply strategy, and a plurality of direction adjustment prediction results are obtained. When the plurality of direction adjustment prediction results exist direction adjustment prediction results better than the optimal solution in the plurality of adjustment prediction results, the direction adjustment matching power supply strategy corresponding to the optimal solution in the plurality of direction adjustment prediction results is taken as the target power supply strategy.

[0061] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. And the above describes a specific embodiment of the present application. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or may be advantageous.

[0062] The above only describes the preferred embodiments of the present application, and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0063] The specification and drawings are merely exemplary of the present application, and any and all modifications, variations, combinations or equivalents that fall within the scope of the present application are considered to be covered by the present application. Obviously, those skilled in the art can make various modifications and changes to the present application without departing from the scope of the present application. Thus, if these modifications and changes of the present application belong to the scope of the present application and its equivalent technology, the present application is intended to include these modifications and changes.

Claims

1. A network resource collaborative scheduling method for photovoltaic power generation, characterized in that, The method comprises: A working memory snapshot mechanism is constructed, and working memory snapshots of a photovoltaic power generation system, an energy storage system, an uninterruptible power supply, and a network load are collected according to a preset period to obtain a real-time site operation memory snapshot; Based on a historical site operation memory snapshot set and a corresponding historical power supply strategy set, a long-term collaborative scheduling memory library is constructed, wherein the long-term collaborative scheduling memory library comprises a plurality of collaborative scheduling memories, and each collaborative scheduling memory has a scene tag; According to the real-time site operation memory snapshot, the long-term collaborative scheduling memory library is matched and identified in combination with the scene tag to determine a matched collaborative scheduling memory; Based on a matched power supply strategy in the matched collaborative scheduling memory, network resource collaborative scheduling prediction is performed on the real-time site operation memory snapshot, and if the prediction result meets the requirement, scheduling is performed based on the matched power supply strategy. 2.The network resource collaborative scheduling method for photovoltaic power generation of claim 1, wherein, A working memory snapshot mechanism is constructed, and working memory snapshots of a photovoltaic power generation system, an energy storage system, an uninterruptible power supply, and a network load are collected according to a preset period to obtain a real-time site operation memory snapshot, comprising: Aligning the photovoltaic power generation system, the energy storage system, the uninterruptible power supply, and the network load according to the same clock reference; Retrieving state parameters of the photovoltaic power generation system, the energy storage system, the uninterruptible power supply, and the network load in the working memory snapshot mechanism, collecting data of the photovoltaic power generation system, the energy storage system, the uninterruptible power supply, and the network load according to the preset period to obtain the real-time site operation memory snapshot; The real-time site operation memory snapshot comprises a photovoltaic side operation situation, an energy storage side operation situation, a power supply side operation situation, and a network side operation situation. 3.The network resource collaborative scheduling method for photovoltaic power generation of claim 1, wherein, Based on a historical site operation memory snapshot set and a corresponding historical power supply strategy set, a long-term collaborative scheduling memory library is constructed, wherein the long-term collaborative scheduling memory library comprises a plurality of collaborative scheduling memories, and each collaborative scheduling memory has a scene tag, comprising: Based on the historical site operation memory snapshot set, a plurality of clustered historical site operation memory snapshot sets are obtained through clustering division; According to the plurality of clustered historical site operation memory snapshot sets, a plurality of scene tags are determined through scene tag identification; According to the plurality of clustered historical site operation memory snapshot sets, a plurality of clustered historical power supply strategy sets are obtained through mapping clustering of the historical power supply strategy set; The plurality of collaborative scheduling memories are identified through collaborative scheduling memory identification of the plurality of clustered historical power supply strategy sets; The plurality of collaborative scheduling memories are identified through collaborative scheduling memory identification of the plurality of clustered historical power supply strategy sets. 4.The method for network resource coordinated scheduling for photovoltaic power generation of claim 3, wherein, According to the plurality of clustered historical site operation memory snapshot sets, a plurality of scene tags are determined through scene tag identification, comprising: Scene feature sets are obtained through scene feature extraction of the plurality of clustered historical site operation memory snapshot sets respectively; A plurality of interactive enhanced scene feature sets are obtained through two-by-two scene feature non-repeated interaction enhancement within the set of the plurality of scene feature sets respectively; The plurality of scene tags are obtained through mean value processing of the plurality of interactive enhanced scene feature sets.

5. The network resource collaborative scheduling method for photovoltaic power generation of claim 3, wherein, The multiple clustering historical power supply strategy sets are cooperatively scheduled and memory-identified to obtain multiple cooperative scheduling memories, including: Respectively extract the historical power supply strategies with the highest similarity to other historical power supply strategies in the multiple clustering historical power supply strategy sets as multiple first cooperative scheduling memories; Aggregate the historical power supply strategies in the multiple clustering historical power supply strategy sets with the similarity to the multiple first cooperative scheduling memories within a preset similarity threshold to construct multiple first cooperative scheduling memory spaces; Count the number of historical power supply strategies in the multiple first cooperative scheduling memory spaces to obtain multiple first memory quantities; Iteratively update the multiple first cooperative scheduling memories in combination with the multiple first memory quantities under the memory identification constraint of the multiple first cooperative scheduling memory spaces to obtain multiple cooperative scheduling memories.

6. The network resource collaborative scheduling method for photovoltaic power generation of claim 5, wherein, Iteratively update the multiple first cooperative scheduling memories in combination with the multiple first memory quantities under the memory identification constraint of the multiple first cooperative scheduling memory spaces to obtain multiple cooperative scheduling memories, including: Respectively randomly extract a historical power supply strategy in the multiple first cooperative scheduling memory spaces as multiple to-be-analyzed historical power supply strategies under the memory identification constraint of the multiple first cooperative scheduling memory spaces; Count the number of historical power supply strategies in the multiple clustering historical power supply strategy sets with the similarity to the multiple to-be-analyzed historical power supply strategies within a preset similarity threshold to obtain multiple second memory quantities; Determine whether the multiple second memory quantities are greater than or equal to the multiple first memory quantities, if yes, take the multiple to-be-analyzed historical power supply strategies as multiple second cooperative scheduling memories, and continue to iteratively update the multiple second cooperative scheduling memories in combination with the multiple second memory quantities under the memory identification constraint of the multiple first cooperative scheduling memory spaces until a preset update number is met to obtain multiple cooperative scheduling memories.

7. The network resource collaborative scheduling method for photovoltaic power generation of claim 6, wherein, If not, again extract a historical power supply strategy in the multiple first cooperative scheduling memory spaces under the memory identification constraint of the multiple first cooperative scheduling memory spaces, and iteratively update the multiple first memory quantities until a preset re-selection number is met to obtain multiple cooperative scheduling memories. 8.The method for network resource coordinated scheduling for photovoltaic power generation of claim 1, wherein, If the prediction result does not meet the requirement, perform multiple random adjustments on the matching power supply strategy in a random direction to obtain multiple adjusted power supply strategies; Perform network resource cooperative scheduling prediction on the real-time station operation memory snapshot based on the multiple adjusted power supply strategies to obtain multiple adjusted prediction results; Determine whether there is an adjusted prediction result meeting the requirement in the multiple adjusted prediction results, if yes, obtain an adjustment direction, adjust and screen the matching power supply strategy according to a preset amplitude to obtain a target power supply strategy, and perform scheduling based on the target power supply strategy. 9.The network resource collaborative scheduling method for photovoltaic power generation of claim 8, wherein, If not, obtain an adjustment direction, adjust and screen the matching power supply strategy according to a preset amplitude to obtain a target power supply strategy, including: Take the optimal solution in the multiple adjusted prediction results as the adjustment direction; Adjust the matching power supply strategy according to a preset amplitude to obtain multiple direction-adjusted matching power supply strategies. The real-time station operation memory snapshot is predicted for network resource collaborative scheduling based on the direction adjustment matching power supply strategy, and a plurality of direction adjustment prediction results are obtained; When the plurality of direction adjustment prediction results exist a direction adjustment prediction result better than the optimal solution in the plurality of adjustment prediction results, the direction adjustment matching power supply strategy corresponding to the optimal solution in the plurality of direction adjustment prediction results is taken as the target power supply strategy.

10. A network resource collaborative scheduling system for photovoltaic power generation, characterized in that, The system is used to implement the network resource collaborative scheduling method for photovoltaic power generation according to any one of claims 1-9, and the system comprises: A memory snapshot acquisition module is configured to construct a working memory snapshot mechanism, acquire working memory snapshots of a photovoltaic power generation system, an energy storage system, an uninterruptible power supply and a network load according to a preset period, and obtain a real-time station operation memory snapshot; A memory library construction module is configured to construct a long-term collaborative scheduling memory library based on a historical station operation memory snapshot set and a corresponding historical power supply strategy set, wherein the long-term collaborative scheduling memory library comprises a plurality of collaborative scheduling memories, and each collaborative scheduling memory has a scene tag; A matching collaborative scheduling memory determination module is configured to match and identify the long-term collaborative scheduling memory library according to the real-time station operation memory snapshot and in combination with the scene tag, and determine a matching collaborative scheduling memory; A collaborative scheduling module is configured to predict the real-time station operation memory snapshot for network resource collaborative scheduling based on a matching power supply strategy in the matching collaborative scheduling memory, and schedule based on the matching power supply strategy if the prediction result meets the requirements.

Images