Intelligent park device ad hoc network energy consumption ai optimization management system

Abstract

The application belongs to the technical field of energy management, and discloses an intelligent park equipment ad hoc network energy consumption AI optimization management system; including: collecting multi-source operation data of each device node in the intelligent park equipment ad hoc network in real time and performing data cleaning to obtain an energy consumption management dataset; performing a preheating stage to identify and generate a set of adjustable resource parameters; performing recovery event correlation analysis and regional peak value determination correction to obtain regional effective demand data; constructing a multi-device collaborative candidate strategy table and performing segmented benefit attribution determination to obtain a target collaborative control strategy; performing false effective section identification and re-marking to obtain a strategy execution log; constructing an updated energy consumption optimization management strategy and sending it to the corresponding device node; the accuracy of park energy consumption scheduling, the controllability of strategy execution and the overall energy saving optimization effect are improved.

Description

Technical Field

[0001] This application relates to the field of energy management technology, and more specifically, to a smart park equipment self-organizing network energy consumption AI optimization management system. Background Technology

[0002] Energy consumption optimization management in the context of self-organizing networks within smart parks is a crucial component of smart park operation and management. Due to the large number of devices and complex operating scenarios within the park, comprehensive management of multi-source information is necessary to achieve scheduling, control, and optimization of overall park energy consumption. Especially in self-organizing network environments, the communication links and response timing of device nodes are often unstable, making it difficult for traditional optimization management systems to form an effective management mechanism for resource allocation and strategy optimization applicable to the size of the park.

[0003] In real-world multi-device ad hoc network scenarios, the preheating power, startup impact, and short-term fluctuations of restored online device nodes after reconnecting to the ad hoc network are often directly included in the stable load statistics. This causes the stable load profile to deviate from the actual operating state, thus affecting the subsequent data calculation process. Furthermore, when multiple device nodes in the same area are restored to online status in a short period of time, traditional optimization management systems are prone to misjudging the restorative load as the actual peak or abnormal load, thereby triggering unnecessary peak shaving or load limiting actions, affecting the continuity of park services. Moreover, in scheduling scenarios close to the peak-valley electricity price switching boundary, traditional optimization management systems usually assume that all devices take effect simultaneously, without fully considering the response time difference and the step-by-step load reduction process between different devices in the ad hoc network environment. This leads to distorted attribution of segmented benefits and inconsistencies between strategy ranking results and actual benefits. The transition period between the issuance of control commands and the actual effect of devices in traditional optimization management systems is often directly regarded as the policy effective sample, causing shadow period data to pollute the training and execution evaluation results of the model, reducing the adaptability and update reliability of subsequent optimization strategies, resulting in reduced energy consumption scheduling accuracy for device ad hoc networks in smart parks and situations where generated strategies are not compatible with devices.

[0004] In view of this, this application proposes a smart park equipment self-organizing network energy consumption AI optimization management system to solve the above problems. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the existing technology and to achieve the above objectives, this application provides the following technical solution: a smart park equipment self-organizing network energy consumption AI optimization management system, comprising: a data acquisition module, which collects multi-source operating data of each device node in the smart park equipment self-organizing network in real time and performs data cleaning to obtain an energy consumption management dataset, including energy consumption parameters and status parameters.

[0006] The stable load identification module identifies and restores online device nodes based on status parameters and performs preheating phase identification on the device nodes to obtain stable load profile data; based on the stable load profile data, it generates a set of adjustable resource parameters.

[0007] The recovery analysis module performs recovery event correlation analysis based on the set of adjustable resource parameters, energy consumption parameters, and status parameters, and outputs recovery load filtering results. Based on these recovery load filtering results, regional peak value determination correction is performed to obtain regional effective demand data.

[0008] The power resource allocation module acquires power information from the smart park and constructs a multi-device collaborative candidate strategy table by combining the set of adjustable resource parameters and regional effective demand data. Based on each strategy in the multi-device collaborative candidate strategy table, a stepped load reduction curve is generated. Based on the stepped load reduction curve, a segmented revenue attribution determination is performed to obtain the target collaborative control strategy.

[0009] The strategy execution optimization module designs a sequence of control instructions based on the target collaborative control strategy and collects control instruction parameters in real time; it identifies false effective sections based on the control instruction parameters and status parameters to obtain the strategy effective boundary constraints; and it re-marks the false effective sections based on the strategy effective boundary constraints to obtain the strategy execution log.

[0010] The final strategy update module updates the target collaborative control strategy and energy management dataset based on the strategy execution log, constructs an updated energy optimization management strategy, and sends it to the corresponding device nodes; the modules are connected to each other via wired and / or wireless means.

[0011] Furthermore, the methods for performing the warm-up phase identification include: identifying the network connection marker belonging to each device node in the status parameters, as well as the most recent communication end timestamp of the corresponding device node.

[0012] When the network connection marker of the device node changes from disconnected to connected, the time interval between the start timestamp of the connection state and the end timestamp of the most recent communication is calculated. If the time interval is greater than the preset retention threshold, the corresponding device node is determined to be a restored online device node.

[0013] Identify the device type of the restored online device node and obtain the stable power range corresponding to that device type; at the same time, collect the energy consumption power of continuous sampling points after the connection status start timestamp in real time to form an initial energy consumption sequence.

[0014] Each sampling point in the initial energy consumption sequence is compared with the stable power range one by one. A preset number of consecutive sampling points that are all within the stable power range are selected, and the first sampling point among them is taken as the steady-state switching point. The part between the start timestamp of the initial energy consumption sequence and the steady-state switching point is removed to obtain the updated energy consumption sequence.

[0015] Identify the corresponding load level label of the updated energy consumption sequence and mark the updated energy consumption sequence to obtain the stable load profile data of the corresponding online device node.

[0016] Furthermore, the method for generating the set of adjustable resource parameters includes: matching the rated power upper limit and safe power lower limit of each recovered online device node based on the load level label of that recovered online device node.

[0017] Calculate the power mean of the updated energy consumption sequence in the stable load profile data, and calculate the difference between the rated power upper limit and the safe power lower limit and the power mean to obtain the increase adjustment margin and decrease adjustment margin respectively.

[0018] Identify the network layer to which each recovered online device node belongs in the corresponding device ad hoc network, and calculate the response deviation value based on the network layer and the corresponding preset layer delay.

[0019] The adjustment margins (increase and decrease) and response deviation values ​​are integrated and matched with the corresponding online recovery device nodes to obtain a set of adjustable resource parameters.

[0020] Furthermore, the method for performing recovery event correlation analysis includes: identifying the self-organizing network area marker corresponding to each recovery online device node in the status parameters, and merging recovery online device nodes with the same self-organizing network area marker into the same area device node cluster.

[0021] Set a statistical time window to count the number of times the connection status start timestamp appears in a single regional device node cluster within the statistical time window; when the number of occurrences exceeds a preset recovery quantity threshold, mark the corresponding regional device node cluster as a batch recovery cluster.

[0022] A load prediction model is constructed, and this model is used to perform baseline extrapolation on each online recovery device node in the batch recovery cluster to obtain the corresponding historical reference power.

[0023] The difference between the real-time power and the historical reference power of each online recovery device node in the batch recovery cluster is calculated to obtain the single-point recovery fluctuation amount; the single-point recovery fluctuation amounts of all online recovery device nodes in the batch recovery cluster are summed to obtain the regional recovery load interference amount; the regional recovery load interference amount is associated with the corresponding ad hoc network area to obtain the recovery load filtering result.

[0024] Furthermore, the method for correcting the regional peak value determination includes: extracting the total regional load of the corresponding self-organizing network area from the energy consumption parameters, calculating the difference between the total regional load and the regional recovery load interference, and obtaining the regional corrected load of the self-organizing network area.

[0025] Obtain the peak load alarm threshold corresponding to each self-organizing network area, and compare the area-corrected load with the peak load alarm threshold.

[0026] If the area's corrected load is not greater than the load peak alarm threshold, the corresponding self-organizing network area is determined to be in a load safe state; if the area's corrected load is greater than the load peak alarm threshold, the difference between the area's corrected load and the load peak alarm threshold is calculated to obtain the actual area load reduction.

[0027] If the self-organizing network area is in a load safe state, the corresponding real area load reduction is zero; integrate the real area load reduction of each self-organizing network area to obtain the effective demand data of the area.

[0028] Furthermore, the method of constructing a multi-device collaborative candidate strategy table includes: using a preset device combination optimization model to combine and search the online recovery device nodes corresponding to each self-organizing network area to generate multiple candidate device groups, and at the same time counting the total number of scheduling nodes in each candidate device group.

[0029] Calculate the sum of the reduction adjustment margins of all restored online device nodes in each candidate device group to obtain the total combined load reduction of that candidate device group.

[0030] Identify the maximum number of concurrent regulating nodes corresponding to each self-organizing network area in the power information of the smart park; select candidate equipment groups whose total load reduction is not less than the actual load reduction of the corresponding self-organizing network area and whose total number of scheduling nodes is not greater than the maximum number of concurrent regulating nodes, as effective collaborative equipment combinations.

[0031] The device identifiers and related parameters of the online recovery device nodes corresponding to each effective collaborative device combination are integrated and stored in a table to obtain a multi-device collaborative candidate strategy table.

[0032] Furthermore, the method for determining the segmented revenue allocation includes: identifying the response deviation value of each restored online device node in each effective collaborative device combination in the multi-device collaborative candidate strategy table and reducing the adjustment margin.

[0033] The timestamp of the preset strategy is summed with the response deviation value of each restored online device node to obtain the actual effective timestamp of each restored online device node.

[0034] Based on the order of the actual effective timestamps, the reduction adjustment margins of each restored online device node are accumulated, and a stepped load reduction curve for the corresponding effective collaborative device combination is plotted.

[0035] Extract the time-of-use electricity price range corresponding to each scheduling cycle from the power information of the smart park; divide the tiered load reduction curve into time intervals based on the time-of-use electricity price range to obtain the load reduction segment.

[0036] The load reduction amount for each load reduction segment is obtained by integral calculation. The actual power revenue of the load reduction segment is calculated based on the load reduction amount and the time-of-use price of the corresponding segment. The total power resource revenue of the corresponding effective coordinated equipment combination is obtained by summing the actual power revenue of all load reduction segments.

[0037] The total power resource revenue of all effective collaborative device combinations in the multi-device collaborative candidate strategy table is input into the preset strategy optimization model. The effective collaborative device combination that maximizes the revenue is then used as the target collaborative control strategy.

[0038] Furthermore, the method for identifying fake effective sections includes: identifying the target load reduction command corresponding to each restored online device node in the control command sequence, executing the application according to the preset command execution sequence, and recording the corresponding command execution timestamp.

[0039] Extract the instruction feedback data from the control instruction parameters and identify the instruction reception timestamp; using the instruction reception timestamp as the starting point, extract the real-time operating power at subsequent sampling times in the control instruction parameters and calculate the real-time power change rate.

[0040] Mark the first moment when the real-time power change rate exceeds the preset negative power change threshold as the actual load reduction response moment; use the instruction execution timestamp as the starting point of the fake effective segment and the actual load reduction response moment as the ending point of the fake effective segment; integrate the time interval of each fake effective segment as the policy effective boundary constraint for the corresponding online device node recovery.

[0041] Furthermore, the method for remarking the false effective section includes: combining the energy consumption parameters and status parameters corresponding to each restored online device node with the policy effective boundary constraints to obtain a transition parameter set; modifying the pre-determined policy effective markers in the transition parameter set to false effective transition state markers to obtain a remarked sample set.

[0042] The system acquires energy consumption parameter records collected after the actual load reduction response time, and calculates the command execution compliance rate based on the power at each time point and the expected power of the target load reduction command.

[0043] The re-marked sample set and instruction execution compliance rate corresponding to each recovered online device node are integrated, and the instruction execution timestamp and instruction reception timestamp are associated to generate a policy execution log.

[0044] Furthermore, the method of constructing an updated energy consumption optimization management strategy includes: using the relabeled sample set as feedback samples to input into the load prediction model, iteratively updating the parameters of the load prediction model; using the updated load prediction model to correct the set of adjustable resource parameters in the energy consumption management dataset, and adjusting the weights of the preset strategy optimization model in conjunction with the instruction execution compliance rate, and outputting an updated energy consumption optimization management strategy.

[0045] The technical effects and advantages of the Smart Park Equipment Self-Organizing Network Energy Consumption AI Optimization Management System proposed in this application are as follows: By collecting and cleaning multi-source operating data from each device node in the smart park equipment self-organizing network, an energy consumption management dataset is constructed, thereby achieving optimized management of energy consumption in the smart park equipment self-organizing network; compared with existing experience, by identifying and restoring online device nodes and removing preheating power segments, the pollution of the stable load profile by non-steady-state power in the early stage of restoration is eliminated; by constructing a restoration event correlation analysis model and extracting regional restoration load interference, the short-term load increase caused by batch restoration and the load increase of real business are distinguished. The system is designed to prevent restorative loads from being misjudged as peak loads, thus avoiding unnecessary peak shaving. By segmenting revenue allocation, it corrects the revenue miscalculation problem caused by asynchronous activation of multiple devices. By identifying false effective sections and remarking transitional samples, it avoids data during communication delays after command issuance being incorrectly included in policy effective samples. By iteratively updating the load prediction model and policy optimization model based on policy execution logs, it improves the accuracy of subsequent adjustable resource parameter corrections and the adaptability of collaborative control strategies, thereby enhancing the accuracy of park energy consumption scheduling, the controllability of policy execution, and the overall energy-saving optimization effect. Attached Figure Description

[0046] Figure 1 This is a schematic diagram of the smart park equipment self-organizing network energy consumption AI optimization management system of this application. Detailed Implementation

[0047] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application. Example

[0048] Please see Figure 1 As shown in this embodiment, the smart park equipment self-organizing network energy consumption AI optimization management system includes: a data acquisition module, which collects multi-source operating data of each device node in the smart park equipment self-organizing network and performs data cleaning to obtain an energy consumption management dataset, including energy consumption parameters and status parameters.

[0049] The stable load identification module identifies and restores online device nodes based on status parameters and performs preheating phase identification on the device nodes to obtain stable load profile data; based on the stable load profile data, it generates a set of adjustable resource parameters.

[0050] The recovery analysis module performs recovery event correlation analysis based on the set of adjustable resource parameters, energy consumption parameters, and status parameters, and outputs recovery load filtering results. Based on these recovery load filtering results, regional peak value determination correction is performed to obtain regional effective demand data.

[0051] The power resource allocation module acquires power information from the smart park and constructs a multi-device collaborative candidate strategy table by combining the set of adjustable resource parameters and regional effective demand data. Based on each strategy in the multi-device collaborative candidate strategy table, a stepped load reduction curve is generated. Based on the stepped load reduction curve, a segmented revenue attribution determination is performed to obtain the target collaborative control strategy.

[0052] The strategy execution optimization module designs a sequence of control instructions based on the target collaborative control strategy and collects control instruction parameters in real time; it identifies false effective sections based on the control instruction parameters and status parameters to obtain the strategy effective boundary constraints; and it re-marks the false effective sections based on the strategy effective boundary constraints to obtain the strategy execution log.

[0053] The final strategy update module updates the target collaborative control strategy and energy management dataset based on the strategy execution log, constructs an updated energy optimization management strategy, and sends it to the corresponding device nodes; the modules are connected to each other via wired and / or wireless means.

[0054] In this embodiment, the data acquisition terminal of the smart park device self-organizing network collects relevant data belonging to each device node in real time, and obtains a higher quality energy consumption management dataset through time alignment and filtering and noise reduction. The energy consumption parameters include parameters that can reflect the operating energy consumption level of the corresponding device node, such as the real-time operating power, cumulative power consumption, peak power, and load change sequence of each device node. The status parameters refer to parameters that reflect the network access status and control execution status of the corresponding device node, such as the online or offline status of each device node, network connection mark, network layer, and command feedback record.

[0055] The methods for performing the warm-up phase identification include: identifying the network connection flag belonging to each device node in the status parameters, and the timestamp of the most recent communication end for the corresponding device node.

[0056] The network connection flag is an identifier used to indicate whether a device node is in a disconnected, connected, or reconnected state; the most recent communication end timestamp is the timestamp when the corresponding device node ended its most recent communication with other nodes in the device ad hoc network.

[0057] When the network connection marker of the device node changes from disconnected to connected, the time interval between the start timestamp of the connection state and the end timestamp of the most recent communication is calculated. If the time interval is greater than the preset retention threshold, the corresponding device node is determined to be a restored online device node.

[0058] The connection state start timestamp refers to the starting time when a device node that has transitioned from a disconnected state to a connected state reconnects to the ad hoc network and resumes communication. The preset retention threshold is a threshold used to limit the offline duration of a device node. It is set based on the communication specifications of the corresponding ad hoc network. If the time interval is greater than the preset retention threshold, it indicates that the device node has been offline for a relatively long time and is a node that has truly recovered from an offline state rather than a node experiencing short-term network fluctuations. Therefore, it is determined to be a device node that has recovered to online status.

[0059] Identify the device type of the online device node to be restored, and obtain the stable power range corresponding to that device type.

[0060] In this embodiment, the equipment types include, for example, air conditioning equipment, lighting equipment, and control terminals. By using the equipment type as an index, the stable power range corresponding to the corresponding equipment type is searched from the equipment operation database based on the smart park equipment operation specifications. The stable power range is used to represent the power value range of the equipment node of the corresponding equipment type under normal and stable operation.

[0061] Simultaneously, the energy consumption power of continuous sampling points after the start timestamp of the connection status is collected in real time to form an initial energy consumption sequence.

[0062] The process involves using the connection state start timestamp as the starting point, collecting the power consumption of subsequent continuous sampling points, and sorting them according to the timestamp order to obtain the initial power consumption sequence. This sequence is used to completely record the power change process of the restored online device node from the start of reconnecting to the self-organizing network.

[0063] Each sampling point in the initial energy consumption sequence is compared with the stable power range one by one. A preset number of consecutive sampling points that are all within the stable power range are selected, and the first sampling point among them is taken as the steady-state switching point.

[0064] The preset continuous quantity refers to the threshold number of sampling points that continuously fall into the stable power range, which is used to ensure that the corresponding device node has truly entered stable operation. The specific value of this value is set based on the operation specifications of the corresponding type of equipment. By continuously comparing and identifying several continuous sampling points with stable power, the first sampling point is taken as the steady-state switching point, which is used to mark the critical moment when the online device node enters the stable operation stage from the preheating stage.

[0065] The updated energy consumption sequence is obtained by removing the portion between the start timestamp and the steady-state switching point of the initial energy consumption sequence.

[0066] The removed portion corresponds to the preheating power segment of online device nodes after reconnecting to the ad hoc network. This segment may include power surges or short-term unstable fluctuations during device startup; failure to remove it would introduce characteristics of unsteady-state power, thus affecting the accuracy of subsequent operations. By removing this portion, an updated energy consumption sequence reflecting only the stable operating state of the equipment is obtained.

[0067] Identify the corresponding load level label of the updated energy consumption sequence and mark the updated energy consumption sequence to obtain the stable load profile data of the corresponding online device node.

[0068] By comparing the energy consumption changes in the updated energy consumption sequence with the theoretical operating conditions set based on the operating specifications of the corresponding type of equipment, the load level of the updated energy consumption sequence under the current stable operating state is identified, and the corresponding load level label is output. By associating the updated energy consumption sequence with the load level label, data reflecting the power changes and load characteristics of the restored online equipment nodes after removing the preheating power is obtained, which serves as stable load profile data.

[0069] The methods for generating adjustable resource parameter sets include: matching the rated power upper limit and safe power lower limit of each recoverable online device node based on its load level label.

[0070] The corresponding power upper and lower limits are matched based on the load level label of the corresponding online recovery device node; the rated power upper limit is used to represent the maximum stable power value of the corresponding online recovery device node under the current load level; the safe power lower limit is used to represent the minimum stable power value that the corresponding online recovery device node is allowed to maintain under the current load level.

[0071] Calculate the power mean of the updated energy consumption sequence in the stable load profile data, and calculate the difference between the rated power upper limit and the safe power lower limit and the power mean to obtain the increase adjustment margin and decrease adjustment margin respectively.

[0072] The average power value is used to represent the average power level of the restored online device node under stable operating conditions. The power margin for adjustment is obtained by calculating the difference between the rated power upper limit and the safe power lower limit and the average power value. Increasing the adjustment margin reflects the maximum power space that the corresponding device node can be adjusted upward, and decreasing the adjustment margin reflects the maximum power space that the device node can be adjusted downward.

[0073] Identify the network layer to which each recovered online device node belongs in the corresponding device ad hoc network, and calculate the response deviation value based on the network layer and the corresponding preset layer delay.

[0074] The network layer refers to the access depth of the corresponding online recovery device node in the ad hoc network. The network layer is identified by matching each online recovery device node with the network status records of the corresponding nodes in the ad hoc network control terminal. The preset layer delay is used to represent the time required for control commands to be transmitted from the control terminal to the corresponding device node under different network layers.

[0075] The formula for calculating the response deviation is: ;in This represents the response deviation value of a specific online recovery device node; This indicates the network layer of the corresponding restored online device node; This indicates the preset level latency of the restored online device node; the response deviation value is used to reflect the difference in command transmission latency between different restored online device nodes in an ad hoc network environment.

[0076] The adjustment margins (increase and decrease) and response deviation values ​​are integrated and matched with the corresponding online recovery device nodes to obtain a set of adjustable resource parameters.

[0077] The adjustable resource parameter set is used to uniformly store the up-adjustment capability, down-adjustment capability, and command response time difference of each recovery online device node, and obtain them by associating and matching them with the corresponding recovery online device node.

[0078] The methods for performing recovery event correlation analysis include: identifying the self-organizing network area marker corresponding to each recovery online device node in the status parameters, and merging recovery online device nodes with the same self-organizing network area marker into the same area device node cluster.

[0079] The self-organizing network area refers to the local area of ​​the self-organizing network in each corresponding partition of the smart park, which covers the entire smart park. For example, each floor or each unit building. Each local area of ​​the self-organizing network corresponds to a local communication topology range. By merging the restored online device nodes belonging to the same self-organizing network area, the device node cluster of that area is obtained.

[0080] Set a statistical time window to count the number of times the start timestamp of the connection status appears in a single area device node cluster within the statistical time window.

[0081] The specific size of the statistical time window is set based on the theoretical duration of the device recovery event. By counting the number of times a device node cluster in a certain area recovers from disconnection within a single statistical time window, it is used to identify whether there are multiple devices that have recovered online within a certain time interval.

[0082] When the number of occurrences exceeds the preset recovery quantity threshold, the corresponding regional device node cluster will be marked as a batch recovery cluster.

[0083] If the number of occurrences exceeds the preset recovery number threshold set based on the minimum recovery number required to meet the recovery event requirements under theoretical conditions, it indicates that there are multiple online recovery device nodes in the corresponding self-organizing network area that are centrally recovering communication or operation, which is a batch recovery phenomenon. Therefore, the corresponding device node cluster is marked as a batch recovery cluster.

[0084] A load prediction model is constructed, and this model is used to perform baseline extrapolation on each online recovery device node in the batch recovery cluster to obtain the corresponding historical reference power.

[0085] In this embodiment, historical operating data of the same device type, the same region, and similar operating time as the online recovery device nodes in the batch recovery cluster are obtained as a training set in the historical period. The existing machine learning model is used as the model basis, and the training set is used to train the machine learning model to obtain the load prediction model. The baseline extrapolation refers to using the load prediction model to predict the reference power of the corresponding online recovery device node under the condition that no recovery disturbance occurs, as a historical reference power.

[0086] The difference between the real-time power and the historical reference power of each recovery online device node in the batch recovery cluster is used to obtain the single-point recovery fluctuation.

[0087] Real-time power refers to the power value actually collected by the corresponding online recovery device node at the current moment. The single-point recovery fluctuation is obtained by calculating the difference between the real-time power and the corresponding historical reference power. The single-point recovery fluctuation is used to represent the additional power change of the corresponding online recovery device node relative to the historical reference power due to the recovery of online status or operation.

[0088] The regional recovery load interference is obtained by summing the single-point recovery fluctuations of all online recovery device nodes in the batch recovery cluster.

[0089] The formula for calculating the load disturbance in the calculation area is as follows: ;in This indicates the amount of regional recovery load interference for a specific batch recovery cluster; Indicates the number of the corresponding batch recovery clusters. The single-point recovery fluctuation of each online device node should be noted. If the value is negative, it should be set to 0 to avoid negative value interference. This indicates the number of online recovery device nodes in the corresponding batch recovery cluster; the recovery load interference in this area is used to represent the total load fluctuation caused by the centralized recovery of multiple online recovery device nodes in the same batch recovery cluster.

[0090] The load restoration interference in this area is correlated with the corresponding self-organizing network area to obtain the load restoration filtering results.

[0091] The methods for correcting regional peak load determination include: extracting the total regional load of the corresponding self-organizing network area from the energy consumption parameters, calculating the difference between the total regional load and the regional recovery load interference, and obtaining the corrected regional load of the self-organizing network area.

[0092] The total regional load refers to the overall energy consumption level of all device nodes in the corresponding self-organizing network area at the current moment. By subtracting the total regional load from the regional recovery load interference, and deducting the recovery load interference caused by batch recovery, the regional corrected load of the corresponding self-organizing network area is obtained, which represents the remaining effective energy consumption level of the self-organizing network area.

[0093] Obtain the peak load alarm threshold corresponding to each self-organizing network area, and compare the area-corrected load with the peak load alarm threshold.

[0094] The load peak alarm threshold is a threshold set based on the safety load specification of the corresponding local area of ​​the self-organizing network, which is used to indicate the maximum safe load limit reached by the self-organizing network area during the current scheduling cycle.

[0095] If the corrected load in a region is not greater than the peak load alarm threshold, the corresponding self-organizing network region will be determined to be in a load safe state.

[0096] The process involves comparing the regional corrected load of each self-organizing network area with the corresponding self-organizing network area. If the regional corrected load is less than or equal to the load peak alarm threshold, it indicates that after deducting the load recovery interference, the actual load has not reached the level that requires triggering grid peak shaving or load limiting control, and is therefore determined to be in a load safe state.

[0097] If the corrected load for a region is greater than the peak load alarm threshold, the difference between the corrected load for that region and the peak load alarm threshold is calculated to obtain the actual load reduction for that region.

[0098] If the area correction load is greater than the load peak alarm threshold, it means that the area load of the self-organizing network area has exceeded the allowable limit. Therefore, the actual area load reduction is obtained by calculating the difference between the area correction load and the load peak alarm threshold, which is used to reflect the load reduction that needs to be performed in the self-organizing network area.

[0099] If the self-organizing network area is in a load safe state, the corresponding real area load reduction is zero; integrate the real area load reduction of each self-organizing network area to obtain the effective demand data of the area.

[0100] If a self-organizing network area is in a load safe state, it means that the self-organizing network area does not exceed the limit after deducting the restoration load interference. Therefore, the corresponding real area load reduction is set to 0. The effective demand data of the area is used to summarize the real load reduction demand that still exists in all self-organizing network areas after deducting the restoration disturbance, thereby eliminating the false demand caused by the restoration load.

[0101] The method of constructing a multi-device collaborative candidate strategy table includes: using a preset device combination optimization model to combine and search the online recovery device nodes corresponding to each self-organizing network area to generate multiple candidate device groups, and at the same time counting the total number of scheduling nodes in each candidate device group.

[0102] In this embodiment, the preset equipment combination optimization model refers to a model trained using samples of successfully running equipment combinations in historical records, as well as corresponding load conditions and power resource revenue data. This model is used to combine and search for several equipment combinations that represent the normal power resource allocation of the load, which are the candidate equipment groups. The total number of scheduling nodes is used to represent the number of equipment nodes that actually participate in coordinated regulation in each candidate equipment group.

[0103] Calculate the sum of the reduction adjustment margins of all restored online device nodes in each candidate device group to obtain the total combined load reduction of that candidate device group.

[0104] The combined load reduction total refers to the sum of the reduction adjustment margins of all restored online device nodes in the candidate device group, which is used to reflect the total load reduction capacity that the candidate device group can provide under the current scheduling conditions.

[0105] Identify the maximum number of concurrent adjustment nodes corresponding to each self-organizing network area in the power information of the smart park.

[0106] The maximum number of concurrent adjustment nodes is used to represent the upper limit of device nodes that are allowed to participate in adjustment simultaneously in the current scheduling cycle of the corresponding self-organizing network area. This upper limit is set by the line safety bearing specification in the smart park power information.

[0107] Select a group of candidate devices whose total load reduction is not less than the actual load reduction of the corresponding self-organizing network area and whose total number of scheduling nodes is not greater than the maximum number of concurrent adjustment nodes, and use them as an effective collaborative device combination.

[0108] The requirement that the total combined load reduction is not less than the actual load reduction in the area is to ensure that the candidate equipment group has the ability to cover the actual load reduction demand of the current self-organizing network area after deducting restorative load interference. The requirement that the total number of scheduling nodes is not greater than the maximum number of concurrent adjustment nodes is to limit the candidate equipment group to meet the execution constraints of the park's power and communication layers. Only when both conditions are met simultaneously will the candidate equipment group be determined as an effective collaborative equipment combination, in order to exclude candidate equipment groups with insufficient load reduction capacity or too many nodes that lead to high execution risks.

[0109] The device identifiers and related parameters of the online recovery device nodes corresponding to each effective collaborative device combination are integrated and stored in a table to obtain a multi-device collaborative candidate strategy table.

[0110] The device identifier and related parameters refer to information such as device type, area label, reduction adjustment margin, response deviation value, load level label and network layer. The tabular storage is used to record the relevant data of effective collaborative device combinations in a structured manner to obtain a multi-device collaborative candidate strategy table.

[0111] The methods for determining the segmented revenue allocation include: identifying the response deviation values ​​of each restored online device node in each effective collaborative device combination in the multi-device collaborative candidate strategy table and reducing the adjustment margin; The timestamp of the preset strategy is summed with the response deviation value of each restored online device node to obtain the actual effective timestamp of each restored online device node.

[0112] The policy issuance timestamp is used to record the moment when control commands are sent to device nodes, and the actual effective timestamp is used to record the moment when device nodes begin to generate effective responses. By adding the policy issuance timestamp to the response deviation value, the communication delays and execution delays caused by different network layers are uniformly converted to the same time base, thereby determining the actual effective sequence of each restored online device node.

[0113] Based on the order of the actual effective timestamps, the reduction adjustment margins of each restored online device node are accumulated, and a stepped load reduction curve for the corresponding effective collaborative device combination is plotted.

[0114] The stepped load reduction curve is used to represent the cumulative load reduction trajectory of the effective coordinated equipment combination in the time dimension. Each step corresponds to the additional load reduction of a restored online equipment node at the actual effective timestamp. The reduction adjustment margin of each node is accumulated according to the order of the actual effective timestamps, so that the stepped load reduction curve can truly reflect the time-segmented load reduction formed by the gradual effectiveness of each restored online equipment node, and avoid misjudging the load reduction at different effective times as synchronous effectiveness.

[0115] Extract the time-of-use electricity price range corresponding to each scheduling cycle from the power information of the smart park. The time-of-use electricity price refers to the peak, normal and valley electricity prices corresponding to different time periods within the scheduling cycle of the power system.

[0116] Based on the time-of-use pricing range, the tiered load reduction curve is divided into time intervals to obtain load reduction segments. Specifically, by dividing the tiered load reduction curve according to the critical points of each interval in the time-of-use pricing range, curve segments corresponding to different time-of-use prices are obtained, which are the load reduction segments.

[0117] The load reduction amount for each load reduction segment is obtained by integral calculation, and the actual electricity revenue of the load reduction segment is calculated based on the load reduction amount and the time-of-use price of the corresponding segment.

[0118] The method involves integrating the local area of ​​each load reduction segment on the time axis based on mathematical knowledge to obtain the load reduction power of that segment. This load reduction power is used to represent the cumulative reduction in electricity consumption within that segment.

[0119] The formula for calculating the actual electricity revenue of the corresponding load shedding section is as follows: ;in, This represents the actual electricity revenue of a certain load reduction section; This indicates the amount of electricity reduced in the corresponding load reduction section; This represents the time-of-use electricity price for the corresponding load shedding zone; this revenue is used to represent the contribution to the reduction of electricity costs in the smart park within the corresponding load shedding zone.

[0120] The total power resource revenue of the corresponding effective coordinated equipment combination is obtained by summing the actual power revenue of all load reduction sections. The total power resource revenue is used to represent the cumulative revenue of the same effective coordinated equipment combination across multiple load reduction sections in the entire power system dispatch cycle.

[0121] The total power resource revenue of all effective collaborative device combinations in the multi-device collaborative candidate strategy table is input into the preset strategy optimization model. The effective collaborative device combination that maximizes the revenue is then used as the target collaborative control strategy.

[0122] In this embodiment, the preset strategy optimization model refers to a model trained using historical power resource allocation and electricity price information. The objective function of the model is to maximize the total power resource revenue. The model is used to predict the effective equipment combination that maximizes the total power resource revenue, and the combination and the parameters corresponding to each equipment node are encoded into a target collaborative control strategy that can be recognized by the control terminal.

[0123] The methods for identifying false effective sections include: In this embodiment, the target collaborative control strategy is used to represent the optimal collaborative device combination obtained after segmented benefit attribution determination, as well as its corresponding reduction margin and actual effective timestamp, etc.; the control command sequence is used to decompose the target collaborative control strategy into the command sequence corresponding to each restored online device node, as well as the issuance order and execution interval of each command, including, for example, load reduction command, start / stop command, and power adjustment command, etc.; the control command parameters refer to the feedback information recorded during the execution of each command, including, for example, command reception timestamp, command acknowledgment status, real-time operating power, power sampling timestamp, and command execution status, etc.

[0124] Identify the target load reduction command corresponding to each online recovery device node in the control command sequence, execute the application according to the preset command execution sequence, and record the corresponding command execution timestamp.

[0125] In this embodiment, the target load reduction command refers to the power regulation control content issued to the online recovery device node. The preset command execution sequence is used to specify the order in which each online recovery device node receives and executes various commands, and is obtained based on the relevant command execution specifications set in the control terminal. The command execution timestamp is used to record the timestamp when the target load reduction command is determined to be issued.

[0126] Extract the instruction feedback data from the control instruction parameters and identify the instruction reception timestamp.

[0127] The instruction feedback data is used to indicate the response status of the corresponding device node to the target load reduction instruction, including receipt confirmation, execution confirmation, or acknowledgment status; the instruction reception timestamp is used to record the moment when the device node confirms that it has received the target load reduction instruction.

[0128] Starting from the command reception timestamp, extract the real-time operating power of the control command parameters at subsequent sampling times and calculate the real-time power change rate.

[0129] The real-time operating power is used to represent the continuous power sampling value of the corresponding online device node after receiving the command, and the real-time power change rate is used to represent the direction and magnitude of the power change between adjacent sampling times. By taking the command receiving timestamp as the starting point and continuously extracting subsequent power sampling values, it is used to observe whether the device has a significant load reduction trend after receiving the control command.

[0130] The first moment when the real-time power change rate exceeds the preset negative power change threshold is marked as the actual load reduction response moment.

[0131] The preset negative power change threshold is set based on the parameter configuration of the corresponding device. This threshold is used to limit the judgment condition when the power of a certain online recovery device node drops to a preset magnitude. The actual load reduction response time is used to indicate the critical moment when the corresponding online recovery device node begins to produce an effective load reduction effect.

[0132] The instruction execution timestamp is used as the starting point of the sham effective segment, and the actual load reduction response time is used as the ending point of the sham effective segment.

[0133] The false effective period refers to the time interval during which the target load reduction command has been issued but the device node has not yet started to produce the actual load reduction effect. This time interval may include processes such as communication transmission, node confirmation, and execution delay. By extracting the time interval between the command execution timestamp and the actual load reduction response time separately, the power sampling data in this false effective period is prevented from contaminating other data.

[0134] The time interval of each fake effective period is integrated as the policy effective boundary constraint for the corresponding online device node to be restored.

[0135] Among them, the policy effective boundary constraint refers to the constraint used to limit which time segments are false effective segments and which time segments are real effective segments during the execution of the target load reduction command by the recovery online device node.

[0136] The methods for remarking false effective sections include: combining the energy consumption parameters and status parameters corresponding to each restored online device node with the policy effective boundary constraints to obtain a set of transition parameters.

[0137] The transition parameter set is obtained by combining and encapsulating energy consumption parameters, status parameters, and policy effective boundary constraints. This set is used to simultaneously reflect the device's power changes, command reception status, and effective boundary range.

[0138] The pre-determined policy activation flags in the transition parameter set are modified to false activation transition state flags, resulting in a re-labeled sample set.

[0139] Since the control terminal may collect samples in real time during this period and regard them as having taken effect, there is a pre-determined policy effective flag. The false effective transition state flag refers to the false effective state in which the samples during this period are actually in the communication delay or execution transition phase and have not truly entered the load reduction state. By modifying the original policy effective flag to the false effective transition state flag, the samples in the false effective section are separated from the real effective sample data, reducing the risk of data pollution.

[0140] The system acquires energy consumption parameter records collected after the actual load reduction response time, and calculates the command execution compliance rate based on the power at each time point and the expected power of the target load reduction command.

[0141] The system retrieves energy consumption parameter records starting from the actual load reduction moment to represent the power change of the corresponding online recovery device node after the actual load reduction begins; the formula for calculating the instruction execution compliance rate is as follows: ,in Command execution success rate; This indicates the number of sampling moments after the actual load reduction response time; Indicates the first Operating power at any given moment; This indicates the first [unit / item] set based on the load reduction target. The target load reduction command corresponds to the expected power at each moment; the closer the actual operating power is to the expected power, the higher the command execution success rate; the greater the deviation, the lower the command execution success rate.

[0142] The re-marked sample set and instruction execution compliance rate corresponding to each recovered online device node are integrated, and the instruction execution timestamp and instruction reception timestamp are associated to generate a policy execution log.

[0143] The strategy execution log is a complete backtracking record of policy execution, which is formed by structuring and integrating the relabeled sample set, instruction execution success rate, instruction execution timestamp, and instruction reception timestamp. This log is used to record the execution process, execution boundaries, sample relabeling results, and execution completion degree of the target collaborative control policy on the online device node.

[0144] The method of constructing an updated energy consumption optimization management strategy includes: using the relabeled sample set as feedback samples to input into the load prediction model, and iteratively updating the parameters of the load prediction model.

[0145] Feedback samples refer to sample data after being remarked through false effective sections, including the device identifier, status parameters, stable load profile data, policy effective boundary constraints, actual load reduction response time, and command execution compliance rate of the corresponding restored online device nodes. Iterative parameter updates for the load forecasting model involve inputting feedback samples and historical training samples into the model training process, and then reversing the model's prediction parameters based on the power deviation after the actual load reduction response. This allows the model to more accurately reflect the true stable operation of restored online device nodes in subsequent predictions, thereby reducing the contamination of model parameters by false effective samples.

[0146] The updated load forecasting model is used to correct the set of adjustable resource parameters in the energy consumption management dataset, and the weights of the preset strategy optimization model are adjusted in combination with the command execution compliance rate to output an updated energy consumption optimization management strategy.

[0147] The process involves recalculating the increase and decrease of adjustment margins and response deviation values ​​for online device nodes based on the updated baseline power projection results, and reassigning values ​​to the adjustable resource parameter set to achieve correction. When the command execution compliance rate changes, the ratio of benefit weight to execution deviation weight in the strategy optimization model is adjusted accordingly to ensure that the output updated energy consumption optimization management strategy can take into account load reduction benefits, execution reachability, and equipment coordination constraints.

[0148] This embodiment collects and cleans multi-source operational data from each device node in a smart park's self-organizing network, constructing an energy consumption management dataset to achieve optimized energy consumption management within the smart park's self-organizing network. Compared to existing experience, by identifying and removing online device nodes that have recovered and eliminating preheating power segments, the contamination of the stable load profile by unsteady power in the initial recovery phase is eliminated. By constructing a recovery event correlation analysis model and extracting regional recovery load interference, the short-term load increase caused by batch recovery is distinguished from the actual business load growth, avoiding the misjudgment of recovery load as peak load and thus triggering unnecessary peak shaving. By performing segmented revenue attribution, the revenue miscalculation problem caused by asynchronous activation of multiple devices is corrected. By identifying false effective segments and re-marking transitional samples, data during the communication delay period after command issuance is prevented from being incorrectly included in the policy effective samples. By iteratively updating the load prediction model and policy optimization model based on policy execution logs, the accuracy of subsequent adjustable resource parameter corrections and the adaptability of collaborative control strategies are improved, thereby enhancing the accuracy of park energy consumption scheduling, the controllability of policy execution, and the overall energy-saving optimization effect.

[0149] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0150] All formulas in this manual are dimensionless and calculated numerically. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0151] Although embodiments of this application have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of this application, the scope of which is defined by the claims and their equivalents.

Claims

1. A smart park equipment self-organizing network energy consumption AI optimization management system, characterized in that, include: The data acquisition module collects multi-source operating data from each device node in the smart park's self-organizing network in real time and performs data cleaning to obtain an energy management dataset, including energy consumption parameters and status parameters. The stable load identification module identifies and restores online device nodes based on status parameters and performs preheating phase identification on the device nodes to obtain stable load profile data; based on the stable load profile data, it generates a set of adjustable resource parameters. The recovery analysis module performs recovery event correlation analysis based on the set of adjustable resource parameters, energy consumption parameters, and status parameters, and outputs recovery load filtering results. Based on the results of this restorative load filtering, the regional peak value determination is corrected to obtain the effective demand data for the region. The power resource allocation module acquires power information from the smart park and constructs a multi-device collaborative candidate strategy table by combining the set of adjustable resource parameters and regional effective demand data; it then generates a stepped load reduction curve based on each strategy in the multi-device collaborative candidate strategy table. Based on the step load reduction curve, segmented revenue attribution is determined, and a target collaborative control strategy is obtained. The strategy execution optimization module designs a sequence of control instructions based on the target collaborative control strategy and collects control instruction parameters in real time; it identifies false effective sections based on the control instruction parameters and status parameters to obtain the strategy effective boundary constraints; and it re-marks the false effective sections based on the strategy effective boundary constraints to obtain the strategy execution log. The final strategy update module updates the target collaborative control strategy and energy management dataset based on the strategy execution log, constructs an updated energy optimization management strategy, and sends it to the corresponding device nodes; the modules are connected to each other via wired and / or wireless means.

2. The smart park equipment self-organizing network energy consumption AI optimization management system according to claim 1, characterized in that, The methods for identifying the preheating phase include: Identify the network connection marker belonging to each device node in the status parameters, as well as the timestamp of the most recent communication end for the corresponding device node; When the network connection marker of the device node changes from disconnected to connected, the time interval between the start timestamp of the connection state and the end timestamp of the most recent communication is calculated. If the time interval is greater than the preset retention threshold, the corresponding device node is determined to be a restored online device node. Identify the device type of the online device node to be restored and obtain the stable power range corresponding to the device type; at the same time, collect the energy consumption power of continuous sampling points after the connection status start timestamp in real time to form an initial energy consumption sequence. Each sampling point in the initial energy consumption sequence is compared with the stable power range one by one. A preset number of consecutive sampling points that are all within the stable power range are selected, and the first sampling point among them is taken as the steady-state switching point. The part between the start timestamp of the initial energy consumption sequence and the steady-state switching point is removed to obtain the updated energy consumption sequence. Identify the corresponding load level label of the updated energy consumption sequence and mark the updated energy consumption sequence to obtain the stable load profile data of the corresponding online device node.

3. The smart park equipment self-organizing network energy consumption AI optimization management system according to claim 2, characterized in that, The methods for generating the adjustable resource parameter set include: Match the rated power upper limit and safe power lower limit of each online recovery device node based on its load level label; Calculate the power mean of the updated energy consumption sequence in the stable load profile data, and calculate the difference between the rated power upper limit and the safe power lower limit and the power mean to obtain the increase adjustment margin and decrease adjustment margin respectively; Identify the network level of each online recovery device node in the corresponding device ad hoc network, and calculate the response deviation value based on the network level and the corresponding preset level latency; The adjustment margins (increase and decrease) and response deviation values ​​are integrated and matched with the corresponding online recovery device nodes to obtain a set of adjustable resource parameters.

4. The smart park equipment self-organizing network energy consumption AI optimization management system according to claim 3, characterized in that, The methods for performing recovery event correlation analysis include: Identify the self-organizing network area marker corresponding to each online recovery device node in the status parameters, and merge online recovery device nodes with the same self-organizing network area marker into the same area device node cluster; Set a statistical time window to count the number of times the connection status start timestamp appears in a single regional device node cluster within the statistical time window; when the number of occurrences exceeds a preset recovery quantity threshold, the corresponding regional device node cluster is marked as a batch recovery cluster; A load prediction model is constructed, and the model is used to perform baseline extrapolation on each online recovery device node in the batch recovery cluster to obtain the corresponding historical reference power. The difference between the real-time power and the historical reference power of each online recovery device node in the batch recovery cluster is calculated to obtain the single-point recovery fluctuation amount; the single-point recovery fluctuation amounts of all online recovery device nodes in the batch recovery cluster are summed to obtain the regional recovery load interference amount; the regional recovery load interference amount is associated with the corresponding ad hoc network area to obtain the recovery load filtering result.

5. The smart park equipment self-organizing network energy consumption AI optimization management system according to claim 4, characterized in that, The methods for correcting the regional peak value determination include: Extract the total regional load of the corresponding self-organizing network area from the energy consumption parameters, calculate the difference between the total regional load and the regional recovery load interference, and obtain the regional corrected load of the self-organizing network area. Obtain the peak load alarm threshold corresponding to each self-organizing network area, and compare the area-corrected load with the peak load alarm threshold. If the area's corrected load is not greater than the load peak alarm threshold, the corresponding self-organizing network area is determined to be in a load safe state; if the area's corrected load is greater than the load peak alarm threshold, the difference between the area's corrected load and the load peak alarm threshold is calculated to obtain the actual area load reduction. If the self-organizing network area is in a load safe state, the corresponding real area load reduction is zero; integrate the real area load reduction of each self-organizing network area to obtain the effective demand data of the area.

6. The smart park equipment self-organizing network energy consumption AI optimization management system according to claim 5, characterized in that, The methods for constructing the multi-device collaborative candidate strategy table include: Using a preset device combination optimization model, the online recovery device nodes corresponding to each self-organizing network area are combined and searched to generate multiple candidate device groups. At the same time, the total number of scheduling nodes in each candidate device group is counted. Calculate the sum of the reduction adjustment margins of all restored online device nodes in each candidate device group to obtain the total combined load reduction of that candidate device group; Identify the maximum number of concurrent adjustment nodes corresponding to each self-organizing network area in the power information of the smart park; select candidate equipment groups whose total load reduction is not less than the actual load reduction of the corresponding self-organizing network area and whose total number of scheduling nodes is not greater than the maximum number of concurrent adjustment nodes, as effective collaborative equipment combinations; The device identifiers and related parameters of the online recovery device nodes corresponding to each effective collaborative device combination are integrated and stored in a table to obtain a multi-device collaborative candidate strategy table.

7. The smart park equipment self-organizing network energy consumption AI optimization management system according to claim 6, characterized in that, The methods for determining the ownership of segmented revenue include: Identify the response deviation values ​​of each recovered online device node in each effective collaborative device combination in the multi-device collaborative candidate strategy table and reduce the adjustment margin; The timestamp of the preset strategy is issued is summed with the response deviation value of each online recovery device node to obtain the actual effective timestamp of each online recovery device node; Based on the order of the actual effective timestamps, the reduction adjustment margins of each restored online device node are accumulated, and a step-down curve for the corresponding effective collaborative device combination is plotted. Extract the time-of-use electricity price range corresponding to each scheduling cycle from the power information of the smart park; divide the tiered load reduction curve into time intervals based on the time-of-use electricity price range to obtain the load reduction segment; The load reduction amount for each load reduction segment is obtained by integral calculation. The actual power revenue of the load reduction segment is calculated based on the load reduction amount and the time-of-use price of the corresponding interval. The total power resource revenue of the corresponding effective coordinated equipment combination is obtained by summing the actual power revenue of all load reduction segments. The total power resource revenue of all effective collaborative device combinations in the multi-device collaborative candidate strategy table is input into the preset strategy optimization model. The effective collaborative device combination that maximizes the revenue is then used as the target collaborative control strategy.

8. The smart park equipment self-organizing network energy consumption AI optimization management system according to claim 7, characterized in that, The methods for identifying false effective segments include: Identify the target load reduction command corresponding to each online recovery device node in the control command sequence, execute the application according to the preset command execution sequence, and record the corresponding command execution timestamp; Extract the instruction feedback data from the control instruction parameters and identify the instruction reception timestamp; using the instruction reception timestamp as the starting point, extract the real-time operating power at subsequent sampling times in the control instruction parameters and calculate the real-time power change rate; Mark the first moment when the real-time power change rate exceeds the preset negative power change threshold as the actual load reduction response moment; use the instruction execution timestamp as the starting point of the fake effective segment and the actual load reduction response moment as the ending point of the fake effective segment; integrate the time interval of each fake effective segment as the policy effective boundary constraint for the corresponding online device node recovery.

9. The smart park equipment self-organizing network energy consumption AI optimization management system according to claim 8, characterized in that, The method for remarking false effective segments includes: The energy consumption parameters and status parameters corresponding to each restored online device node are combined with the policy effective boundary constraints to obtain a set of transition parameters; the pre-determined policy effective flags in the set of transition parameters are modified to false effective transition state flags to obtain a set of re-labeled samples. The energy consumption parameter records collected after the actual load reduction response time are obtained, and the command execution compliance rate is calculated based on the power at each time and the expected power of the target load reduction command. The re-marked sample set and instruction execution compliance rate corresponding to each recovered online device node are integrated, and the instruction execution timestamp and instruction reception timestamp are associated to generate a policy execution log.

10. The smart park equipment self-organizing network energy consumption AI optimization management system according to claim 9, characterized in that, The methods for constructing and updating energy consumption optimization management strategies include: The relabeled sample set is used as a feedback sample input to the load prediction model, and the parameters of the load prediction model are iteratively updated. The updated load prediction model is used to correct the set of adjustable resource parameters in the energy consumption management dataset, and the weights of the preset strategy optimization model are adjusted in combination with the instruction execution compliance rate, and an updated energy consumption optimization management strategy is output.

Images