Multi-target-oriented power grid dispatching decision-making method and system

Abstract

The invention discloses a multi-target-oriented power grid dispatching decision-making method and system, and relates to the technical field of power grid dispatching, and the method comprises the steps: constructing a normal dispatching target through minimum load fluctuation, constructing an abnormal weather dispatching target through minimum dispatching amount and minimum dispatching loss, constructing a device abnormal dispatching target through minimum switching loss, and constructing a device abnormal dispatching target through minimum switching loss; combining a normal scheduling target, an abnormal weather scheduling target and a device abnormal scheduling target according to the initial priority weight to construct a multi-target optimization function; obtaining a probability prediction value according to the regional historical load data, the regional real-time load data, the regional historical weather data, the regional real-time weather data, the regional historical device maintenance data and the regional real-time device data; obtaining the scheduling weight of each scheduling target according to the probability prediction value; and updating the multi-objective optimization function according to the scheduling weight, and updating the multi-objective-oriented power grid scheduling decision. The method has the beneficial effects that the scene switching flexibility and the power dispatching accuracy are considered in the face of an actual complex scene.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power grid dispatching, and particularly relates to a multi-target-oriented power grid dispatching decision method and system. BACKGROUND

[0002] With the large-scale access of distributed energy, energy storage systems and flexible loads, the traditional power grid operation mode is developing towards high proportion of new energy, multi-source integration and regional autonomy. The current power grid dispatching system is gradually evolving from static control to dynamic optimization, and more emphasis is placed on the intelligence and multi-target orientation of dispatching decisions, such as considering economy, safety, green low carbon and operation stability at the same time.

[0003] In the related art, a single dispatching target is usually used for the output of dispatching decisions. In normal operation scenarios, the dispatching decisions mainly focus on load balancing as the core target and only pay attention to the static matching of power supply and demand. When abnormal weather or device failure occurs, the remaining dispatching decisions need to be called for adjustment, which is low in efficiency and easy to cause the shock of key operation parameters such as power grid frequency and voltage, thereby affecting the power supply quality.

[0004] In the related art, static statistical analysis is used for load fluctuation, abnormal weather or device abnormality, which cannot adapt to the dynamically changing power grid, resulting in weak risk prediction ability, poor accuracy of prediction results relied on by power grid dispatching, and low reliability of power grid dispatching. SUMMARY

[0005] The present application is aimed at the problem that the power grid dispatching in the prior art is limited to a single target scenario, and the scene switching flexibility and power dispatching accuracy are difficult to adapt to actual complex scenarios. A multi-target-oriented power grid dispatching decision method and system are provided. The method is constructed by a multi-target function facing normal dispatching state, abnormal weather dispatching state and device abnormal dispatching state, and the optimization priority is adjusted according to the risk probability in different scenarios, so as to uniformly output a decision that can meet the current high-risk scenario dispatching and can be used as a benchmark for dispatching adjustment after the risk changes in the remaining scenarios, thereby improving the adaptability of dispatching decisions, reducing the dispatching complexity of complex scene switching, and improving the dispatching efficiency.

[0006] To achieve the above technical purposes, a technical solution provided by the present application is a power grid scheduling decision method for multiple targets, comprising the following steps: constructing a normal scheduling target with minimum load fluctuation, constructing an abnormal weather scheduling target with minimum scheduling quantity and minimum scheduling loss, constructing a device abnormal scheduling target with minimum switching loss, and combining the normal scheduling target, the abnormal weather scheduling target and the device abnormal scheduling target with initial priority weights to construct a multi-target optimization function; obtaining a current load fluctuation probability prediction value according to the fluctuation matching of regional historical load data and regional real-time load data, obtaining a current weather abnormality probability prediction value according to the matching of regional historical weather data and regional real-time weather data, and obtaining a current device abnormality probability prediction value according to the matching of regional historical device maintenance data and regional real-time device data; obtaining scheduling weights of the normal scheduling target, the abnormal weather scheduling target and the device abnormal scheduling target respectively according to the current load fluctuation probability prediction value, the current weather abnormality probability prediction value and the current device abnormality probability prediction value; correcting the initial priority weights with the scheduling weights, updating the multi-target optimization function, obtaining a power grid scheduling decision for multiple targets with the updated multi-target optimization function and regional real-time data, and performing power grid scheduling with the power grid scheduling decision for multiple targets.

[0007] Further, the abnormal weather scheduling target is constructed with minimum scheduling quantity and minimum scheduling loss, comprising: calculating the scheduling quantity by multiplying the negative value of the scheduling frequency of the distributed energy resource scheduling times; calculating the scheduling loss by the scheduling loss of the distributed energy resource and the reserved distributed energy resource disaster resistance loss; and constructing the constraint condition of the abnormal weather scheduling target by the line safety margin threshold and the renewable energy power abandonment threshold.

[0008] Further, the abnormal weather scheduling target is constructed with minimum scheduling quantity and minimum scheduling loss, further comprising: setting the scheduling quantity weight and the scheduling loss weight according to the difference between the average value and the median value of the actual regional distributed energy resource scheduling frequency; and constructing the abnormal weather scheduling target with the scheduling quantity weight, the scheduling loss weight, the scheduling quantity and the scheduling loss.

[0009] Further, the device abnormal scheduling target is constructed with minimum switching loss, comprising: obtaining a first switching loss according to the line difference loss of the abnormal distributed energy resource device and the distributed energy resource device to be switched; obtaining a second switching loss according to the transient loss of the distributed energy resource device switching; and calculating the switching loss with the first switching loss and the second switching loss.

[0010] Further, the current load fluctuation probability prediction value is obtained according to matching of the regional historical load data and the regional real-time load data, comprising: obtaining the precursor load data of a preset precursor time period, calculating the similarity between the regional precursor load data and the regional historical load data by a similarity algorithm, and constructing a prediction trend set with historical load data whose similarity is greater than a preset similarity threshold; obtaining a deviation ratio by comparing the regional real-time load data with the prediction trend set; and obtaining the current load fluctuation probability prediction value according to the deviation ratio and the future load fluctuation in the prediction trend.

[0011] Further, the current weather abnormality probability prediction value is obtained according to matching of the regional historical weather data and the regional real-time weather data, comprising: obtaining the regional real-time weather data, screening matching historical weather data in the regional historical weather data that matches the regional real-time weather data at the same time sequence, and obtaining the current weather abnormality probability prediction value according to a proportion of abnormal weather existing in a preset period in all matching historical weather data.

[0012] Further, the current device abnormality probability prediction value is obtained according to matching of the regional historical device maintenance data and the regional real-time device data, comprising: constructing a device maintenance model of each device type according to the association between the device parameters and the maintenance time sequence in the regional historical device maintenance data; obtaining the remaining life by the regional real-time device data and the corresponding device maintenance model, and obtaining the current device abnormality probability prediction value according to the proportion of the remaining life and the initial life.

[0013] Further, the initial priority weight is corrected by the dispatching weight, the multi-objective optimization function is updated, and the multi-objective power grid dispatching decision is obtained according to the updated multi-objective optimization function and the regional real-time data, comprising: obtaining a special dispatching association relationship according to the human dispatching data and the weather abnormality type under the regional historical abnormal weather; obtaining the current weather abnormality type according to the matching of the regional historical weather data and the regional real-time weather data; obtaining a special dispatching correction weight according to the current weather abnormality type and the special dispatching association relationship; correcting the dispatching weight by the special dispatching correction weight, updating the multi-objective optimization function by the corrected dispatching weight, and obtaining the multi-objective power grid dispatching decision according to the updated multi-objective optimization function and the regional real-time data.

[0014] Further, the normal dispatching target is constructed according to the minimum load fluctuation, comprising: calculating the load fluctuation according to the load fluctuation standard deviation of the regional real-time load data and the regional average load data; and constructing the constraint condition of the normal dispatching target according to the regional equipment power balance and the equipment capacity.

[0015] Another technical solution provided in the application is a multi-target-oriented power grid scheduling decision system for implementing the method as described above, comprising: a storage layer for receiving and storing regional load data, weather data, maintenance data and device data; a development layer for storing construction logic of normal scheduling targets, abnormal weather scheduling targets, device abnormal scheduling targets and scheduling weights; and an application layer for calling data in the storage layer and construction logic in the development layer to perform power grid scheduling decision calculation.

[0016] The application has the following beneficial effects: 1. The scheduling targets in the normal state, the scheduling targets in the abnormal weather and the scheduling targets in the device abnormality are constructed in advance, the scheduling weights of the targets are adjusted according to the current regional load fluctuation probability, the weather abnormality probability and the device abnormality probability, the priority of each target in different scenarios is adjusted, the power grid scheduling decision can not only meet the maximum demand of the current scenario, but also can ensure the adjustment efficiency as much as possible when the remaining demand occurs, and the global adaptability of the power grid scheduling is improved.

[0017] 2. The distributed energy resources of high-frequency scheduling are reversely given higher calling priority, the characteristics that the response mechanism and stability of the distributed energy resources of high-frequency scheduling are usually higher are utilized, the reliability and stability of scheduling in abnormal weather are realized, and the failure caused by the excessive fragility of the distributed energy resources of low-frequency scheduling is avoided.

[0018] 3. The prediction deviation caused by accidental deviation of a single trend is avoided by multi-trend fusion calculation, and the prediction reliability is further enhanced by credibility, in the extreme case that the deviation of regional real-time load data from each prediction trend is large, the overall credibility is dispersed, so that the prediction trend with large fluctuation is more prominent, and the misjudgment of fluctuation in the extreme case can be avoided.

[0019] 4. The reliability of the current remaining life can be reflected by the calculation of the ratio of the remaining life to the initial life, the reliability is lower when the device is closer to the maintenance timing, and the failure problem is more likely to occur, so that the failure probability is judged according to the ratio of the remaining life to the initial life. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 The flowchart of the multi-target-oriented power grid scheduling decision method of the application.

[0021] Figure 2 The flowchart of the load fluctuation probability prediction in the multi-target-oriented power grid scheduling decision method of the application. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical scheme and advantages of the present application more clear, the present application is further described in detail below in combination with the drawings and examples. It should be understood that the specific embodiments described herein are only the best mode of the present application, which are used to explain the present application and do not limit the protection scope of the present application. All other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the present application.

[0023] As shown in Figure 1 , as an embodiment of the present application, a multi-target oriented power grid scheduling decision method comprises the following steps: constructing a normal scheduling target with minimum load fluctuation, constructing an abnormal weather scheduling target with minimum scheduling quantity and minimum scheduling loss, constructing a device abnormal scheduling target with minimum switching loss, and combining the normal scheduling target, the abnormal weather scheduling target and the device abnormal scheduling target with initial priority weights to construct a multi-target optimization function; obtaining a current load fluctuation probability prediction value according to the matching of regional historical load data and regional real-time load data, obtaining a current weather abnormality probability prediction value according to the matching of regional historical weather data and regional real-time weather data, and obtaining a current device abnormality probability prediction value according to the matching of regional historical device maintenance data and regional real-time device data; obtaining scheduling weights of the normal scheduling target, the abnormal weather scheduling target and the device abnormal scheduling target respectively according to the current load fluctuation probability prediction value, the current weather abnormality probability prediction value and the current device abnormality probability prediction value; correcting the initial priority weights with the scheduling weights, updating the multi-target optimization function, obtaining a multi-target oriented power grid scheduling decision with the updated multi-target optimization function and regional real-time data, and executing power grid scheduling with the multi-target oriented power grid scheduling decision.

[0024] In the present embodiment, the multi-target optimization function is constructed in advance according to the initial priority weights, and is simultaneously directed to the scheduling scenarios under normal state, under abnormal weather and under device abnormality, and is guided to the scheduling scenario with high priority according to the initial priority weights, but does not affect the power grid scheduling decision to simultaneously adapt to the remaining scheduling scenarios with low priority. Then, the scheduling weights of each target are adjusted through the current regional load fluctuation probability, weather abnormality probability and device abnormality probability, the initial priority weights are updated with the scheduling weights, so that the power grid scheduling decision is more adapted to the current high-risk probability scenario, but simultaneously has the adaptability to switch the remaining scenarios, ensuring the accuracy of power scheduling and the flexibility of complex scenario switching.

[0025] Specifically, constructing a normal scheduling target with minimum load fluctuation comprises: The load fluctuation is calculated based on the regional real-time load data and the load fluctuation standard deviation of the regional average load data. The normal scheduling target is constructed based on the regional device power balance and the device capacity.

[0026] Under normal circumstances, the power grid scheduling needs to prioritize the peak and valley difference of the load, and reduce the frequent start and stop of units or the repeated charging and discharging of energy storage caused by sudden changes in load. Therefore, the regional real-time load data can be obtained by accessing the distribution network SCADA system and user electricity collection terminal through the data access adaptation module to obtain regional real-time load data and regional historical load data, and the regional average load data is obtained based on the regional historical load data matched with the regional real-time load data.

[0027] The normal scheduling target is: ; The normal scheduling target is: The normal scheduling target function is represented by The regional real-time load data at time t is represented by The regional average load data is represented by

[0028] Since the power grid scheduling usually needs to meet the load balance in a scheduling period, the normal scheduling target is constructed based on the regional real-time load data in a scheduling period. For example, the normal scheduling target is: ; The normal scheduling target is: The normal scheduling target function of the period is represented by The regional average load data in the same period as the period in the regional historical load data is represented by

[0029] At the same time, the regional power balance constraint is constructed based on the same input and output power in the region, and the regional device capacity constraint is constructed based on the upper and lower limits of the generator output and the upper and lower limits of the energy storage system.

[0030] The abnormal weather scheduling target is constructed based on the minimum scheduling quantity and the minimum scheduling loss, including: The scheduling quantity is calculated based on the negative product of the distributed energy resource scheduling frequency and the scheduling frequency; The scheduling loss is calculated based on the distributed energy resource scheduling loss and the reserved distributed energy resource disaster loss; The constraint condition of the abnormal weather scheduling target is constructed based on the line safety margin threshold and the renewable energy curtailment threshold.

[0031] In abnormal weather conditions, the output of renewable energy may drop sharply due to weather interference. Therefore, it is necessary to reduce unnecessary dispatch operations, but at the same time, it is necessary to avoid damage to grid equipment under abnormal weather conditions. For example, in typhoon weather, it is necessary to control the wind turbine to adjust the blade angle to avoid damage to the blade, but at the same time, it is necessary to ensure the power supply as much as possible.

[0032] Specifically, the objectives for managing abnormal weather conditions are: ; in, This represents the objective function for scheduling during abnormal weather. Indicates the first Number of times distributed energy resource scheduling is performed. This represents the total number of distributed energy resources. Indicates the first The truth value of distributed energy resource scheduling Indicates the first Frequency of distributed energy resource dispatch Indicates the scheduling quantity weight. Indicates the scheduling loss weight. This indicates the loss in the dispatching of distributed energy resources. This indicates that distributed energy resources should be retained to mitigate disaster losses. .

[0033] When the When the state of a distributed energy resource changes, it is considered that the distributed energy resource has been dispatched. When the first If the state of a distributed energy resource remains unchanged, it is considered that the distributed energy resource has not been dispatched. Therefore, when distributed energy resources are dispatched, the number of dispatches and the frequency of dispatch of the distributed energy resources are calculated simultaneously, as well as the dispatch losses of the dispatched distributed energy resources and the disaster resistance losses of the undispatched distributed energy resources.

[0034] It is understandable that Distributed Energy Resources (DER) include distributed renewable energy, energy storage systems, and controllable loads.

[0035] In abnormal weather, each scheduling change can cause damage to distributed energy resources or trigger equipment failure that is already at the limit. Therefore, the number of equipment scheduling in abnormal weather is reduced as much as possible, and the scheduling quantity is calculated in combination with the negative value of scheduling frequency. The more frequently the distributed energy resources are scheduled, the lower the scheduling quantity when they are scheduled, so that the distributed energy resources that are scheduled more frequently are more likely to be scheduled. In this embodiment, the distributed energy resources that are scheduled frequently are given higher priority for calling, and the characteristics of higher response mechanism and stability of the distributed energy resources that are scheduled more frequently are utilized to realize the reliability and stability of scheduling in abnormal weather, and to avoid failure caused by the vulnerability of the distributed energy resources that are scheduled less frequently. At the same time, the more the scheduling times, the lower the scheduling quantity, and the single scheduling quantity of the distributed energy resources is reflected by the scheduling times, so as to avoid power grid mutation caused by too high single scheduling power and ensure the stability of the power grid.

[0036] In this embodiment, the distributed energy resource loss in abnormal weather is divided into distributed energy resource scheduling loss and reserved distributed energy resource disaster resistance loss. The distributed energy resource scheduling loss is obtained according to the distributed energy resource state change loss of the scheduled distributed energy resources and the disaster loss of the distributed energy resources after scheduling. The disaster resistance loss of the reserved distributed energy resources is obtained according to the disaster loss of the unscheduled distributed energy resources. The disaster loss model of each type of distributed energy resource can be built according to the disaster situation of each state of the historical distributed energy resources in abnormal weather. When scheduling loss is performed, the corresponding disaster loss data in the disaster loss model is called.

[0037] By dividing the distributed energy resource loss in abnormal weather into distributed energy resource scheduling loss and reserved distributed energy resource disaster resistance loss, the inaction of the distributed energy resources is avoided when calculating the disaster loss, and the accuracy of the loss calculation is improved.

[0038] The load reserved by the line is taken as the load safety margin threshold to avoid overloading of the line in disaster. The acceptable renewable energy curtailment value preset by the renewable energy is taken as the renewable energy curtailment threshold to avoid excessive waste of resources in avoiding abnormal weather.

[0039] In this embodiment, the scheduling quantity weight and scheduling loss weight are set based on the difference between the average scheduling frequency and the median scheduling frequency of distributed energy resources in the actual area. The larger the difference, the larger the scheduling quantity weight; the smaller the difference, the larger the scheduling loss weight. The difference between the average scheduling frequency and the median scheduling frequency reflects the uneven distribution of scheduling frequency. When the difference is large, it indicates that a few distributed energy resources are scheduled more frequently, while most are scheduled less frequently. In this case, the scheduling quantity weight is larger, prioritizing the scheduling of distributed energy resources with high frequency of scheduling to avoid resource overload or poor scheduling effect caused by scheduling too many distributed energy resources that are not frequently adjusted. When the difference is small, it indicates that the distributed energy resources in the current area have high adjustability. In this case, the scheduling loss weight is increased to reduce the loss generated during the scheduling process. It can be understood that in some cases, the ratio of the difference to the average scheduling frequency can be directly used as the scheduling quantity weight.

[0040] The objectives for constructing device anomaly scheduling with minimum switching loss include: The first switching loss is obtained based on the line difference loss between the abnormal distributed energy resource device and the distributed energy resource device to be switched. The second switching loss is obtained based on the transient loss of the switching of distributed energy resource devices; The switching loss is calculated using the first switching loss and the second switching loss.

[0041] In the event of an anomaly in a distributed energy resource device, it is necessary to ensure normal power supply through device switching. Therefore, it is necessary to minimize the additional losses during the switching process.

[0042] Specifically, the device anomaly scheduling objective is: ; in, This represents the objective function for abnormal device scheduling. This represents the total number of distributed energy resource devices. Indicates the first The line length of each distributed energy resource device Indicates corresponding to the first The line length of the distributed energy resource device to be switched is [length missing]. This represents the line loss per unit length. Indicates corresponding to the first The transient switching loss of the distributed energy resource device to be switched.

[0043] It can be understood that the line loss difference is often small in a region, so the line length change before and after switching is used to obtain the line switching loss. The transient switching loss of the distributed energy resource device to be switched can be simulated and obtained in advance according to the transient simulation software and pre-stored, and called when calculating the switching loss.

[0044] In the embodiment, constructing the device abnormal scheduling target with the minimum switching loss further includes: The preset switching time and the capacity of the distributed energy resource device to be switched are used to construct the constraint condition of the device abnormal scheduling target.

[0045] According to the safety standard, the preset switching time is set to 3 minutes, and in actual application, the preset switching time can also be set according to actual needs.

[0046] The initial priority weight is used to combine the normal scheduling target, the abnormal weather scheduling target and the device abnormal scheduling target to construct the multi-objective optimization function, including: The initial priority weight is constructed according to the ratio of the number of abnormal weather days, the number of device abnormal days and the number of normal days in history.

[0047] In the embodiment, taking a year as a calculation period, the initial priority weight of each scheduling target in the year is obtained according to the ratio of the number of abnormal weather days, the number of device abnormal days and the number of normal days in the previous year to the total number of days in the year, and the calculation and output of the current initial power grid scheduling decision are carried out. It can be understood that in other cases, the initial priority weight can also be set according to expert experience.

[0048] As shown in Figure 2 The current load fluctuation probability prediction value is obtained according to the fluctuation matching of the regional historical load data and the regional real-time load data, including: The pre-sequence load data of the preset pre-sequence time period of the region is obtained, the similarity between the regional pre-sequence load data and the regional historical load data is calculated by a similarity algorithm, and the historical load data with a similarity greater than a preset similarity threshold is used to construct a prediction trend set; The deviation ratio is obtained by comparing the regional real-time load data with the prediction trend set; The current load fluctuation probability prediction value is obtained according to the deviation ratio and the future load fluctuation in the prediction trend.

[0049] In the embodiment, the preset preceding time period is 24 hours, i.e. the load data of the preceding 24 hours of the current time sequence is taken as the preceding load data. The dynamic time warping algorithm or the cosine similarity algorithm can be used to calculate the curve similarity of the preceding load data and the historical load data, and the historical load data curve with a curve similarity greater than a preset similarity threshold is taken as the prediction trend. The regional load is acquired in real time, and the deviation ratio of the regional real-time load and the predicted value in each prediction trend is used to calculate the load fluctuation probability prediction value.

[0050] Specifically, the deviation ratio is acquired by comparing the regional real-time load data with the prediction trend set, including: acquiring the predicted load data corresponding to the regional real-time load data time sequence of each prediction trend in the prediction trend set; acquiring the deviation ratio by taking the difference between each predicted load data and the regional real-time load data and the ratio of the regional real-time load data.

[0051] the current load fluctuation probability prediction value is acquired by taking the deviation ratio and the future load fluctuation of the prediction trend, including: acquiring the credibility of each prediction trend by taking the deviation ratio of each prediction trend, and calculating the future load fluctuation standard deviation of each prediction trend in a preset period; calculating the overall load fluctuation standard deviation according to the credibility of all prediction trends and the future load fluctuation standard deviation, and acquiring the current load fluctuation probability prediction value by taking the ratio of the overall load fluctuation standard deviation and the regional historical maximum load fluctuation standard deviation.

[0052] In the embodiment, the load at the current time point is predicted according to the load curve with the same trend as the preceding time period, and the relative deviation of the predicted value and the regional real-time load data is taken to acquire the deviation ratio. The smaller the deviation ratio is, the more consistent the prediction trend and the actual trend are, and the higher the credibility is. The credibility of each prediction trend can be acquired by using the exponential decay function and the normalization algorithm. Further, the load fluctuation standard deviation of each prediction trend in a preset period is calculated. In the embodiment, the preset period is 24 hours. The credibility is taken as the weight to perform weighted summation on the future load fluctuation standard deviation of all prediction trends, to acquire the overall load fluctuation standard deviation. The ratio of the overall load fluctuation standard deviation and the regional historical maximum load fluctuation standard deviation is taken as the current load fluctuation probability prediction value. The prediction deviation caused by accidental deviation of a single trend is avoided by the fusion calculation of multiple trends, and the prediction reliability is further enhanced by the credibility. In the extreme case, when the deviation of the regional real-time load data and each prediction trend is large, the overall credibility is dispersed, so that the prediction trend with large fluctuation is more highlighted, and the misjudgment of the fluctuation in the extreme case is avoided.

[0053] The regional historical maximum load fluctuation standard deviation reflects the fluctuation state in the most extreme case. The current overall load fluctuation standard deviation and the regional historical maximum load fluctuation standard deviation reflect the current risk position. If the current overall load fluctuation standard deviation is much lower than the regional historical maximum load fluctuation standard deviation, even if there is fluctuation, it is more consistent with the stable load low fluctuation of an ordinary working day. At this time, the fluctuation does not need to be adjusted too much, and therefore, it can be considered as a low fluctuation probability.

[0054] The current weather abnormality probability prediction value is obtained according to the matching of the regional historical weather data and the regional real-time weather data, and includes: The regional real-time weather data is obtained, the matching historical weather data that matches the regional real-time weather data in the same time sequence is screened from the regional historical weather data, and the current weather abnormality probability prediction value is obtained according to the proportion of abnormal weather existing in a preset period in all matching historical weather data.

[0055] In this embodiment, the weather data at least includes temperature and humidity. The historical weather data that matches the regional real-time weather data in the time sequence and the weather data is taken as the matching historical weather data, and the proportion of abnormal weather existing in a preset period in the matching historical weather data is used to determine the probability of future weather abnormality.

[0056] The current device abnormality probability prediction value is obtained according to the matching of the regional historical device maintenance data and the regional real-time device data, and includes: A device maintenance model of each device type is constructed according to the association relationship between the device parameters and the maintenance time sequence in the regional historical device maintenance data; The remaining life is obtained according to the regional real-time device data and the corresponding device maintenance model, and the current device abnormality probability prediction value is obtained according to the proportion of the remaining life and the initial life.

[0057] In this embodiment, the historical device maintenance data at least includes device type, device maintenance time sequence and device maintenance period parameter, the regional real-time device data at least includes real-time device parameter, the device maintenance model corresponding to each device type is constructed according to the device maintenance time sequence and the device maintenance period parameter, the device remaining life under the regional real-time device data is obtained according to the device maintenance model, and the current device abnormality probability prediction value is obtained according to the proportion of the remaining life and the initial life of the device in the unused state. Through the calculation of the proportion of the remaining life and the initial life, the reliability of the current remaining life can be reflected. The closer the device is to the maintenance time sequence, the lower the reliability is, and the more likely the device is to fail. Therefore, the failure probability is determined according to the proportion of the remaining life and the initial life.

[0058] The scheduling weights of the normal scheduling target, the abnormal weather scheduling target and the device abnormal scheduling target are obtained according to the current load fluctuation probability prediction value, the current weather abnormality probability prediction value and the current device abnormality probability prediction value, and the scheduling weights of the normal scheduling target, the abnormal weather scheduling target and the device abnormal scheduling target include: The scheduling weights of the normal scheduling target, the abnormal weather scheduling target and the device abnormal scheduling target are obtained by normalizing the current load fluctuation probability prediction value, the current weather abnormality probability prediction value and the current device abnormality probability prediction value, and the sum of the scheduling weights of the normal scheduling target, the abnormal weather scheduling target and the device abnormal scheduling target is 1.

[0059] In the embodiment, the scheduling priority in the multi-objective optimization is determined by the probability of occurrence of each risk, the scheduling demand of the high-risk scenario is ensured to be met, and the adjustment margin that may occur in the remaining risk scenarios is reserved, so that the remaining risk scenarios can be quickly adjusted when they occur.

[0060] Specifically, the initial priority weight is corrected by the scheduling weight, the multi-objective optimization function is updated, and the multi-objective power grid scheduling decision is obtained according to the updated multi-objective optimization function and the regional real-time data, and the multi-objective power grid scheduling decision includes: The initial priority weight is replaced by the scheduling weight, the multi-objective optimization function is updated, and the multi-objective power grid scheduling decision corresponding to the minimum weighted sum is obtained by iterative calculation according to the updated multi-objective optimization function and the regional real-time data.

[0061] In the process of obtaining the power grid scheduling decision according to the initial priority weight, the regional real-time load data, the real-time weather data and the real-time device data are obtained, the probabilities of different scenarios that may occur in the future are calculated, the scheduling weight is obtained according to the probabilities, and the initial priority weight is corrected, so that the optimal scheduling scheme in the multi-objective case is selected by the weighted sum of the minimum value of each objective function value and the corresponding scheduling weight, without complex single-scene calculation and comprehensive solution, the scheduling efficiency is improved, and at the same time, the optimization priority of each objective is determined by the scheduling weight, so that the power grid scheduling decision is more suitable for the current risk higher scenario, while the remaining scenarios are not abandoned. In this way, the current power grid scheduling decision not only adapts to the current high-risk scenario but also adapts to the remaining scenarios, and when the risk value of the remaining scenarios increases, the adjustment decision can be made based on the current power grid scheduling decision, the scheduling when the risk changes is reduced, and the stability of the power grid is improved.

[0062] In this case, the regional real-time data at least includes regional real-time load data, regional real-time weather data and regional real-time device data. The initial priority weight is corrected by the scheduling weight, the multi-objective optimization function is updated, and the multi-objective power grid scheduling decision is obtained according to the updated multi-objective optimization function and the regional real-time data, and the multi-objective power grid scheduling decision further includes: The current power grid dispatching decision is taken as an iteration basis, the dispatching weight corresponding to the current power grid dispatching decision is taken as an initial priority weight combination, the latest dispatching weight is calculated in real time, the initial priority weight is corrected by using the latest dispatching weight, the multi-objective optimization function is updated, and the power grid dispatching decision method is updated on the basis of the updated multi-objective optimization function.

[0063] At this time, each time the dispatching weight is updated on the basis of the original power grid dispatching decision. Since the original power grid dispatching decision itself meets the three risk scenarios, even if the risk probability of one of the risk scenarios increases, the device used to compensate for the risk in the original power grid dispatching decision can remain unchanged, and the dispatching of the remaining devices is performed, thereby reducing the number of times of dispatching the devices in the risk scenario switching process, improving the dispatching efficiency, and compensating for the efficiency reduction problem caused by the multiple dispatching of high-frequency dispatching devices in abnormal weather scenarios, while taking into account reliability and efficiency.

[0064] In some other cases, the initial priority weight is corrected by using the dispatching weight, the multi-objective optimization function is updated, and the multi-objective power grid dispatching decision based on the updated multi-objective optimization function and the regional real-time data further includes: obtaining a special dispatching correlation relationship according to the human dispatching data and the weather abnormal type under the regional historical abnormal weather; obtaining the current weather abnormal type according to the matching of the regional historical weather data and the regional real-time weather data; obtaining a special dispatching correction weight according to the current weather abnormal type according to the special dispatching correlation relationship; correcting the dispatching weight by using the special dispatching correction weight, updating the multi-objective optimization function by using the corrected dispatching weight, and obtaining the multi-objective power grid dispatching decision based on the updated multi-objective optimization function and the regional real-time data.

[0065] In this case, considering that in the case of special abnormal weather, the harm caused by power failure may be greater than the damage to the device, for example, in extremely cold weather, power supply needs to be ensured, and therefore the dispatching weight correction under special weather is performed according to the human dispatching data under the regional historical abnormal weather, the dispatching weight of the abnormal weather dispatching target and the device abnormal dispatching target is reduced, and the dispatching weight of the normal dispatching target is increased, to ensure that the dispatching decision meets the demand in the actual situation.

[0066] Specifically, obtaining a special dispatching correlation relationship according to the human dispatching data and the weather abnormal type under the regional historical abnormal weather includes: obtaining the dispatching type, the corresponding weather abnormal type, the weather range, the distance between the dispatching device and the extreme weather center in the human dispatching data, and constructing the special dispatching correlation relationship according to the distance range between the dispatching device and the extreme weather center of each dispatching type corresponding to each weather abnormal type under different weather ranges.

[0067] In the embodiment, the distance range of the scheduling device corresponding to the same scheduling type and the extreme weather center is taken as the association reference, the number of scheduling devices in the distance range is obtained, the special scheduling correction ratio is obtained from the number of unscheduled devices, the modified scheduling weight is obtained from the product of the special scheduling correction ratio and the scheduling weight. If there is a weather abnormality probability in the current area, the current weather abnormality type and the current extreme weather center are obtained according to the matching of the area historical weather data and the area real-time weather data. The extreme weather center can be obtained according to the most serious disaster area center in abnormal weather, for example, in extremely cold weather, the center with the lowest temperature is taken as the extreme weather center. The distance range in the special scheduling association relationship is used to obtain the region to be modified, and the scheduling weight of all devices in the region to be modified is corrected according to the special scheduling correction ratio corresponding to the scheduling type, so as to reduce the number of devices in the extreme weather range to avoid disaster, ensure the basic power supply demand of the current area, avoid power imbalance caused by excessive disaster avoidance, and cause more serious disaster loss.

[0068] As the second embodiment of the present application, a multi-target-oriented power grid scheduling decision system comprises: A storage layer is configured to receive and store area load data, weather data, maintenance data and device data. A development layer is configured to store normal scheduling target, abnormal weather scheduling target, device abnormal scheduling target and scheduling weight construction logic. An application layer is configured to call the data in the storage layer and the construction logic in the development layer to perform power grid scheduling decision calculation.

[0069] The normal scheduling target, the abnormal weather scheduling target, the device abnormal scheduling target and the scheduling weight construction logic at least include: The normal scheduling target is constructed based on the minimum load fluctuation, the abnormal weather scheduling target is constructed based on the minimum scheduling quantity and the minimum scheduling loss, the device abnormal scheduling target is constructed based on the minimum switching loss, and the multi-target optimization function is constructed by combining the normal scheduling target, the abnormal weather scheduling target and the device abnormal scheduling target based on the initial priority weight. The current load fluctuation probability prediction value is obtained according to the fluctuation matching of the area historical load data and the area real-time load data, the current weather abnormality probability prediction value is obtained according to the matching of the area historical weather data and the area real-time weather data, and the current device abnormality probability prediction value is obtained according to the matching of the area historical device maintenance data and the area real-time device data. The scheduling weight of the normal scheduling target, the abnormal weather scheduling target and the device abnormal scheduling target is obtained based on the current load fluctuation probability prediction value, the current weather abnormality probability prediction value and the current device abnormality probability prediction value.

[0070] Specifically, the storage layer receives real-time data through Kafka / RabbitMQ, Kafka and RabbitMQ build a message queue system to realize asynchronous communication, task triggering and event-driven mechanism between modules, and synchronize historical data with the power grid system through an interface. In the storage layer, the historical load, historical weather, and historical device maintenance data are stored in a relational database (RDS / Dameng), and the real-time load, real-time weather, and real-time device data are stored in Redis, which also temporarily stores the probability prediction value, scheduling weight, and decision result. The metadata management service is used to uniformly arrange data to ensure that the application layer can quickly read the required data.

[0071] In the development layer, Python is used to write calculation models of minimum load fluctuation, minimum scheduling quantity, minimum scheduling loss, and minimum switching loss, as well as mapping algorithms of probability prediction value and scheduling weight, and Java is used to encapsulate the algorithms into standardized interfaces for the application layer to call.

[0072] In the application layer, the initial priority weight, regional historical load data, regional real-time load data, regional historical weather data, regional real-time weather data, regional historical device maintenance data, and regional real-time device data of the storage layer are called, the probability calculation interface of the development layer is called, the current load fluctuation probability prediction value, the current weather anomaly probability prediction value, and the current device anomaly probability prediction value are calculated and stored in the storage layer. The probability prediction value of the storage layer is called, the weight calculation interface of the development layer is called, the scheduling weight of the normal scheduling target, the abnormal weather scheduling target, and the device abnormal scheduling target is calculated and stored in the storage layer. The weight result of the storage layer is called, the multi-objective decision interface of the development layer is called, the power grid scheduling decision is calculated by fusing the three scheduling targets and stored in the storage layer and output to the execution end.

[0073] In some other cases, a multi-objective oriented power grid scheduling decision system further includes: The deployment layer encapsulates the build logic in the development layer in a Docker container to provide a containerized running environment for the application layer.

[0074] In the deployment layer, a distributed containerized deployment method is adopted, the front-end service is responsible for static resource hosting and access routing reverse proxy by Nginx, the back-end service deploys each micro-service component based on Tomcat, supports service-level load balancing and hot deployment, encapsulates different build logics in the development layer into independent Docker containers respectively, ensures the running environment isolation of each logic module, and avoids dependency conflicts.

[0075] In the embodiment, the storage layer, the development layer, the application layer and the deployment layer are built based on the Java EE architecture. The development layer is built based on the Spring Boot and the Spring Cloud to build the service components, supports mixed development of languages such as Java, Python and C++, and uses the Vue and the HTML to build the front-end interface. The deployment layer includes the Nginx front-end service, the Tomcat back-end service and the algorithm container environment based on the Docker, and supports online calling of multiple models and isolated running of heterogeneous environments.

[0076] The above described specific embodiment is a preferred embodiment of the power grid dispatching decision method and system of the application, and is not intended to limit the specific implementation range of the application. The range of the application includes but is not limited to the specific embodiment. Any equivalent changes made in accordance with the shape and structure of the application are within the protection scope of the application.

Claims

1. A multi-objective power grid dispatching decision-making method, characterized in that: Includes the following steps: The normal scheduling objective is constructed with the minimum load fluctuation, the abnormal weather scheduling objective is constructed with the minimum scheduling amount and the minimum scheduling loss, the abnormal device scheduling objective is constructed with the minimum switching loss, and the multi-objective optimization function is constructed by combining the normal scheduling objective, the abnormal weather scheduling objective and the abnormal device scheduling objective with the initial priority weight. The current load fluctuation probability prediction value is obtained by matching the fluctuation of regional historical load data with regional real-time load data; the current weather anomaly probability prediction value is obtained by matching regional historical weather data with regional real-time weather data; and the current device anomaly probability prediction value is obtained by matching regional historical device maintenance data with regional real-time device data. The scheduling weights for normal scheduling targets, abnormal weather scheduling targets, and equipment abnormality scheduling targets are obtained based on the current load fluctuation probability prediction value, the current weather abnormality probability prediction value, and the current equipment abnormality probability prediction value, respectively. The initial priority weights are corrected by adjusting the scheduling weights, the multi-objective optimization function is updated, and the updated multi-objective optimization function and real-time regional data are used to obtain multi-objective power grid scheduling decisions. Power grid scheduling is then executed based on the multi-objective power grid scheduling decisions.

2. The power grid dispatching decision-making method oriented towards multiple objectives as described in claim 1, characterized in that: The method of constructing an abnormal weather scheduling objective with minimum scheduling amount and minimum scheduling loss includes: The scheduling quantity is calculated by multiplying the number of times distributed energy resources are scheduled by the negative value of the scheduling frequency. Dispatch losses are calculated based on the dispatch losses of distributed energy resources and the disaster resistance losses of retained distributed energy resources. Constraints for abnormal weather dispatch targets are constructed using line safety margin thresholds and renewable energy curtailment thresholds.

3. The power grid dispatching decision-making method oriented towards multiple objectives as described in claim 2, characterized in that: The method of constructing an abnormal weather scheduling objective with minimum scheduling amount and minimum scheduling loss also includes: The scheduling quantity weight and scheduling loss weight are set based on the difference between the average scheduling frequency and the median scheduling frequency of distributed energy resources in the actual area. An abnormal weather scheduling target is constructed using scheduling quantity weight, scheduling loss weight, scheduling quantity, and scheduling loss.

4. The power grid dispatching decision-making method oriented towards multiple objectives as described in claim 1, characterized in that: The objective of constructing abnormal device scheduling with minimum switching loss includes: The first switching loss is obtained based on the line difference loss between the abnormal distributed energy resource device and the distributed energy resource device to be switched. The second switching loss is obtained based on the transient loss of the switching of distributed energy resource devices; The switching loss is calculated using the first switching loss and the second switching loss.

5. The power grid dispatching decision-making method oriented towards multiple objectives as described in claim 1, characterized in that: The step of obtaining the current load fluctuation probability prediction value based on the fluctuation matching of regional historical load data and regional real-time load data includes: Obtain the preceding load data of the region within a preset preceding time period, calculate the similarity between the preceding load data and the historical load data of the region using a similarity algorithm, and construct a prediction trend set using historical load data with similarity greater than a preset similarity threshold; The deviation ratio is obtained by comparing the real-time regional load data with the predicted trend set. The predicted probability of current load fluctuation is obtained by using the deviation ratio and the future load fluctuation situation in the forecast trend.

6. The power grid dispatching decision-making method oriented towards multiple objectives as described in claim 1, characterized in that: The step of obtaining the current weather anomaly probability prediction value by matching regional historical weather data with regional real-time weather data includes: Obtain real-time weather data for the region, filter historical weather data that match the real-time weather data for the same time period in the region's historical weather data, and obtain the current weather anomaly probability prediction value based on the proportion of abnormal weather within a preset period in all matched historical weather data.

7. The power grid dispatching decision-making method oriented towards multiple objectives as described in claim 1, characterized in that: The step of obtaining the current device anomaly probability prediction value by matching regional historical device maintenance data with regional real-time device data includes: Based on the correlation between device parameters and maintenance sequence in the historical device maintenance data of the region, device maintenance models for each device type are constructed; The remaining lifespan is obtained by using real-time regional device data and the corresponding device maintenance model, and the current device anomaly probability prediction value is obtained by using the ratio of the remaining lifespan to the initial lifespan.

8. The power grid dispatching decision-making method oriented towards multiple objectives as described in claim 1, characterized in that: The process of correcting the initial priority weights with scheduling weights, updating the multi-objective optimization function, and obtaining multi-objective power grid scheduling decisions based on the updated multi-objective optimization function and real-time regional data includes: Special scheduling relationships are obtained based on historical human scheduling data under abnormal weather conditions in the region and the types of weather anomalies. The current weather anomaly type is obtained by matching regional historical weather data with regional real-time weather data; Based on the current weather anomaly type, special scheduling correction weights are obtained according to special scheduling correlations; The scheduling weights are corrected using special scheduling correction weights, and the multi-objective optimization function is updated with the corrected scheduling weights. The updated multi-objective optimization function and regional real-time data are used to obtain multi-objective power grid scheduling decisions.

9. The power grid dispatching decision-making method oriented towards multiple objectives as described in claim 1, characterized in that: The goal of establishing normal scheduling with minimal load fluctuations includes: Load fluctuation is calculated using the standard deviation of load fluctuations from regional real-time load data and regional average load data. Constraints for normal scheduling objectives are constructed based on regional equipment power balance and equipment capacity.

10. A multi-objective power grid dispatching decision system, used to implement the method as described in any one of claims 1 to 9, characterized in that: include: The storage layer is used to receive and store regional load data, weather data, maintenance data, and equipment data; The development layer is used to store the construction logic for normal scheduling targets, abnormal weather scheduling targets, device abnormality scheduling targets, and scheduling weights; The application layer is used to call data in the storage layer and the construction logic in the development layer to perform power grid dispatching decision calculations.

Images