Quantitative evaluation method, device and equipment for power imbalance of new energy power system

Abstract

The application provides a new energy power system power imbalance quantitative evaluation method, device and equipment, by considering the random volatility of uncertain power output and load demand and the mutual influence therebetween, a joint probability prediction model for different prediction time scales of a future to-be-evaluated period is constructed, and by using the joint probability prediction model and the working state information of the uncertain power and load in the time dimension provided by the operation scheduling plan, a probability distribution of the net difference of the uncertain power in the to-be-evaluated period is generated, the net difference of the deterministic power and load is considered, and the power imbalance distribution of the whole power system is accurately predicted, so as to provide accurate data support for the scheduling and operation of the power system.

Description

Technical Field

[0001] This application relates to the field of power system analysis technology, and in particular to a method, apparatus and equipment for quantitatively assessing power imbalance in a new energy power system. Background Technology

[0002] The output of new energy power sources is inherently subject to random fluctuations and is easily affected by various factors such as weather and seasons, resulting in strong uncertainty in output. Furthermore, with the diversification of power-side equipment types and the integration of large-scale power electronic interface equipment, the load changes of the power system exhibit more complex random characteristics.

[0003] In scenarios where there are significant errors in the forecasting of renewable energy output and load, such as extreme freezing weather, the output of renewable energy sources and the demand for loads may exhibit inverse changes. For example, when the output of renewable energy sources decreases due to weather conditions, the demand for loads may increase due to heating or other reasons, resulting in a large power shortage in the power system during a specific period. If the reserve capacity of the power system is insufficient, this may pose a significant risk to safe operation and could even lead to the collapse of the power system.

[0004] Therefore, there is an urgent need for a new method for predicting and quantifying power imbalance in power systems, which can accurately predict and assess the power imbalance state of power systems, thereby providing data support for power system dispatching and operation. Summary of the Invention

[0005] In view of this, in order to solve the above-mentioned technical problems, this application provides a method, apparatus, equipment, storage medium and computer program product for quantitative assessment of power imbalance in new energy power systems.

[0006] Specifically, this application is implemented through the following technical solution:

[0007] According to a first aspect of the embodiments of this application, a method for quantitatively assessing the power imbalance in a new energy power system is provided, the method comprising:

[0008] Using a first probability prediction model for the predicted object, the power probability distribution of the predicted object at each prediction time point within the period to be evaluated is obtained; the predicted object includes uncertain power sources and loads; the model is constructed based on the error data between the historical predicted values ​​and the historical actual values ​​of the predicted object's power.

[0009] A joint probability prediction model is constructed based on the power probability distribution of each predicted object at each prediction time point and the correlation coefficient between the power probability distributions.

[0010] Obtain the operation scheduling plan for the period to be evaluated and determine the power superposition matrix; the power superposition matrix is ​​used to characterize the running time of each predicted object in each time period within the period to be evaluated; the time period refers to the time period between each predicted time point within the period to be evaluated;

[0011] Based on the joint probability prediction model and the power superposition matrix, the probability distribution of the net difference in uncertain power between uncertain power sources and loads during the period to be evaluated is determined.

[0012] Based on the operation scheduling plan, the net difference between the deterministic power and the load during the period to be evaluated is determined, and a predicted distribution of power imbalance during the period to be evaluated is generated based on the probability distribution of the net difference between the deterministic power and the net difference between the uncertain power.

[0013] Optionally, the method further includes a first probabilistic prediction model construction step, comprising:

[0014] Obtain the historical true value of the predicted object within the historical period;

[0015] Obtain the historical point prediction value of the historical true value at the historical time point. The historical point prediction value includes the point prediction value of the prediction object at the historical time point obtained by making forward predictions with different forward time points.

[0016] Modeling is performed on the error data distribution between the predicted historical point value and the actual historical value at each look-ahead time, resulting in multiple probabilistic prediction models at different look-ahead time scales, which serve as the first probabilistic prediction model for the predicted object; the first probabilistic prediction model has linear superposition property.

[0017] Optionally, when the first probabilistic prediction model includes multiple probabilistic prediction models with different look-ahead time scales, obtaining the power probability distribution of the predicted object at each prediction time point within the period to be evaluated includes:

[0018] The time interval from the current moment of power system assessment to the predicted time point is obtained, and a target probability prediction model whose forward time scale conforms to the time interval is determined from the first probability prediction model of the predicted object.

[0019] Obtain the actual power of the predicted object at the current time;

[0020] Based on the actual power and the target probability prediction model, the power probability distribution of the predicted object at the prediction time point is obtained.

[0021] Optionally, the method further includes a step of determining the correlation coefficient between the power probability distributions, comprising:

[0022] Using a set duration as the look-ahead time, a historical prediction sequence is obtained, consisting of historical point prediction values ​​of the power of the predicted object under the look-ahead time. The set duration refers to the time interval from the current moment of power system assessment to the prediction time point.

[0023] Based on the true value sequence formed by the historical true values ​​of the predicted object power and the historical prediction sequence, a historical prediction error sequence at the prediction time point is generated.

[0024] Determine the correlation coefficients between each pair of historical prediction error sequences for each prediction object at each prediction time point, and use these coefficients as the correlation coefficients between the corresponding power probability distributions.

[0025] Optionally, determining the probability distribution of the net difference in uncertain electricity quantities between the uncertain power source and the load during the period to be evaluated includes:

[0026] Based on the joint probability prediction model, a first joint probability distribution of each predicted object is obtained within the period to be evaluated; the first joint probability distribution includes multiple different independent variables; each independent variable is used to represent the power value of a predicted object at one prediction time point within the period to be evaluated.

[0027] Based on the power superposition matrix, determine the first linear transformation relationship between the net difference in uncertain power consumption during the period to be evaluated and the independent variable of the first joint probability distribution;

[0028] Based on the first joint probability distribution of each predicted object during the period to be evaluated, and combined with the first linear transformation relationship, the probability distribution of the uncertain net difference in electricity consumption is derived.

[0029] Optionally, if the first joint probability distribution satisfies a Gaussian mixture distribution, the step of deriving the probability distribution of the uncertain net electricity difference based on the first joint probability distribution of each predicted object during the period to be evaluated, combined with the first linear transformation relationship, includes:

[0030] Obtain the first statistical parameters of the first joint probability distribution, wherein the first statistical parameters include at least the mean and the covariance matrix;

[0031] Based on the first linear transformation relationship, the parameter transformation relationship between the first statistical parameter and the second statistical parameter of the probability distribution of the uncertain net difference in electricity consumption is obtained;

[0032] Based on the parameter transformation relationship, the first statistical parameter is transformed, and the transformed statistical parameter is determined as the second statistical parameter.

[0033] Based on the second statistical parameter, a Gaussian mixture distribution of the uncertain net difference in electricity consumption is determined as the probability distribution of the uncertain net difference in electricity consumption.

[0034] Optionally, determining the probability distribution of the net difference in uncertain electricity quantities between the uncertain power source and the load during the period to be evaluated includes:

[0035] Using the joint probability prediction model, the second joint probability distribution of each prediction object at each prediction time point within the period to be evaluated is obtained;

[0036] For each time period between adjacent prediction time points, a second linear transformation relationship is determined between the net difference in uncertain electricity consumption within the time period and the independent variables of the second joint probability distribution, based on the electricity consumption superposition matrix.

[0037] Based on the second linear transformation relationship and the second joint probability distribution at the predicted time points included in the time period, the probability distribution of the net difference in uncertain electricity consumption for the time period is obtained.

[0038] By combining the probability distributions of the net difference in uncertain electricity consumption in each time period within the period to be evaluated, the probability distribution of the net difference in uncertain electricity consumption for the entire period to be evaluated is obtained.

[0039] Optionally, generating the predicted distribution of the power imbalance within the time period to be evaluated includes:

[0040] For the period to be evaluated, a third linear transformation relationship is determined between the power imbalance and the uncertain net power difference, as well as the certain net power difference.

[0041] Based on the probability distribution of the third linear transformation relationship and the net difference in uncertain electricity, a predicted distribution of the electricity imbalance is generated.

[0042] Optionally, generating the predicted distribution of the power imbalance within the time period to be evaluated includes:

[0043] Samples are drawn from the probability distribution of the uncertain net power imbalance, and the power imbalance value corresponding to the sample is determined according to the deterministic net power imbalance and the reserve capacity, thus obtaining a sample set of power imbalance values; the predicted distribution of the power imbalance is determined according to the sample set of power imbalance values.

[0044] According to a second aspect of the embodiments of this application, a quantitative assessment device for power imbalance in a new energy power system is provided, the device comprising:

[0045] The power probability distribution prediction module is used to obtain the power probability distribution of the predicted object at each prediction time point within the period to be evaluated using a first probability prediction model of the predicted object; the predicted object includes uncertain power sources and loads; the model is constructed based on the error data between the historical predicted values ​​and historical true values ​​of the power of the predicted object.

[0046] The joint probability prediction construction module is used to construct a joint probability prediction model based on the power probability distribution of each prediction object at each prediction time point and the correlation coefficient between the power probability distributions.

[0047] The superposition relationship acquisition module is used to acquire the operation scheduling plan within the period to be evaluated and determine the power superposition matrix; the power superposition matrix is ​​used to characterize the running time of each prediction object in each time period within the period to be evaluated; the time period refers to the time period between each prediction time point within the period to be evaluated;

[0048] The uncertain power difference distribution prediction module is used to determine the probability distribution of the uncertain net power difference between uncertain power sources and loads during the period to be evaluated, based on the joint probability prediction model and the power superposition matrix.

[0049] The power imbalance distribution prediction module is used to determine the net difference between deterministic power and load power in the period to be evaluated based on the operation scheduling plan, and generate a predicted distribution of power imbalance in the period to be evaluated based on the probability distribution of the net difference between deterministic power and the net difference between uncertain power.

[0050] Optionally, the device further includes:

[0051] Obtain the historical true value of the predicted object within the historical period;

[0052] Obtain the historical point prediction value of the historical true value at the historical time point. The historical point prediction value includes the point prediction value of the prediction object at the historical time point obtained by making forward predictions with different forward time points.

[0053] Modeling is performed on the error data distribution between the predicted historical point value and the actual historical value at each look-ahead time, resulting in multiple probabilistic prediction models at different look-ahead time scales, which serve as the first probabilistic prediction model for the predicted object; the first probabilistic prediction model has linear superposition property.

[0054] Optionally, when the first probabilistic prediction model includes multiple probabilistic prediction models with different look-ahead time scales, the power probability distribution prediction module is specifically used for:

[0055] The time interval from the current moment of power system assessment to the predicted time point is obtained, and a target probability prediction model whose forward time scale conforms to the time interval is determined from the first probability prediction model of the predicted object.

[0056] Obtain the actual power of the predicted object at the current time;

[0057] Based on the actual power and the target probability prediction model, the power probability distribution of the predicted object at the prediction time point is obtained.

[0058] Optionally, the device further includes:

[0059] Using a set duration as the look-ahead time, a historical prediction sequence is obtained, consisting of historical point prediction values ​​of the power of the predicted object under the look-ahead time. The set duration refers to the time interval from the current moment of power system assessment to the prediction time point.

[0060] Based on the true value sequence formed by the historical true values ​​of the predicted object power and the historical prediction sequence, a historical prediction error sequence at the prediction time point is generated.

[0061] Determine the correlation coefficients between each pair of historical prediction error sequences for each prediction object at each prediction time point, and use these coefficients as the correlation coefficients between the corresponding power probability distributions.

[0062] Optionally, the uncertainty power difference distribution prediction module is specifically used for:

[0063] Based on the joint probability prediction model, a first joint probability distribution of each predicted object is obtained within the period to be evaluated; the first joint probability distribution includes multiple different independent variables; each independent variable is used to represent the power value of a predicted object at one prediction time point within the period to be evaluated.

[0064] Based on the power superposition matrix, determine the first linear transformation relationship between the net difference in uncertain power consumption during the period to be evaluated and the independent variable of the first joint probability distribution;

[0065] Based on the first joint probability distribution of each predicted object during the period to be evaluated, and combined with the first linear transformation relationship, the probability distribution of the uncertain net difference in electricity consumption is derived.

[0066] Optionally, when the first joint probability distribution satisfies a Gaussian mixture distribution, the uncertainty electricity difference distribution prediction module, in deriving the probability distribution of the net uncertainty electricity difference based on the first joint probability distribution of each prediction object within the time period to be evaluated, combined with the first linear transformation relationship, includes:

[0067] Obtain the first statistical parameters of the first joint probability distribution, wherein the first statistical parameters include at least the mean and the covariance matrix;

[0068] Based on the first linear transformation relationship, the parameter transformation relationship between the first statistical parameter and the second statistical parameter of the probability distribution of the uncertain net difference in electricity consumption is obtained;

[0069] Based on the parameter transformation relationship, the first statistical parameter is transformed, and the transformed statistical parameter is determined as the second statistical parameter.

[0070] Based on the second statistical parameter, a Gaussian mixture distribution of the uncertain net difference in electricity consumption is determined as the probability distribution of the uncertain net difference in electricity consumption.

[0071] Optionally, the uncertainty power difference distribution prediction module is specifically used for:

[0072] Using the joint probability prediction model, the second joint probability distribution of each prediction object at each prediction time point within the period to be evaluated is obtained;

[0073] For each time period between adjacent prediction time points, a second linear transformation relationship is determined between the net difference in uncertain electricity consumption within the time period and the independent variables of the second joint probability distribution, based on the electricity consumption superposition matrix.

[0074] Based on the second linear transformation relationship and the second joint probability distribution at the predicted time points included in the time period, the probability distribution of the net difference in uncertain electricity consumption for the time period is obtained.

[0075] By combining the probability distributions of the net difference in uncertain electricity consumption in each time period within the period to be evaluated, the probability distribution of the net difference in uncertain electricity consumption for the entire period to be evaluated is obtained.

[0076] Optionally, the power imbalance distribution prediction module is specifically used for:

[0077] For the period to be evaluated, a third linear transformation relationship is determined between the power imbalance and the uncertain net power difference, as well as the certain net power difference.

[0078] Based on the probability distribution of the third linear transformation relationship and the net difference in uncertain electricity, a predicted distribution of the electricity imbalance is generated.

[0079] Optionally, the power imbalance distribution prediction module is specifically used for:

[0080] Samples are drawn from the probability distribution of the uncertain net power imbalance, and the power imbalance value corresponding to the sample is determined according to the deterministic net power imbalance and the reserve capacity, thus obtaining a sample set of power imbalance values; the predicted distribution of the power imbalance is determined according to the sample set of power imbalance values.

[0081] According to a third aspect of the embodiments of this application, an electronic device is provided, the electronic device comprising: a memory and a processor; the memory being used to store a computer program; the processor being used to execute the above-described quantitative assessment method for power imbalance in a new energy power system by invoking the computer program.

[0082] According to a fourth aspect of the embodiments of this application, a computer-readable storage medium is provided, on which a computer program is stored, wherein when the program is executed by a processor, it implements the above-described quantitative assessment method for power imbalance in a new energy power system.

[0083] The technical solutions provided in this application embodiment may include the following beneficial effects:

[0084] In the technical solution provided in this application, a joint probability prediction model is constructed to achieve power prediction by combining the predicted power probability distribution of uncertain power sources or loads in the power system at different prediction time points in the future assessment period, and the mutual influence between the prediction results of uncertain power sources or loads at different prediction time points. This improves the accuracy of predicting uncertain power sources or loads. By using the joint probability prediction model and the working status information of uncertain power sources and loads in the time dimension provided by the operation and scheduling plan, the probability distribution of the net difference of uncertain electricity in the assessment period is generated. Considering the net difference between deterministic electricity and load, the power imbalance state of the entire power system is accurately predicted, thereby providing accurate data support for the scheduling and operation of the power system. Attached Figure Description

[0085] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0086] Figure 1 This is a schematic diagram of a quantitative assessment method for power imbalance in a new energy power system, as illustrated in an exemplary embodiment of this application.

[0087] Figure 2 This is a schematic diagram illustrating a process of pre-constructing a first probability prediction model for a prediction object, as shown in an exemplary embodiment of this application.

[0088] Figure 3A This is a flowchart illustrating an exemplary embodiment of the present application of obtaining the power probability distribution at each prediction time point under multiple probabilistic prediction models with different look-ahead time scales;

[0089] Figure 3B This is a schematic diagram illustrating the relationship between the current time and the predicted time point within the period to be evaluated, as shown in an exemplary embodiment of this application.

[0090] Figure 4 This is a flowchart illustrating an exemplary embodiment of the present application of obtaining the correlation coefficient between the power distributions of each predicted object at each prediction time point;

[0091] Figure 5 This is a flowchart illustrating the steps for deriving the probability distribution of uncertain net electricity difference, as shown in an exemplary embodiment of this application.

[0092] Figure 6A This is a schematic diagram illustrating another calculation process for deriving the probability distribution of the net difference in uncertain electricity consumption, as shown in an exemplary embodiment of this application.

[0093] Figure 6B This is a flowchart illustrating another implementation step for deriving the probability distribution of the uncertain net difference in electricity consumption, as shown in an exemplary embodiment of this application.

[0094] Figure 7 This is a flowchart illustrating an exemplary embodiment of the present application of a quantitative assessment method for power imbalance in a new energy power system under a scenario of predicting power imbalance in the next 3-5 days from the current point in time;

[0095] Figure 8 This is a schematic diagram of the structure of a quantitative assessment device for power imbalance in a new energy power system, as shown in an exemplary embodiment of this application.

[0096] Figure 9 This is a hardware schematic diagram of an electronic device illustrated in an exemplary embodiment of this application. Detailed Implementation

[0097] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0098] This application provides a quantitative assessment method for power imbalance in a new energy power system. It is applicable to power system assessment scenarios where power output or load changes are uncertain. This method can comprehensively consider the uncertainty of new energy power output and the uncertainty of load changes, and accurately predict and quantitatively assess the power imbalance state of the power system, thereby providing effective support for the safe and stable operation of the power system.

[0099] This method can be applied to a terminal or a server, or to a system including both a terminal and a server, and is implemented through the interaction between the terminal and the server. The terminal can include, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, etc., and portable wearable devices can include smartwatches, smart bracelets, head-mounted devices, etc. The server can include a standalone server or a server cluster consisting of multiple servers.

[0100] like Figure 1 As shown, the method in this embodiment may include at least the following steps:

[0101] S101, using the first probability prediction model of the prediction object, obtain the power probability distribution of the prediction object at each prediction time point in the period to be evaluated; the prediction object includes uncertain power sources and loads; the model is constructed based on the error data between the historical predicted values ​​and the historical true values ​​of the power of the prediction object.

[0102] In power systems, uncertain power sources refer to sources whose output is random and easily affected by various factors such as weather conditions. These uncertain power sources can include at least renewable energy sources such as wind power and solar power. For example, due to variations in wind speed and solar radiation intensity, the output of wind turbines and solar photovoltaic panels can fluctuate, resulting in uncertainty in power output.

[0103] Uncertain loads are loads whose power output is random. Changes in this type of load may be affected by a variety of factors, such as changes in equipment usage patterns, changes in user electricity consumption behavior, weather conditions, and economic activities, making it difficult for the power system to accurately forecast demand and plan power supply for this type of load.

[0104] The specific type of the aforementioned prediction object is determined based on the distribution of uncertain power sources and loads in the actual power system. The distribution of uncertain power sources and loads in the power system can include: the presence of only uncertain power sources, or only uncertain loads, or both uncertain power sources and uncertain loads. Therefore, when obtaining the aforementioned power probability distribution based on the first probability prediction model of the prediction object:

[0105] For a power system with only uncertain power sources, a probabilistic prediction model constructed for the uncertain power sources can be used separately to predict the power output probability distribution at each prediction time point from the current time.

[0106] For power systems with only uncertain loads, the probability prediction model built for the uncertain loads can also be used separately to obtain the power demand probability distribution at each prediction time point.

[0107] For a power system that includes both uncertain power sources and uncertain loads, the probabilistic prediction models for the uncertain power sources and uncertain loads can be used in parallel to obtain the power probability distribution at each prediction time point.

[0108] The time intervals between the aforementioned prediction time points can be fixed time intervals. That is, the period to be evaluated is divided into multiple sampling points according to a fixed sampling period, and the entire period to be evaluated is divided into multiple time periods. The sampling period can be determined based on the accuracy of the power imbalance quantification demand and the dynamic characteristics of the system. For example, if the period to be evaluated is the 8th to 15th day in the future, the prediction time points are set to the time points at 24-hour intervals every 8 to 15 days in the future.

[0109] The first probability prediction model for the aforementioned prediction object is constructed based on the error data between the historical predicted value and the actual value of the power of the prediction object. The modeling method includes, but is not limited to, parameter estimation and non-parametric estimation methods. The established first probability prediction model has linear superposition. Furthermore, in order to improve the prediction accuracy, a first probability prediction model can be constructed separately for each prediction object. For example, if the power system includes two wind power generation units, the first probability prediction model can be constructed separately for each wind power generation unit.

[0110] The first probabilistic prediction model can be a hybrid time-scale probabilistic prediction model established for each prediction object included in the power system. When constructing the probabilistic prediction model based on historical data, the uncertainty of the prediction object can be captured and simulated through methods such as time series analysis, machine learning, or deep learning, so that the same model can output the probability distribution of the prediction object at each prediction time point. Alternatively, the first probabilistic prediction model can also be multiple probabilistic prediction models with different forward time scales established for the prediction object, each model corresponding to a different prediction time range or prediction duration. When obtaining the power probability distribution of the prediction object at each prediction time point within the aforementioned period to be evaluated, the most suitable forward time scale prediction model is selected from the first probabilistic prediction models for prediction based on the prediction duration from the current time to the prediction time point.

[0111] The aforementioned power probability distribution refers to the probability density function, which describes all possible power values ​​and their corresponding probabilities that the predicted object may generate at a predicted time point within the period to be evaluated. For example, for a wind farm, its power output is affected by various factors such as weather conditions. By establishing a probabilistic prediction model, the probability of the wind farm generating different power values ​​at a specified predicted time point in the future can be obtained.

[0112] Taking the sampling point with the prediction time point set as the sampling period of 8-15 days or 24 hours in the future of the current time as an example, assuming that the power system includes the uncertain power sources P1 and P2 and the uncertain load L, the power probability distribution of P1 at each prediction time point in the next 8-15 days can be obtained by using the first probability prediction model of P1 that was established in advance. Similarly, the power probability distribution of P2 and L at each prediction time point can be obtained by using the first probability prediction models of P2 and L respectively.

[0113] S102, construct a joint probability prediction model based on the power probability distribution of each predicted object at each prediction time point and the correlation coefficient between the power probability distributions;

[0114] The correlation coefficient is used to represent the strength and direction of the linear relationship between two or more variables. In this embodiment, the correlation coefficient is used to describe the degree of association and mutual influence of the power probability distributions of different predicted objects at different prediction times. The correlation coefficient can be determined based on the historical data sequences corresponding to each predicted object at different prediction time points, and can be stored in the form of a correlation coefficient matrix. For example, two wind farms located in close geographical locations may be affected by similar weather conditions, and therefore their power probability distributions may have a high correlation.

[0115] For example, a power system includes two forecast objects, such as an uncertain power source P1 and an uncertain load L. The power probability distribution for each forecast object corresponds to two future forecast time points. For example, the probability distribution of P1 includes f...11 with f 12 The probability distribution of P2 includes f 21 f 22 The correlation coefficient between the power probability distributions is referenced below.

[0116] Table 1 illustrates the following (the values ​​highlighted in dark in the table represent the correlation coefficients between probability distributions; the table is symmetrical):

[0117]

[0118] Table 1 illustrates the correlation coefficients between power probability distributions.

[0119] The aforementioned joint probability prediction model is used to provide more comprehensive and accurate joint prediction results based on the power probability distribution of the predicted objects and considering the mutual influence and correlation between power probability distributions. Its output joint probability distribution describes the comprehensive probability distribution prediction after the interaction of multiple predicted objects, providing the joint probability of different combinations of values ​​of multiple predicted objects occurring simultaneously. Since multiple prediction time points are set within the evaluation period, the power probability distribution of the same predicted object at these multiple prediction time points belongs to the prediction results under different prediction time scales, enabling the constructed joint probability prediction model to predict the joint probability distribution of each predicted object at different time scales.

[0120] For example, if wind power, photovoltaic power, and uncertain load 1 are taken as the prediction objects, the joint probability prediction model can output the joint probability distribution of the three prediction objects at each prediction time point within the period to be evaluated. This distribution describes all possible combinations of values ​​of the three prediction objects at that prediction time point and their corresponding probabilities, such as the probability of wind power being X1, photovoltaic power being X2, and load being X3.

[0121] Given the power probability distribution and correlation coefficient of each predicted object at each prediction time point, the joint probability prediction model can be modeled by parametric estimation, nonparametric estimation, Copula model, Gaussian mixture model, etc., and the joint probability model also has linear superposition.

[0122] Taking parameter estimation methods as an example, the joint probability prediction model can be constructed in the following way:

[0123] For the power probability distribution of each predicted object at each predicted time point, determine the distribution parameters of the power probability distribution; for example, if the power probability distribution is a normal distribution, then obtain its mean and standard deviation.

[0124] Based on the correlation coefficients between the power probability distributions and the distribution parameters of each power probability distribution, estimate the covariance between the power probability distributions; for example, the distribution parameter is the standard deviation, assuming r i,jσ is the correlation coefficient between the i-th and j-th power probability distributions. i and σ j If the standard deviations of the power probability distributions are given, then the elements ∑ of the covariance matrix Σ are... i,j It can be done through ∑ i,j =r i,j *σ i *σ j The calculation shows that the elements on the diagonal ∑ i,i The variance of each power probability distribution;

[0125] A joint probability prediction model is constructed based on the power probability distribution of each predicted object at each prediction time point and the covariance between the power probability distributions. This joint probability prediction model can use a multivariate probability distribution (such as a multivariate normal distribution) to describe the joint probability distribution of each predicted object.

[0126] After the joint probability prediction model is built, its performance can be verified using historical data or independent test datasets, such as calculating the model's prediction accuracy, error metrics, and other relevant statistical indicators.

[0127] S103, obtain the operation scheduling plan for the period to be evaluated, and determine the power superposition matrix; the power superposition matrix is ​​used to characterize the running time of each predicted object in each time period within the period to be evaluated; the time period refers to the time period between each predicted time point within the period to be evaluated;

[0128] A power system operation and dispatch plan refers to a pre-formulated operation plan for the period to be evaluated, based on certain deterministic and uncertain optimization strategies and the actual situation and forecast data of the power system. This plan includes the output of each power source, reserve capacity, and load power within the power system, aiming to achieve optimal allocation of power resources and stable operation of the power system. The setting period of the operation and dispatch plan can be less than or equal to the sampling period of each forecast time point within the period to be evaluated, in order to obtain a more accurate power superposition matrix.

[0129] The aforementioned time periods refer to the time periods between two adjacent prediction times within the period to be evaluated. The period to be evaluated can be divided into several time periods based on the frequency of prediction time points within the period to be evaluated. For example, if the prediction is made hourly, then each time period is one hour; if the prediction is made on a daily basis, then each time period is one day.

[0130] After obtaining the operation scheduling plan, it can be parsed to determine the operating status of each predicted object, which may include information such as power-on, power-off, and output. Then, a power superposition matrix is ​​constructed based on the parsed information. The number of rows in this matrix can be equal to the number of predicted objects in the power system, and the number of columns can be equal to the number of time periods mentioned above. All elements of the initialized matrix can be 0 or indicate "not running". When filling the matrix with the operating status of each predicted object in each time period according to the parsed scheduling plan, the operating status can be represented by a number (e.g., 1 for running, 0 for not running) or a specific operating duration (e.g., hours). This operating status can indicate the power superposition relationship between predicted objects when subsequently calculating the net difference in uncertain power.

[0131] S104, Based on the joint probability prediction model and the power superposition matrix, determine the probability distribution of the net difference in uncertain power between uncertain power sources and loads during the period to be evaluated;

[0132] The net difference in uncertain power output between uncertain power sources and loads refers to the difference between the total power output of all uncertain power sources and the total power consumption of all uncertain loads during the period to be evaluated. It reflects the situation of power surplus or shortage in the power system caused by the mismatch between the output of uncertain power sources and the demand of uncertain loads during the period to be evaluated.

[0133] The probability distribution of the uncertain net difference in electricity consumption describes the possible values ​​and probabilities of the uncertain electricity consumption difference in the power system during the entire evaluation period. The joint probability prediction model outputs the probability of each predicted object occurring simultaneously under different power value combinations. The electricity consumption superposition matrix represents the operating time of each predicted object in each time period within the evaluation period. Based on the linear transformation relationship that electricity consumption is the product of power and operating time, the probability distribution of the uncertain net difference in electricity consumption during the entire evaluation period can be obtained from the joint probability prediction model and the electricity consumption superposition matrix.

[0134] To determine the probability distribution of the uncertain net difference in electricity consumption, methods such as sampling, approximation, linear superposition, and transformation can be used. A joint probability prediction model can be employed to obtain a first joint probability distribution for the power values ​​of each predicted object at each prediction time point over the entire evaluation period. This first joint probability distribution can then be combined with the electricity consumption superposition matrix to deduce the probability distribution of the aforementioned uncertain net difference in electricity consumption. Specifically, the linear superposition and transformation method is based on the linear superposition property of the joint probability prediction model. By establishing a linear transformation relationship between the uncertain net difference in electricity consumption over the entire evaluation period and the independent variables of the first joint probability distribution, as well as the electricity consumption superposition matrix, the probability distribution of the uncertain net difference in electricity consumption can be derived from the first joint probability distribution concerning power.

[0135] Taking the sampling method as an example, the probability distribution of the uncertain net difference in electricity consumption can be determined in the following way: a large number of power value combination samples of each predicted object at each prediction time point are generated according to the joint probability prediction model; for each power value combination sample, the uncertain net difference in electricity consumption of each power value combination sample is calculated according to the working time of each predicted object in each time period provided by the electricity consumption superposition matrix; the uncertain net difference in electricity consumption of all samples is statistically analyzed to obtain the probability distribution of the uncertain net difference in electricity consumption.

[0136] S105, according to the operation scheduling plan, determine the net difference between the deterministic power and the load during the period to be evaluated, and generate a predicted distribution of power imbalance during the period to be evaluated based on the probability distribution of the net difference between the deterministic power and the net difference between the uncertain power.

[0137] The deterministic power supply and load refer to the power supply and demand that can be predicted relatively accurately based on plans, historical data and known factors. The predicted values ​​have a small fluctuation range. The deterministic power source can include controllable generating equipment or energy sources, such as traditional fossil fuel power plants and nuclear power plants. The deterministic load can include user demand from fixed power consumption plans and power consumption patterns, such as industrial power consumption, commercial power consumption and some residential power consumption.

[0138] Based on the definition of deterministic power and load, the net difference in deterministic power refers to the expected power surplus or shortage in the power system during the period to be evaluated without considering uncertainties. The net difference in deterministic power can be calculated as the difference between the total planned power generation of all deterministic power sources and the total planned power consumption of all deterministic loads during the period to be evaluated, based on the planned values ​​of deterministic power and load in the operation and dispatch plan.

[0139] The predicted distribution of the power imbalance is a probabilistic description of the power supply-demand imbalance during the evaluation period. It can be determined through linear superposition transformation, approximation methods, etc. The specific method of determining the predicted distribution depends on the specific form and properties of the probability distribution of the uncertain net power difference. In this embodiment, the power imbalance in the power system refers to the power supply-demand imbalance. A deterministic net power difference can be linearly superimposed on the probability distribution of the uncertain net power difference to obtain the predicted distribution of the power imbalance during the evaluation period. When the operation and scheduling plan includes reserve capacity during the evaluation period, the linear superposition operation also includes superimposing the reserve capacity. This reserve capacity refers to the power generation capacity reserved by the power system to cope with uncertainties, used to supplement power when the system power is insufficient.

[0140] The predicted distribution of this power imbalance comprehensively considers the probability distribution of both deterministic and uncertain net power imbalances, thus providing a series of possible power imbalance values ​​and their probabilities for the power system during the period to be assessed. This can provide important reference information for power system decision-makers, such as helping power system operators assess the risks under different power imbalance levels, identify potential risk points, and formulate corresponding risk response measures. It can also provide a basis for power system operation scheduling and planning, so as to reasonably adjust the operation scheduling plan.

[0141] This disclosure provides a quantitative assessment scheme for predicting power system imbalances that considers uncertainties. By constructing a first probabilistic prediction model for uncertain power sources and uncertain loads, the power probability distribution of each uncertain predictor during the future assessment period is obtained. Then, based on the power probability distribution and correlation coefficients of each predictor, a joint probabilistic prediction model is constructed to comprehensively consider the mutual influence between different predictors, thereby effectively improving the accuracy and reliability of power system imbalance prediction. Subsequently, a power superposition matrix is ​​introduced according to the operation and scheduling plan to provide the operating status information of each predictor in the power system for power imbalance prediction. Combining the joint probabilistic prediction model and the power superposition matrix, the probability distribution of the net difference in uncertain power volume for the entire assessment period can be obtained. Considering the net difference between deterministic power volume and load, a predicted distribution of power imbalance is generated. This comprehensively considers the random fluctuations of uncertain power source output and load demand, as well as their mutual influence, accurately predicting the power imbalance state of the entire power system. This provides precise data support for power system scheduling and operation, helps dispatchers formulate more scientific and reasonable scheduling plans, and reduces the safety risks of the power system.

[0142] In some embodiments, the duration of the current moment in the power system assessment varies depending on the distance between different prediction time points within the assessment period. To improve the prediction accuracy of the power probability distribution of the predicted object at that prediction time point, the power probability distribution at each prediction time point can be obtained through a probabilistic prediction model with a corresponding look-ahead time scale. Based on this, the first probabilistic prediction model of the predicted object in step S101 may include multiple probabilistic prediction models with different look-ahead time scales established for the predicted object. The method may also include a step of pre-constructing the first probabilistic prediction model, such as... Figure 2 As shown, taking a prediction object as an example, the steps may include:

[0143] S201, Obtain the historical true value of the predicted object power within the historical period;

[0144] The historical true value refers to the actual operating power data of the predicted object over a period of time. This historical true value can be collected by power system monitoring equipment or data recording system, reflecting the power output or load demand of the predicted object in its past actual operation.

[0145] S202, For each historical true value, obtain the historical point prediction value of the historical time point where the historical true value is located. The historical point prediction value includes the point prediction value of the predicted object at the historical time point obtained by making forward predictions with different forward time.

[0146] Forward forecasting refers to the act of predicting the situation at a future point in time or over a period of time from the current moment. Using different forward timeframes in forward forecasting indicates that different time spans or lead times are considered during the forecasting process. For example, a forward forecast with one forward timeframe predicts the situation for the next hour, while another forward forecast with a different forward timeframe predicts the situation for the next 24 hours or longer. Different forward timeframes allow the forecasting model to consider influencing factors and change patterns at different time scales, thereby improving the accuracy and applicability of the forecast.

[0147] Based on the above definition of forward prediction, the historical point prediction value at the historical time point where the historical true value is located refers to multiple prediction values ​​obtained by forward prediction of the power of the prediction object at that historical time point at different historical moments before that historical time point. For example, if the historical time point where the historical true value is located is 12:00 on April 30, then the predicted value at the historical point can include the power value of the predicted object at 12:00 on April 30, predicted from any historical time before 12:00 on April 30 (such as 12:00 on April 28, 12:00 on April 29, or 00:00 on April 30). Specifically, predicting 12:00 on April 30 from 12:00 on April 28 is a forward prediction with a look-ahead time of 2 days, predicting 12:00 on April 30 from 12:00 on April 29 is a forward prediction with a look-ahead time of 1 day, and predicting 12:00 on April 30 from 00:00 on April 30 is a forward prediction with a look-ahead time of 12 hours.

[0148] In one example, the look-ahead time mentioned in step S202 can be determined based on the time interval between the current moment of the power system assessment and each predicted time point within the aforementioned assessment period. For example, if the assessment period is the 3rd to 5th day in the future, and three predicted time points are set in days, then the time intervals between the current moment and each predicted time point within the aforementioned assessment period are 3 days, 4 days, and 5 days, respectively. Based on this, when obtaining the historical point prediction value corresponding to the historical true value, the historical point prediction value can include the prediction value made N days before the historical time point where the historical true value is located, where N takes values ​​from 3 to 5, i.e., the aforementioned look-ahead times are 3 days, 4 days, and 5 days, respectively.

[0149] S203, Model the error data distribution between the predicted historical point value and the actual historical value for each look-ahead time period to obtain multiple first probability prediction models for different look-ahead time scales, which serve as the first probability prediction models for the prediction object; the first probability prediction models have linear superposition.

[0150] After obtaining the historical true values ​​and corresponding predicted values ​​at each historical time point, data cleaning can be performed first, such as removing outliers and missing values, to ensure data quality and completeness. For the error data between the predicted and true historical values ​​at each look-ahead time scale, statistical analysis can be performed to understand the distribution characteristics of the error, such as mean, variance, skewness, and kurtosis. Based on the statistical analysis of the error data, a modeling method can be selected to construct a first probabilistic prediction model for that look-ahead time scale. After the model is constructed, the first probabilistic prediction model can be trained using the aforementioned historical true values, and some historical true value data can be retained for model validation.

[0151] Suppose the obtained historical true value sequence of a certain prediction object is [a1, a2, a3…a… n The corresponding historical point prediction sequence includes predictions made 3, 4, and 5 days prior to each historical true value's historical time point:

[0152] Based on the predicted historical values ​​from 3 days ago: [b 31 b 32 b 33 …b 3n ];

[0153] Based on the predicted historical values ​​from 4 days ago: [b 41 b 42 b 43 …b 4n ];

[0154] Based on the predicted historical values ​​from 5 days ago: [b 51 b52 b 53 …b 5n ];

[0155] Based on the aforementioned historical true values ​​and historical point predictions, and according to the data distribution of the prediction errors between the historical point predictions and historical true values ​​at the same historical time point under each look-ahead time, three first probability prediction models with look-ahead time scales of 3 days, 4 days, and 5 days can be constructed using methods such as parameter estimation and non-parametric estimation. These models are used as the first probability prediction models for the predicted object. The multiple first probability prediction models with different look-ahead time scales have linear superposition.

[0156] In this embodiment of the disclosure, since the time distance between different predicted time points within the future period to be evaluated is different from the current time, the accuracy of the prediction is affected by the time span. By pre-constructing a probability prediction model under different forward time scales for each prediction object, it is possible to more accurately capture the operating characteristics and uncertainties of the prediction object under different time scales, thereby improving the accuracy of the prediction.

[0157] In some embodiments, based on multiple probabilistic prediction models for the prediction object at different look-ahead time scales determined in the above embodiments, for the first probabilistic prediction model of the prediction object described in step S101 above, the power probability distribution of the prediction object at each prediction time point within the period to be evaluated can be obtained. Power probability distribution prediction can be performed by selecting a suitable probabilistic prediction model with a suitable look-ahead time scale for each prediction time point. Figure 3A As shown, the following steps may be included:

[0158] S301, obtain the time interval from the current moment of power system assessment to the predicted time point, and determine the target probability prediction model that the forward time scale conforms to the time interval from the first probability prediction model of the predicted object;

[0159] The current moment for conducting the power system assessment refers to the time point at which the power probability distribution of the predicted object is obtained at each prediction time point, that is, the time point at which the power system assessment is performed to obtain the power imbalance distribution of the power system; the time interval refers to the duration between the current moment and the prediction time point in the future to be assessed period.

[0160] The first probability prediction model is trained based on historical data. Its look-ahead timescale is closely related to the time range considered during model training. Probability prediction models with different look-ahead timescales learn the influence and correlation of various factors that may affect power output at different times, such as weather patterns and seasonal changes, and can provide accurate predictions at that look-ahead timescale. Therefore, for prediction objects including first probability prediction models with different look-ahead timescales, the appropriate look-ahead timescale probability prediction model can be selected based on the time interval from the current time to the prediction time point to predict the power distribution at each prediction time point, thereby improving the prediction accuracy.

[0161] For example, see Figure 3B The example shows the period to be evaluated and the predicted time point. The period to be evaluated refers to the future time t1 to t2 shown in the figure. n In the time period t0, where t0 is the current time and t2, t3, and t4 are example predicted time points, the aforementioned time interval refers to the duration from time t0 to time t2, t3, and t4, respectively, corresponding to the duration T in the figure. 02 T 03 T 04 Based on the first probability prediction models of the prediction object with different look-ahead time scales pre-constructed in the above embodiments, taking the prediction of the power probability distribution at time t2 as an example, a look-ahead time scale conforming to T is selected from the first probability prediction model of the prediction object. 02 The first probability prediction model is used as the target probability prediction model for the prediction time point t2.

[0162] Taking the future assessment period as the 3rd to 5th day in the future as an example, and setting 3 prediction time points in days, the time interval from the current moment to the first prediction time point is 3 days. Then, the first probability prediction model with a forward time scale of 3 days is selected from the first probability prediction model of the prediction object as the target probability prediction model for the first prediction time point, which is used to predict the power probability distribution of the prediction object at the first prediction time point.

[0163] S302, Obtain the actual power of the predicted object at the current time;

[0164] S303, Based on the actual power and the target probability prediction model, obtain the power probability distribution of the predicted object at the prediction time point.

[0165] In this embodiment, the actual power at the current moment serves as the initial condition for predicting the future state, reflecting the actual performance of the object in the current state. After obtaining the actual power at the current moment, and combining it with the selected target probability prediction model, the power probability distribution of the object at a specific prediction time point can be predicted. This power probability distribution describes the power values ​​that the object may obtain at the prediction time point and their corresponding probabilities.

[0166] In this embodiment of the disclosure, in order to meet the prediction requirements of the prediction object at different time scales, a probabilistic prediction model that matches the time interval between the current time and the prediction time point is first selected. This can maximize the prediction capability of the model and optimize the prediction accuracy. By combining the actual power data at the current time and the selected probabilistic prediction model, the power of the prediction object at the prediction time point is predicted probabilistically, thereby improving the reliability and accuracy of the prediction results.

[0167] In some embodiments, the correlation coefficient between the power distributions described in step S102 can be determined based on the historical data sequences corresponding to each predicted object at different prediction time points, see [link to relevant documentation]. Figure 4 As shown, based on the historical true values ​​and corresponding historical point predicted values ​​obtained from the aforementioned embodiments, the following steps can be used to determine:

[0168] S401, using a set duration as the look-ahead time, obtain a historical prediction sequence consisting of historical point prediction values ​​of the power of the predicted object under the look-ahead time; the set duration refers to the time interval from the current moment of power system assessment to the prediction time point.

[0169] That is, for each predicted time point within the period to be evaluated, based on the different time interval between the predicted time point and the current time, when quantifying the degree of mutual influence between the predicted objects at each predicted time point, the predicted value of historical points with the same forward-looking prediction time is selected for calculation, thereby controlling the influence of the time interval variable on the prediction result.

[0170] Based on the historical true values ​​obtained in the aforementioned embodiments, the predicted historical point values ​​corresponding to the historical true values ​​include: the predicted values ​​of the predicted object at the historical time point where the historical true values ​​are located, obtained by making forward predictions with different forward time periods. In this embodiment, the forward time is taken as the time interval from the current time to each prediction time point. For the true value sequence formed by the historical true values ​​obtained in the aforementioned embodiments, the sequence of historical point predicted values ​​corresponding to each historical true value at the forward time is obtained as the historical prediction sequence corresponding to each prediction time point.

[0171] For example, if the period to be evaluated is the 3rd to 5th day in the future, and three prediction time points are set in days, then the time intervals from the current moment to each prediction time point in the aforementioned period to be evaluated are 3 days, 4 days, and 5 days, respectively. For all historical true values ​​in the true value sequence, the predicted value is obtained by making a forward prediction 3 days before the historical time point where the historical true value is located. The sequence formed is the historical point prediction value sequence corresponding to the first prediction time point with a time interval of 3 days. Similarly, the historical prediction sequences corresponding to the second and third prediction time points can be obtained.

[0172] Suppose the truth sequence of the object to be predicted is [a1, a2, a3…a… n ], then the first prediction time point corresponds to the historical prediction sequence predicted 3 days ago [b 31 b 32 b 33 …b 3n The second prediction time point corresponds to the historical prediction sequence predicted 4 days ago. 41 b 42 b 43 …b 4n The third prediction time point corresponds to the historical prediction sequence made 5 days ago. 51 b 52 b 53 …b 5n ]; where b 3n b 4n b 5n All are a n The predicted value at the given historical time point.

[0173] S402, Generate a historical prediction error sequence at the prediction time point based on the true value sequence formed by the historical true values ​​of the predicted object power and the historical prediction sequence;

[0174] For each prediction time point, each predicted value in the historical prediction sequence corresponding to that prediction time point is compared with the corresponding true value, and the difference (i.e., error) is calculated to obtain the historical prediction error sequence for that prediction time point.

[0175] Given the truth value sequence [a1, a2, a3…a] of a certain prediction object mentioned above. n Taking the historical prediction sequence corresponding to the three prediction time points as an example, the historical prediction error sequence of the prediction object at the first prediction time point can be expressed as W1=[a1-b 31 a2-b 32 …a n -b 3n Similarly, the historical prediction error sequences at the second and third prediction time points can be represented as W2 = [a1 - b] 41a2-b 42 …a n -b 4n W3 = [a1-b] 51 a2-b 52 …a n -b 5n ].

[0176] S403, determine the correlation coefficients between each pair of historical prediction error sequences of each prediction object at each prediction time point, and determine them as the correlation coefficients between the corresponding power probability distributions.

[0177] The correlation coefficient of the historical prediction error sequence can be calculated by means of Pearson correlation coefficient, distance correlation, etc., and this application does not limit it.

[0178] For each predicted object at each predicted time point, the correlation coefficient between each pair of historical prediction error sequences is calculated and used as the correlation coefficient between the power probability distributions of the corresponding two predicted time points. For example, at the three predicted time points within the next 3-5 days, the two predicted objects correspond to six historical prediction error sequences. The correlation coefficient between each pair of these six historical prediction error sequences is calculated and used as the correlation coefficient between the corresponding power probability distributions.

[0179] Taking the historical prediction error sequences W1, W2, and W3 of a certain prediction object at three prediction time points as an example, the correlation coefficient between W1 and W2 is calculated as the correlation coefficient between the power probability distribution of the prediction object at the first prediction time point and the second prediction time point. Similarly, the correlation coefficient between the historical prediction error sequences of two prediction objects at the same prediction time point is the correlation coefficient between the power probability distributions of the two prediction objects at that prediction time point.

[0180] In this embodiment of the disclosure, by utilizing the correlation coefficient of historical prediction error sequences, the correlation between prediction results of the prediction model at different time intervals is evaluated, providing a data foundation for the construction of the joint probability prediction model and helping to improve the prediction accuracy of the joint probability prediction model.

[0181] In some embodiments, a joint probability prediction model is used to output a joint probability distribution regarding power. To convert this joint probability distribution regarding power into a probability distribution regarding the net difference in electricity consumption, runtime information needs to be incorporated. Based on this principle, this embodiment provides an acquisition process based on linear superposition and transformation for the probability distribution of the uncertain net difference in electricity consumption between the uncertain power source and load during the evaluation period as described in step S104 above. See [link to relevant documentation]. Figure 5 The flowchart shown may include:

[0182] S501, According to the joint probability prediction model, obtain the first joint probability distribution of each predicted object in the period to be evaluated; the first joint probability distribution includes multiple different independent variables; each independent variable is used to represent the power value of a predicted object at one prediction time point in the period to be evaluated.

[0183] The first joint probability distribution describes the distribution of the joint probability of the power values ​​of multiple prediction objects included in the power system at different prediction time points within the period to be evaluated. The number of independent variables depends on the combination of prediction time points and multiple prediction objects within the period to be evaluated.

[0184] For example, if the period to be evaluated is the next 2-3 days, and one prediction time point is set on the 2nd and 3rd days respectively, assuming that the power system includes two uncertain power sources and one uncertain load, and the power value of each prediction object at each prediction time point is used as an independent variable, then the first joint probability distribution includes 6 independent variables, and its joint probability is used to represent the probability of the occurrence of different power value combinations of the 6 independent variables.

[0185] For example, consider the following combinations of power values ​​for the three predicted objects at two prediction time points:

[0186] Forecast target 1 (wind farm): Day 2, 12:00 PM: Power range of 50MW to 150MW; Day 3, 12:00 PM: Power range of 40MW to 140MW;

[0187] Forecast Target 2 (Photovoltaic Power Plant): Day 2, 12:00 PM: Power range of 100MW to 200MW; Day 3, 12:00 PM: Power range of 80MW to 180MW;

[0188] Forecast Target 3 (Residential Electricity Load): Day 2, 12:00 PM: Power range 250MW to 350MW; Day 3, 12:00 PM: Power range 240MW to 340MW

[0189] The first joint probability distribution describes the joint probability of all possible power value combinations for the three predicted objects at the two prediction time points. An example of the assumed power value combinations and their corresponding joint probabilities is as follows:

[0190] The wind farm will be charged at 12:00 on the second day, with a capacity of 100MW; the photovoltaic power station will be charged at 150MW; and the residential electricity load will be charged at 300MW.

[0191] The wind farm will be charged at 12:00 on the third day, with a capacity of 120MW; the photovoltaic power station will be charged at 160MW; and the residential electricity load will be charged at 320MW.

[0192] This power combination corresponds to a joint probability value, such as 0.05, indicating that the actual probability of this combination occurring in the next 2-3 days is 5%. The first joint probability distribution will include all possible combinations and their corresponding probability values, thus comprehensively describing the uncertainty of the power values ​​of the three predicted objects in the next 2-3 days.

[0193] S502, Based on the power superposition matrix, determine the first linear transformation relationship between the net difference in uncertain power consumption during the period to be evaluated and the independent variable of the first joint probability distribution;

[0194] The first linear transformation relationship is a linear expression that transforms the independent variables of the first joint probability distribution into the net difference of uncertain electricity through the electricity superposition matrix. This linear transformation relationship describes the direct mapping from power to electricity and takes into account the contributions of all uncertain power sources and loads throughout the entire period to be evaluated.

[0195] Based on the fact that the net difference in uncertain electricity volume during the entire evaluation period is the difference between the total electricity generated by all uncertain power sources in the power system during the evaluation period and the total electricity consumed by all uncertain loads during the evaluation period, and considering that electricity volume is the product of power and operating time, the net difference in uncertain electricity volume can be obtained by linearly transforming the independent variables of the first joint probability distribution through the electricity volume superposition matrix.

[0196] For example, all the independent variables of the first joint probability distribution can be arranged into a one-dimensional matrix in a certain order. By transforming the matrix, this one-dimensional matrix can be multiplied by the power superposition matrix, and the product result can represent the net difference in uncertain power. Then the above first linear transformation relationship is a matrix transformation based on the power superposition matrix.

[0197] S503, based on the first joint probability distribution and combined with the first linear transformation relationship, the probability distribution of the uncertain net difference in electricity consumption is derived.

[0198] The method for deriving the probability distribution of the uncertain net electricity difference from the first joint probability distribution can be determined based on the distribution type of the first joint probability distribution. If the first joint probability distribution is a continuous distribution, the probability distribution of the uncertain net electricity difference can be derived by replacing the independent variable in the first joint probability distribution with the uncertain net electricity difference through variable substitution and integration. If the first joint probability distribution is difficult to derive analytically directly, numerical methods (such as Monte Carlo simulation) can be used to approximate the probability distribution of the uncertain net electricity difference. This involves simulating the combinations of independent variable values ​​in the first joint distribution through random sampling, calculating the corresponding net electricity difference using the electricity superposition matrix, and finally obtaining the probability distribution statistically.

[0199] If the first joint probability distribution satisfies a Gaussian mixture distribution, based on the closure property of the Gaussian mixture distribution under linear transformation (i.e., the Gaussian distribution remains Gaussian after linear transformation), the probability distribution of the uncertain net difference in electricity consumption can be derived from the first joint probability distribution in the following manner, including the following steps:

[0200] Obtain the first statistical parameter of the first joint probability distribution, the first statistical parameter including at least the mean and covariance matrix; according to the first linear transformation relationship, obtain the parameter transformation relationship between the first statistical parameter and the second statistical parameter of the probability distribution of the uncertain net difference in electricity consumption; according to the parameter transformation relationship, transform the first statistical parameter and determine the transformed statistical parameter as the second statistical parameter; according to the second statistical parameter, determine the Gaussian mixture distribution of the uncertain net difference in electricity consumption as the probability distribution of the uncertain net difference in electricity consumption.

[0201] The first statistical parameters include the mean and covariance matrix of each Gaussian component, as well as the weight of the component. The mean and covariance matrix describe the position and shape of each Gaussian component, and the weight indicates the relative importance of the component in the mixed distribution. When the independent variable X of the first joint probability distribution undergoes a first linear transformation to obtain the uncertain net difference in electricity Y, the first statistical parameters of the Gaussian distribution will also change accordingly. Since the Gaussian distribution remains a Gaussian distribution after the linear transformation, a new Gaussian distribution can be constructed based on the result of the transformation of the first statistical parameters, which is the probability distribution of the uncertain net difference in electricity.

[0202] For example, if the first joint probability distribution is a Gaussian mixture distribution, and its independent variable-vector X includes the power variables of each predicted object at each prediction time point, and the net difference in uncertain electricity consumption Y is obtained by transforming the independent variable-vector X of the first joint probability distribution through the electricity consumption superposition matrix A, that is, Y is a linear transformation of X, then the probability distribution of Y also conforms to a Gaussian mixture distribution, and the mean of each component in the probability distribution of Y is A*μ. m +C, the covariance matrix is ​​A∑ m A T A Gaussian distribution with weights ω m , where μ m This is the mean of the components corresponding to this component in the first joint probability distribution, ∑ m ω m These are the covariance matrix and weight coefficients corresponding to the first joint probability distribution, respectively. Based on the mean, covariance matrix, and weight coefficients of Y, the probability distribution of the uncertain net difference in electricity consumption Y can be obtained.

[0203] In some embodiments, given that the joint probability prediction model is known to have linear superposition, it can be assumed that the working time of each predicted object within the time intervals between adjacent prediction time points in the period to be evaluated is independent, and that its impact on the final probability distribution can be linearly superimposed. Therefore, for step S104 above, which involves determining the probability distribution of the net difference in uncertain electricity between uncertain power sources and loads in the period to be evaluated based on the joint probability prediction model and the electricity superposition matrix, see [link to relevant documentation]. Figure 6A The schematic diagram of the calculation process shown illustrates another method for obtaining the probability distribution of the net uncertainty of electricity consumption in this embodiment. For each time period within the period to be evaluated, the probability distribution of the net uncertainty of electricity consumption in each time period can be obtained first. By merging the probability distributions of the net uncertainty of electricity consumption in each time period, the probability distribution of the net uncertainty of electricity consumption in the entire period to be evaluated is generated. Combined with... Figure 6B The flowchart shown illustrates the steps involved. The probability distribution of the net difference in uncertain electricity consumption over the entire evaluation period can be obtained through the following steps:

[0204] S601, using the joint probability prediction model, obtain the second joint probability distribution of each prediction object at each prediction time point within the period to be evaluated;

[0205] In this embodiment, for each prediction time point within the period to be evaluated, the above-mentioned joint probability prediction model can be used to obtain the second joint probability distribution of all predicted objects at each prediction time point. The second joint probability distribution can be expressed as a joint probability distribution density function, which describes the joint probability of different power value combinations of each predicted object at the prediction time point.

[0206] S602, for each time period between adjacent prediction time points, based on the electricity superposition matrix, determine the second linear transformation relationship between the net difference in uncertain electricity within the time period and the independent variable of the second joint probability distribution;

[0207] The period to be evaluated is divided into multiple time periods based on the predicted time point. For each time period, the working time of each predicted object can be determined based on the power superposition matrix. The independent variables of the second joint probability distribution based on each predicted time point include the power variables of each predicted object at the predicted time point. Then, based on the fact that the power is the product of the power and the working time, the second linear transformation relationship between the net difference of uncertain power in the time period where the predicted time point is located and the power variables in the second joint probability distribution can be obtained.

[0208] S603, based on the second linear transformation relationship and the second joint probability distribution at the predicted time points included in the time period, obtain the probability distribution of the net difference in uncertain electricity consumption for the time period;

[0209] After obtaining the second linear transformation relationship between the uncertain net difference in electricity consumption for that time period and the independent variable of the second joint probability distribution, the probability distribution of the uncertain net difference in electricity consumption can be derived from the second joint probability distribution using this second linear transformation relationship, such as by using multiple integrals, Jacobi transformation, etc.

[0210] For example, suppose a power system includes uncertain power sources P1 and P2 and uncertain load L. The power variables corresponding to one time period in the period to be evaluated are defined as random variables x1, x2, and x3, respectively. If the corresponding working hours are 8, 16, and 24 hours, then the net difference in uncertain power is the total power of the two uncertain power sources minus the total power of the uncertain load, i.e., z = 8x1 + 16x2 - 24x3. It is known that the second joint probability density function f(x1,x2,x3) describes the joint probability distribution of the power value combination. Based on the linear transformation relationship between z and the power variables x1, x2, and x3 of each predicted object, the second joint probability density function f(x1,x2,x3) can be integrated and the variables replaced to find the probability distribution density function f(z) of the net difference in uncertain power z.

[0211] Regarding the multiple integration of x1, x2, x3 to eliminate the power variable and leave z as the only variable, the constraint condition of z can be represented by the Dirac delta function δ, and the multiple integration f(z) can be derived by the following expression: f(z)=∫∫∫f(x1,x2,x3)δ(z-(8x1+16x2-24x3))dx1dx2dx3. During the integration process, for cases where the computational cost is high and the integration result is difficult to obtain, numerical integration methods such as Monte Carlo simulation or Gaussian integration can be used to approximate the integration.

[0212] S604, combine the probability distributions of the net difference in uncertain electricity consumption in each time period within the period to be evaluated to obtain the probability distribution of the net difference in uncertain electricity consumption for the entire period to be evaluated.

[0213] When merging the probability distributions of net electricity differences across different time periods, since convolution operations can describe the probability distribution of the sum of two random variables, this can be achieved through convolution merging. Mathematically, the convolution operation can be expressed as... This represents a convolution operation, which combines the probability distributions of the net electricity difference across different time periods to obtain the probability distribution of the uncertain net electricity difference for the entire period to be evaluated. During the convolution operation, numerical methods such as Monte Carlo simulation, discretization, or statistical software can be used to approximate the probability distribution after convolution, thus simplifying the calculation.

[0214] In this embodiment of the disclosure, the probability distribution density function of the net difference in electricity consumption in each time period is derived by means of variable transformation based on the joint probability prediction model. This can accurately reflect the interaction between uncertain power sources and loads, and the probability distributions of each time period are combined to obtain the probability distribution of the net difference in uncertain electricity consumption for the entire period to be evaluated, thereby improving computational efficiency.

[0215] In some embodiments, the generation of the predicted distribution of the power imbalance within the time period to be evaluated in step S105 can be based on the relationship between the power imbalance and the deterministic net power difference and the uncertain net power difference, combined with the probability distribution of the uncertain net power difference. When the operation scheduling plan includes reserve capacity, the relationship between the power imbalance and the reserve capacity in the operation scheduling plan should be further considered when calculating the power imbalance. Based on this, the predicted distribution of the power imbalance can be generated using any of the following methods:

[0216] Linear transformation superposition method: For the period to be evaluated, a third linear transformation relationship is determined between the power imbalance and the uncertain net power difference, as well as the deterministic net power difference; based on the probability distribution of the third linear transformation relationship and the uncertain net power difference, a predicted distribution of the power imbalance is generated;

[0217] Sampling method: Samples are drawn from the probability distribution of the uncertain net power imbalance, and the power imbalance value corresponding to the sample is determined according to the deterministic net power imbalance and the reserve capacity, so as to obtain a sample set of power imbalance values; the predicted distribution of the power imbalance is determined according to the sample set of power imbalance values.

[0218] The linear transformation superposition method is based on the linear relationship between power imbalance and the uncertain net power difference, the deterministic net power difference, and reserve capacity. Specifically, the power imbalance ΔE can be expressed as a linear transformation function of the uncertain net power difference, the deterministic net power difference, and reserve capacity. The linear transformation method can be determined based on the probability distribution type and statistical characteristics of the uncertain net power difference, and the statistical characteristics of the predicted distribution of power imbalance (such as mean, covariance, etc.) can be determined using this linear transformation relationship, thereby deriving the predicted distribution of power imbalance.

[0219] When using the linear transformation superposition method to generate the predicted distribution of power imbalance, a linear transformation model can be constructed based on the above linear relationship. The probability distributions of deterministic net power imbalance, uncertain net power imbalance, and reserve capacity are then used as inputs to the linear transformation model. The predicted distribution of power imbalance is output through the linear transformation relationship. The calculation is simple and easy to understand and implement.

[0220] Taking the probability distribution f(Z) of the uncertain net electricity deficit Z as conforming to a Gaussian mixture model as an example, the deterministic net electricity deficit is a known value E1, and the reserve capacity is a known value E2. The electricity imbalance ΔE can be expressed as ΔE = Z + E1 + E2. Based on this linear transformation, the probability distribution f(ΔE) of ΔE also conforms to a Gaussian mixture model. Assume that f(Z) consists of K Gaussian components, each with a weight w. k The mean is μ k The variance is σ 2 k Then the weight and variance of each component of f(ΔE) remain unchanged, and the mean is (μ k +E1+E2), and thus determine f(ΔE) as the probability distribution of electrical imbalance based on the weight, variance and mean of each component.

[0221] The sampling method is based on statistical simulation. This method first extracts a large number of samples from the probability distribution of the uncertain net power imbalance. For each sample, the power imbalance value corresponding to the sample is calculated based on the deterministic net power imbalance and the reserve capacity. All power imbalance values ​​constitute a new sample set. Based on the statistical characteristics of this new sample set, such as mean and variance, the predicted distribution of power imbalance is determined, which fully considers the uncertainty of power imbalance.

[0222] In this embodiment of the disclosure, the relationship between power imbalance and deterministic net power difference, uncertain net power difference and reserve capacity is used to predict power imbalance by combining the probability distribution of uncertain net power difference. This accurately predicts the power imbalance during the period to be evaluated, providing strong support for power dispatch and energy management.

[0223] In some embodiments, within a power system, after determining the probability distribution of power imbalance for the entire period to be evaluated, given that power dispatching and operational decisions typically require highly accurate time sensitivity, it is possible to further predict the power imbalance distribution for a specific time period within the period to be evaluated. This can be achieved by modifying the power superposition matrix in the scheduling plan for that specific time period, such as setting all matrix elements in the power superposition matrix except for the specified time period to 0. Based on the modified working matrix and the joint probability prediction model, the probability distribution of the uncertain net power difference is determined, thereby obtaining the predicted distribution of power imbalance for that specified time period.

[0224] In this embodiment of the disclosure, by adjusting the power superposition matrix, the output of the joint probability prediction model can be directly set according to the actual situation, avoiding the need to repeatedly re-join the joint probability prediction model when the power system operating state changes. Furthermore, based on the modification of the power superposition matrix, the power imbalance distribution of a specified time period can be obtained, providing more refined data support for power system scheduling and operation.

[0225] To enable those skilled in the art to better understand the implementation of the quantitative assessment method for power imbalance in new energy power systems provided in this application, this embodiment takes the scenario of predicting power imbalance in the next 3-5 days from the current point in time as an example, and uses the Gaussian mixture model to generate power imbalance prediction to illustrate the method.

[0226] First, let's illustrate this prediction scenario with an example:

[0227] Power system: Uncertain power sources include wind farm A and wind farm B; uncertain loads include load L.

[0228] The period to be evaluated: 3-5 days after the current point in time;

[0229] Prediction time points: Using a 24-hour sampling period, a sampling point is set on the 3rd, 4th and 5th days as the prediction time points, dividing the period to be evaluated into 3 sub-periods, with a time interval of 24 hours between the sampling points.

[0230] Based on the above predicted scenarios, such as Figure 7 As shown, the quantitative assessment method for power imbalance in the new energy power system provided in this embodiment can be implemented through the following process:

[0231] S701, based on Gaussian mixture model, constructs a first probability prediction model for each prediction object at multiple time scales, and obtains the power probability distribution of the prediction object at each prediction time point in the next 3-5 days based on the first probability prediction model of the prediction object.

[0232] Based on historical data for each prediction target, probabilistic prediction models with different time scales were pre-constructed as the first probabilistic prediction model, including at least three probabilistic prediction models with time scales of 3 days, 4 days, and 5 days. For the first prediction time point, the 3-day probabilistic prediction model for wind farm A was used as the target probabilistic prediction model, generating the probability density function f shown in Table 2 below. 11 (p), Similarly, using the 3-day timescale probabilistic prediction model of wind farm B and load L as the target probabilistic prediction model, the f shown in Table 2 below are generated respectively. 21 (p), f 31 (p), for the second prediction time point, a probability prediction model with a time scale of 4 days is selected to generate f. 12 (p), f 22 (p), f 32 (p), a probability prediction model with a time scale of 5 days is selected to generate f. 13 (p), f 23 (p), f 33 (p).

[0233] Predicted object / predicted time point Day 3 Day 4 Day 5 Wind Farm A <![CDATA[f 11 (p)]]> <![CDATA[f 12 (p)]]> <![CDATA[f 13 (p)]]> Wind Farm B <![CDATA[f 21 (p)]]> <![CDATA[f 22 (p)]]> <![CDATA[f 23 (p)]]> Uncertain load L <![CDATA[f 31 (p)]]> <![CDATA[f 32 (p)]]> <![CDATA[f 33 (p)]]>

[0234] Table 2. Power probability distribution of each predicted object at each prediction time point in the next 3-5 days.

[0235] The power probability distribution is obtained by modeling the Gaussian mixture model based on the first probability prediction model and has linear superposition. Each power probability distribution includes multiple Gaussian components, and each Gaussian component has a mean, variance, and weight.

[0236] S702, based on the error data between the historical true values ​​and the predicted values ​​of each predicted object, obtain the correlation coefficient matrix between each power probability distribution at each prediction time point, and based on the correlation coefficient matrix and the covariance of each power probability distribution, obtain the covariance matrix between each power probability distribution, and establish their joint probability prediction model; the joint probability prediction model conforms to the Gaussian mixture model and has linear superposition.

[0237] Using the predicted objects and prediction time points shown in Table 2 as a division, and taking the power of the predicted object at each prediction time point as an independent variable, the independent variables of the established joint probability prediction model include 9 power variables, which can be represented as a vector X = [x 11 x 12 x 13 x 21 x 22 x 23 x 31 x 32 x 33 The joint probability distribution obtained from this joint probability prediction model during the evaluation period is used to output the joint probability of the nine power variables taking different possible combinations of values.

[0238] S703 establishes a power overlay matrix based on the operation scheduling plan for the next 3-5 days;

[0239] To facilitate the calculation of linear transformations between probability distributions, the power superposition matrix can be represented as a one-dimensional matrix. Each element indicates the working duration of a predicted object between two adjacent predicted time points, and multiple elements of the same predicted object are placed consecutively. For example, assuming that wind farm A operates for 0.5 days, 1 day, and 1 day in three time periods, wind farm B operates for 1 day, does not operate, and 0.5 days in three time periods, and the uncertain load L operates continuously in three time periods, then the power superposition matrix can be represented as [0.5, 1, 1, 1, 0, 0.5, 1, 1, 1].

[0240] S704, obtain the linear transformation relationship between the uncertain net difference in electricity consumption and multiple independent variables of the joint probability distribution, and the electricity consumption superposition matrix, and determine the mean, variance and weight of the probability distribution of the uncertain net difference in electricity consumption based on the linear transformation relationship, and generate the probability distribution of the uncertain net difference in electricity consumption.

[0241] Based on the premise that electricity is the product of power and time, and that the net difference in uncertain electricity is the difference between the total electricity generated by the uncertain power source and the total electricity consumed by the uncertain load, the above linear transformation relationship can be described as the transformation of a vector X composed of multiple independent variables of a joint probability distribution, where the net difference in uncertain electricity is obtained by matrix A. Matrix A refers to the result matrix obtained by multiplying the electricity superposition matrix by the sampling interval at the prediction time point. Since the total electricity consumed by the uncertain load needs to be subtracted, the elements corresponding to the uncertain load in the electricity superposition matrix can be set to negative numbers.

[0242] For example, in the above-mentioned superimposed electrical quantity matrix [0.5,1,1,1,0,0.5,1,1,1], the vector X = [x 11 x 12 x 13 x 21 x 22 x 23 x 31 x 32 x 33 If the same element in the same position describes the same predictor variable, that is, the element in the power superposition matrix that is in the same position as the independent variable in vector X describes the working state of that independent variable, then matrix A can be represented as A = 24 * [0.5, 1, 1, 1, 0, 0.5, 1, 1, 1]. The net difference in uncertain power Y = A * X T .

[0243] Based on the linear superposition property of Gaussian mixture models, if vector X is represented by a Gaussian mixture model and random variable Y is a linear transformation of X, then the distribution of Y also conforms to a Gaussian mixture model, and each Gaussian component of the Gaussian mixture model of Y has a mean of A*μ. m +C, the covariance matrix is ​​AΣ m A T A Gaussian distribution with weights ω m .

[0244] If the Gaussian mixture model expression for the joint probability prediction model is:

[0245]

[0246] Among them, f X (x) represents the joint probability density function of vector X; ω mThese are weighting coefficients; in the third formula, W in the denominator represents the dimension of vector X; N m (·) represents the multidimensional normal distribution, as the m-th Gaussian component of the Gaussian mixture model; det(·) represents the matrix determinant; M represents the total number of Gaussian components; μ m and σ m Let represent the mean vector and covariance of the m-th Gaussian component, respectively.

[0247] In the case of uncertain net electricity difference Y = A*X T In the case where the mean of each Gaussian component in the probability distribution of Y is A*μ m The covariance matrix is ​​AΣ m A T The weighting coefficient is ω m Y conforms to a Gaussian mixture model, and its probability distribution can be expressed as:

[0248]

[0249] S705: Based on the operation scheduling plan, obtain the deterministic net power imbalance and reserve capacity for the next 3-5 days. Based on the linear transformation relationship between power imbalance and uncertain net power imbalance, and combined with the statistical characteristics of the probability distribution of uncertain net power imbalance, determine the statistical characteristics of the probability distribution of power imbalance, and generate the predicted distribution of power imbalance.

[0250] Based on the probability distribution f of the aforementioned uncertain net electricity difference y Y Taking (y) as an example, given the deterministic net power deficit E1 and reserve capacity E2, the power imbalance ΔE is expressed as ΔE=Y+E1+E2. Based on this linear transformation, for the probability distribution of ΔE, the weight and variance of each component of f(ΔE) and its relationship with f Y In (y), the weights and variances are the same, and the mean is determined by f. Y A*μ in (y) m Transformed into (A*μ) m The predicted distribution of this electrical imbalance (+E1+E2) can be expressed as:

[0251]

[0252] Where, N m (ΔE) represents the mean of each component as (A*μ) m +E1+E2), covariance is AΣ m A T The weighting coefficient is ω m The Gaussian mixture model distribution.

[0253] In this embodiment, a Gaussian mixture model is used to predict the power probability distribution across multiple time scales, and a joint probability prediction model is constructed. This model combines the power superposition matrix with the deterministic power source conditions in the power system dispatch plan to effectively predict and assess the power imbalance in the power system over a specified future period. A series of steps generate the predicted distribution of the power imbalance, providing data support for power system dispatch. Since the Gaussian mixture model can well fit complex probability distributions, the joint probability prediction model constructed using the probability prediction results from the Gaussian mixture model can more accurately capture uncertainties. Furthermore, due to the linear superposition property of the Gaussian mixture model, the predicted distribution of the power imbalance can be calculated through a linear transformation relationship, simplifying the calculation process and making the prediction results easier to interpret and understand.

[0254] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0255] Based on the same inventive concept, and corresponding to the aforementioned implementation of the quantitative assessment method for power imbalance in new energy power systems, see [link to relevant documentation]. Figure 8 As shown, this application also provides an embodiment of a quantitative assessment device for power imbalance in a new energy power system. The solution provided by this device is similar to the solution described in the above-described method. Therefore, the specific limitations in the device embodiments provided below can be found in the limitations of the above-described method embodiments, and will not be repeated here. The device may include:

[0256] The power probability distribution prediction module 801 is used to obtain the power probability distribution of the predicted object at each prediction time point during the period to be evaluated by using a first probability prediction model of the predicted object; the predicted object includes uncertain power sources and loads; the model is constructed based on the error data between the historical predicted values ​​and the historical true values ​​of the power of the predicted object.

[0257] The joint probability prediction construction module 802 is used to construct a joint probability prediction model based on the power probability distribution of each prediction object at each prediction time point and the correlation coefficient between the power probability distributions.

[0258] The superposition relationship acquisition module 803 is used to acquire the operation scheduling plan within the period to be evaluated and determine the power superposition matrix; the power superposition matrix is ​​used to characterize the running time of each prediction object in each time period within the period to be evaluated; the time period refers to the time period between each prediction time point within the period to be evaluated;

[0259] The uncertainty power difference distribution prediction module 804 is used to determine the probability distribution of the net uncertainty power difference between uncertain power sources and loads during the period to be evaluated, based on the joint probability prediction model and the power superposition matrix.

[0260] The power imbalance distribution prediction module 805 is used to determine the net difference between deterministic power and load power in the period to be evaluated according to the operation scheduling plan, and generate a predicted distribution of power imbalance in the period to be evaluated based on the probability distribution of the net difference between deterministic power and the net difference between uncertain power.

[0261] In some embodiments, the apparatus further includes:

[0262] Obtain the historical true value of the predicted object within the historical period;

[0263] Obtain the historical point prediction value of the historical true value at the historical time point. The historical point prediction value includes the point prediction value of the prediction object at the historical time point obtained by making forward predictions with different forward time points.

[0264] Modeling is performed on the error data distribution between the predicted historical point value and the actual historical value at each look-ahead time, resulting in multiple probabilistic prediction models at different look-ahead time scales, which serve as the first probabilistic prediction model for the predicted object; the first probabilistic prediction model has linear superposition property.

[0265] In some embodiments, where the first probabilistic prediction model includes multiple probabilistic prediction models with different look-ahead time scales, the power probability distribution prediction module is specifically used for:

[0266] The time interval from the current moment of power system assessment to the predicted time point is obtained, and a target probability prediction model whose forward time scale conforms to the time interval is determined from the first probability prediction model of the predicted object.

[0267] Obtain the actual power of the predicted object at the current time;

[0268] Based on the actual power and the target probability prediction model, the power probability distribution of the predicted object at the prediction time point is obtained.

[0269] In some embodiments, the apparatus further includes:

[0270] Using a set duration as the look-ahead time, a historical prediction sequence is obtained, consisting of historical point prediction values ​​of the power of the predicted object under the look-ahead time. The set duration refers to the time interval from the current moment of power system assessment to the prediction time point.

[0271] Based on the true value sequence formed by the historical true values ​​of the predicted object power and the historical prediction sequence, a historical prediction error sequence at the prediction time point is generated.

[0272] Determine the correlation coefficients between each pair of historical prediction error sequences for each prediction object at each prediction time point, and use these coefficients as the correlation coefficients between the corresponding power probability distributions.

[0273] In some embodiments, the uncertainty power difference distribution prediction module is specifically used for:

[0274] Based on the joint probability prediction model, a first joint probability distribution of each predicted object is obtained within the period to be evaluated; the first joint probability distribution includes multiple different independent variables; each independent variable is used to represent the power value of a predicted object at one prediction time point within the period to be evaluated.

[0275] Based on the power superposition matrix, determine the first linear transformation relationship between the net difference in uncertain power consumption during the period to be evaluated and the independent variable of the first joint probability distribution;

[0276] Based on the first joint probability distribution of each predicted object during the period to be evaluated, and combined with the first linear transformation relationship, the probability distribution of the uncertain net difference in electricity consumption is derived.

[0277] In some embodiments, when the first joint probability distribution satisfies a Gaussian mixture distribution, the uncertainty electricity difference distribution prediction module, when deriving the probability distribution of the net uncertainty electricity difference based on the first joint probability distribution of each prediction object within the period to be evaluated, combined with the first linear transformation relationship, includes:

[0278] Obtain the first statistical parameters of the first joint probability distribution, wherein the first statistical parameters include at least the mean and the covariance matrix;

[0279] Based on the first linear transformation relationship, the parameter transformation relationship between the first statistical parameter and the second statistical parameter of the probability distribution of the uncertain net difference in electricity consumption is obtained;

[0280] Based on the parameter transformation relationship, the first statistical parameter is transformed, and the transformed statistical parameter is determined as the second statistical parameter.

[0281] Based on the second statistical parameter, a Gaussian mixture distribution of the uncertain net difference in electricity consumption is determined as the probability distribution of the uncertain net difference in electricity consumption.

[0282] In some embodiments, the uncertainty power difference distribution prediction module is specifically used for:

[0283] Using the joint probability prediction model, the second joint probability distribution of each prediction object at each prediction time point within the period to be evaluated is obtained;

[0284] For each time period between adjacent prediction time points, a second linear transformation relationship is determined between the net difference in uncertain electricity consumption within the time period and the independent variables of the second joint probability distribution, based on the electricity consumption superposition matrix.

[0285] Based on the second linear transformation relationship and the second joint probability distribution at the predicted time points included in the time period, the probability distribution of the net difference in uncertain electricity consumption for the time period is obtained.

[0286] By combining the probability distributions of the net difference in uncertain electricity consumption in each time period within the period to be evaluated, the probability distribution of the net difference in uncertain electricity consumption for the entire period to be evaluated is obtained.

[0287] In some embodiments, the power imbalance distribution prediction module is specifically used for:

[0288] For the period to be evaluated, a third linear transformation relationship is determined between the power imbalance and the uncertain net power difference, as well as the certain net power difference.

[0289] Based on the probability distribution of the third linear transformation relationship and the net difference in uncertain electricity, a predicted distribution of the electricity imbalance is generated.

[0290] In some embodiments, the power imbalance distribution prediction module is specifically used for:

[0291] Samples are drawn from the probability distribution of the uncertain net power imbalance, and the power imbalance value corresponding to the sample is determined according to the deterministic net power imbalance and the reserve capacity, thus obtaining a sample set of power imbalance values; the predicted distribution of the power imbalance is determined according to the sample set of power imbalance values.

[0292] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0293] This application also provides an electronic device, the structural schematic diagram of which is shown below. Figure 9As shown, the electronic device 900 includes at least one processor 901, a memory 902, and a bus 903. At least one processor 901 is electrically connected to the memory 902. The memory 902 is configured to store at least one computer-executable instruction, and the processor 901 is configured to execute the at least one computer-executable instruction to perform the steps of any quantitative assessment method for power imbalance in a new energy power system provided in any embodiment or optional implementation of this application.

[0294] Furthermore, the processor 901 can be an FPGA (Field-Programmable Gate Array) or other devices with logic processing capabilities, such as an MCU (Microcontroller Unit) or a CPU (Central Processing Unit).

[0295] This application also provides another readable storage medium storing a computer program that, when executed by a processor, implements the steps of any quantitative assessment method for power imbalance in a new energy power system provided in any embodiment or optional implementation of this application.

[0296] The readable storage media provided in this application include, but are not limited to, any type of disk (including floppy disk, hard disk, optical disk, CD-ROM, and magneto-optical disk), ROM (Read-Only Memory), RAM (Random Access Memory), EPROM (Erasable Programmable Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), flash memory, magnetic cards, or optical cards. In other words, the readable storage media includes any medium by which a device (e.g., a computer) stores or transmits information in a readable form.

[0297] This application also provides a computer program product, which may include a computer program / instruction. When the computer program / instruction is executed by a processor, it implements any quantitative assessment method for power imbalance in a new energy power system provided in any embodiment or optional implementation of this application.

[0298] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The processors involved in the embodiments provided in this application can be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited thereto.

[0299] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification. The above embodiments only illustrate several implementation methods of this application, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of this application. It should be noted that for those skilled in the art, several modifications and improvements can be made without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A quantitative assessment method for power imbalance in a new energy power system, characterized in that, The method includes: Using a first probability prediction model for the predicted object, the power probability distribution of the predicted object at each prediction time point within the period to be evaluated is obtained; the predicted object includes uncertain power sources and loads; the first probability prediction model is constructed based on the error data between the historical predicted values ​​and historical actual values ​​of the predicted object's power; the power probability distribution represents a probability distribution density function, used to describe all possible power values ​​of the predicted object at each prediction time point within the period to be evaluated and their corresponding probabilities; A joint probability prediction model is constructed based on the power probability distribution of each predicted object at each prediction time point and the correlation coefficient between the power probability distributions; the correlation coefficient is used to describe the degree of correlation and mutual influence of the power probability distributions of different predicted objects at different prediction times; the joint probability prediction model is used to provide the probability that each predicted object will occur simultaneously under different power value combinations. Obtain the operation scheduling plan for the period to be evaluated and determine the power superposition matrix; the power superposition matrix is ​​used to characterize the running time of each predicted object in each time period within the period to be evaluated; the time period refers to the time period between each predicted time point within the period to be evaluated; Based on the joint probability prediction model and the power superposition matrix, the probability distribution of the net difference in uncertain power between uncertain power sources and loads during the period to be evaluated is determined; the probability distribution of the net difference in uncertain power describes the possible values ​​and probabilities of the uncertain power difference in the power system during the entire period to be evaluated. According to the operation scheduling plan, the net difference between the deterministic power supply and the load during the period to be evaluated is determined, and a predicted distribution of power imbalance during the period to be evaluated is generated based on the probability distribution of the net difference between the deterministic power supply and the net difference between the uncertain power supply. The predicted distribution of power imbalance is a probabilistic description of the power supply and demand imbalance during the period to be evaluated.

2. The method according to claim 1, characterized in that, The method further includes a first probabilistic prediction model construction step, comprising: Obtain the historical true value of the power of the predicted object; Obtain the historical point prediction value of the historical time point where the historical true value is located. The historical point prediction value includes the point prediction value of the prediction object at the historical time point obtained by making forward predictions with different forward time points. Modeling is performed on the error data distribution between the predicted historical point value and the actual historical value at each look-ahead time, resulting in multiple probabilistic prediction models at different look-ahead time scales, which serve as the first probabilistic prediction model for the predicted object; the first probabilistic prediction model has linear superposition property.

3. The method according to claim 2, characterized in that, The step of using a first probability prediction model of the predicted object to obtain the power probability distribution of the predicted object at each prediction time point within the period to be evaluated includes: The time interval from the current moment of power system assessment to the predicted time point is obtained, and a target probability prediction model whose forward time scale conforms to the time interval is determined from the first probability prediction model of the predicted object. Obtain the actual power of the predicted object at the current time; Based on the actual power and the target probability prediction model, the power probability distribution of the predicted object at the prediction time point is obtained.

4. The method according to claim 2, characterized in that, The method further includes a step of determining the correlation coefficient between the power probability distributions, including: Using a set duration as the look-ahead time, a historical prediction sequence is obtained, consisting of historical point prediction values ​​of the power of the predicted object under the look-ahead time. The set duration refers to the time interval from the current moment of power system assessment to the prediction time point. Based on the true value sequence formed by the historical true values ​​of the predicted object power and the historical prediction sequence, a historical prediction error sequence at the prediction time point is generated. Determine the correlation coefficients between each pair of historical prediction error sequences for each prediction object at each prediction time point, and use these coefficients as the correlation coefficients between the corresponding power probability distributions.

5. The method according to claim 1, characterized in that, The step of determining the probability distribution of the net difference in uncertain electricity levels between uncertain power sources and loads during the period to be evaluated, based on the joint probability prediction model and the electricity superposition matrix, includes: Based on the joint probability prediction model, a first joint probability distribution of each predicted object is obtained within the period to be evaluated; the first joint probability distribution includes multiple different independent variables; each independent variable is used to represent the power value of a predicted object at one prediction time point within the period to be evaluated. Based on the power superposition matrix, determine the first linear transformation relationship between the net difference in uncertain power consumption during the period to be evaluated and the independent variable of the first joint probability distribution; Based on the first joint probability distribution and combined with the first linear transformation relationship, the probability distribution of the uncertain net difference in electricity consumption is derived.

6. The method according to claim 5, characterized in that, When the first joint probability distribution satisfies a Gaussian mixture distribution, the step of deriving the probability distribution of the uncertain net difference in electricity consumption based on the first joint probability distribution and the first linear transformation relationship includes: Obtain the first statistical parameters of the first joint probability distribution, wherein the first statistical parameters include at least the mean and the covariance matrix; Based on the first linear transformation relationship, the parameter transformation relationship between the first statistical parameter and the second statistical parameter of the probability distribution of the uncertain net difference in electricity consumption is obtained; Based on the parameter transformation relationship, the first statistical parameter is transformed, and the transformed statistical parameter is determined as the second statistical parameter. Based on the second statistical parameter, a Gaussian mixture distribution of the uncertain net difference in electricity consumption is determined as the probability distribution of the uncertain net difference in electricity consumption.

7. The method according to claim 1, characterized in that, The step of determining the probability distribution of the net difference in uncertain electricity levels between uncertain power sources and loads during the period to be evaluated, based on the joint probability prediction model and the electricity superposition matrix, includes: Using the joint probability prediction model, the second joint probability distribution of each prediction object at each prediction time point within the period to be evaluated is obtained; For each time period between adjacent prediction time points, a second linear transformation relationship is determined between the net difference in uncertain electricity consumption within the time period and the independent variables of the second joint probability distribution, based on the electricity consumption superposition matrix. Based on the second linear transformation relationship and the second joint probability distribution at the predicted time points included in the time period, the probability distribution of the net difference in uncertain electricity consumption for the time period is obtained. The probability distributions of the net difference in uncertain electricity consumption in each time period within the period to be evaluated are combined to obtain the probability distribution of the net difference in uncertain electricity consumption for the entire period to be evaluated.

8. The method according to claim 1, characterized in that, The step of generating a predicted distribution of the power imbalance during the period to be evaluated based on the probability distributions of the deterministic net power imbalance and the uncertain net power imbalance includes: For the period to be evaluated, a third linear transformation relationship is determined between the power imbalance and the uncertain net power difference, as well as the certain net power difference. Based on the probability distribution of the third linear transformation relationship and the net difference in uncertain electricity, a predicted distribution of the electricity imbalance is generated.

9. A quantitative assessment device for power imbalance in a new energy power system, characterized in that, The device includes: The power probability distribution prediction module is used to obtain the power probability distribution of the predicted object at each prediction time point within the evaluation period using a first probability prediction model of the predicted object; the predicted object includes uncertain power sources and loads; the first probability prediction model is constructed based on the error data between the historical predicted values ​​and historical true values ​​of the predicted object's power; the power probability distribution represents a probability distribution density function, used to describe all possible power values ​​of the predicted object at prediction time points within the evaluation period and their corresponding probabilities; The joint probability prediction construction module is used to construct a joint probability prediction model based on the power probability distribution of each prediction object at each prediction time point and the correlation coefficient between the power probability distributions; the correlation coefficient is used to describe the degree of correlation and mutual influence of the power probability distributions of different prediction objects at different prediction times; the joint probability prediction model is used to provide the probability that each prediction object will occur simultaneously under different power value combinations. The superposition relationship acquisition module is used to acquire the operation scheduling plan within the period to be evaluated and determine the power superposition matrix; the power superposition matrix is ​​used to characterize the running time of each prediction object in each time period within the period to be evaluated; the time period refers to the time period between each prediction time point within the period to be evaluated; The uncertainty power difference distribution prediction module is used to determine the probability distribution of the net uncertainty power difference between uncertain power sources and loads during the period to be evaluated, based on the joint probability prediction model and the power superposition matrix; the probability distribution of the net uncertainty power difference describes the possible values ​​and probabilities of the uncertainty power difference of the power system during the entire period to be evaluated. The power imbalance distribution prediction module is used to determine the deterministic net difference between deterministic power and load in the period to be evaluated according to the operation scheduling plan, and generate a predicted distribution of power imbalance in the period to be evaluated based on the probability distribution of the deterministic net difference and the uncertain net difference; the predicted distribution of power imbalance is a probabilistic description of the power supply and demand imbalance in the period to be evaluated.

10. An electronic device, characterized in that, include: Memory, processor; The memory is used to store computer programs; The processor is configured to invoke the computer program to implement the method as described in any one of claims 1-8.

11. A readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-8.

12. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instruction is executed by the processor, it implements the method as described in any one of claims 1-8.

Images