Power system frequency modulation speed demand quantification method, equipment and medium
By constructing a deep neural network model and an adaptive bandwidth core density estimation model, predicting and quantifying the short-term power fluctuation of the power system's net load, the power balance problem caused by random fluctuations in the existing technology is solved, and the power balance problem caused by random fluctuations in new energy is improved. The safety and economicality of power grid operation.
Patent Information
- Application Number
- CN202410980540.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-22
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2044-07-22
AI Technical Summary
The existing frequency modulation method of power system fails to effectively consider the system's short-term power fluctuations, making it difficult to maintain real-time power balance in the random fluctuations of new energy, threatening the safety of power grid operation.
By constructing a sample feature vector for the prediction of the maximum power fluctuation of the net load, screening the training samples, building a deep neural network model to predict the maximum power fluctuation point of the net load, and using adaptive bandwidth kernel density estimation to construct the short-term power of the net load Maximum fluctuation interval prediction model, obtain the upper limit of the confidence interval as the system frequency modulation speed requirement.
It realizes accurate prediction of the maximum power fluctuation in short-term net load, ensures the abundance of frequency modulation speed requirements, provides reasonable scheduling decision-making reference, and improves the safety and economicality of power grid operation.
Smart Images

Figure CN118920510A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of power systems and automation technologies thereof, and in particular to a method for quantifying frequency regulation speed requirements of power systems. Background Art
[0002] Building a new power system with new energy as the main body is a key way to implement my country's "dual carbon" goals. However, the strong random fluctuation characteristics of new energy have brought severe challenges to the safe and stable operation of the power grid. Reasonable arrangement of resource operation mode through power system dispatching decision-making is a key link in maintaining the real-time balance of supply and demand in the power system. Existing power system dispatching decision-making methods mainly include uncertain dispatching decision-making methods and deterministic dispatching decision-making methods.
[0003] Uncertain scheduling decision-making methods can take into account the impact of uncertainty in renewable energy, and thus have attracted extensive attention from scholars at home and abroad. They mainly include: multi-scenario method, chance constraint method, robust optimization method, etc. Among them, the multi-scenario optimization method generates typical scenarios based on the random fluctuation characteristics of renewable energy, and takes the lowest expected operating cost of the system under the typical scenario as the objective function, and takes the operating safety under different scenarios as the constraint condition to formulate the optimal decision plan. However, the number of scenarios reflecting the random fluctuation characteristics of renewable energy explodes with the increase of the dimension of random variables representing the uncertainty of renewable energy, and the solution efficiency in large-scale power systems is difficult to meet the actual computing needs; the chance constraint method ensures that the probability of system operation constraint exceeding the limit under the influence of random fluctuation of renewable energy is lower than the set threshold, and the balance between power grid operation safety and economy can be achieved by adjusting the threshold size, but the chance constraint model can only be converted into a second-order cone form when the output of renewable energy meets a specific distribution (such as normal distribution), otherwise it needs to be solved by sampling iteration and other methods, and the computational complexity is high. Robust optimization methods ensure that power system dispatch can effectively cope with extreme scenarios of random fluctuations in renewable energy, and can ensure the safety of system operation in the worst scenario. However, this method is relatively conservative and it is difficult to take into account the economic efficiency of system operation. It is generally a non-convex two-layer optimization problem with high computational complexity. In summary, although the uncertainty dispatch decision method can take into account the impact of renewable energy uncertainty, the computational complexity of this type of method is usually high and it is difficult to meet the computational time requirements of actual power system dispatch decisions.
[0004] In order to meet the calculation time requirements of dispatching decisions, the industry currently generally adopts a deterministic dispatching decision method with a small calculation burden. The way to deal with the random fluctuation characteristics of new energy is to introduce the system frequency modulation capacity demand in the dispatching decision link, so that the unit reserves sufficient frequency modulation capacity to ensure that the units participating in frequency modulation in actual operation have sufficient output adjustment space to deal with the random fluctuations of new energy. The principle of setting the frequency modulation capacity demand is to make the unit reserved capacity greater than the maximum deviation of the system net load relative to the unit's planned output, so as to ensure that the unit has sufficient operating margin to deal with the random fluctuations of new energy. The industry usually sets the frequency modulation capacity demand based on operating experience, such as setting the frequency modulation capacity demand to 2%-5% of the load peak in each period, or setting it to a fixed ratio of the load and new energy forecast values. This method of quantifying the frequency modulation capacity demand is simple and easy to operate, but it fails to consider the impact of the random fluctuation characteristics of new energy. To improve this problem, domestic and foreign scholars have successively proposed other improvement methods, including methods based on the distribution of new energy forecast errors and data-driven methods. However, in order to cope with the impact of the random fluctuation characteristics of renewable energy, existing methods mainly focus on the system frequency regulation capacity requirements to ensure the system regulation margin, without considering the impact of short-term power fluctuations in the system. In fact, even if the unit has reserved sufficient frequency regulation capacity, if the adjustment speed of the units involved in frequency regulation is difficult to effectively follow the short-term fluctuations of renewable energy, the system will not be able to effectively maintain the real-time balance of electricity, threatening the safe operation of the power grid. Therefore, in addition to the system frequency regulation capacity requirements, it is necessary to introduce the maximum short-term power fluctuation of the system net load to ensure that the adjustment speed of the units involved in frequency regulation can effectively cope with the short-term power fluctuations of the net load. Summary of the invention
[0005] The present invention aims to at least solve one of the technical problems in the prior art, namely, the lack of a reliable and effective method for frequency regulation of a power system and the lack of consideration for short-term power fluctuations in the traditional power system frequency regulation strategy.
[0006] To this end, a first aspect of the present invention provides a method for quantifying frequency regulation speed requirements of a power system.
[0007] A second aspect of the present invention provides a computer device.
[0008] A third aspect of the present invention provides a computer-readable storage medium.
[0009] The present invention provides a method for quantifying frequency regulation speed demand of a power system, comprising:
[0010] Constructing a sample feature vector for predicting the maximum short-term power fluctuation of net load according to the power grid operation data; wherein the power grid operation data includes load data, new energy output data, meteorological data and special date data; the maximum short-term power fluctuation of net load is defined as the maximum change value of net load per unit time in a fixed scheduling period; the sample feature vector includes an input feature vector and an output feature vector;
[0011] The training samples and sample feature vectors are screened according to the similarity between the training samples and the samples on the forecast day and the matching degree between the input feature vectors of the training samples and the samples on the forecast day;
[0012] A net load short-term power maximum fluctuation point prediction model is constructed, and the net load short-term power maximum fluctuation point prediction model is trained using the screened training samples and sample feature vectors;
[0013] The predicted value of the maximum fluctuation point of the net load short-time power of the system output by the net load short-time power maximum fluctuation point prediction model is input into the net load short-time power maximum fluctuation interval prediction model to obtain the confidence interval of the net load short-time power maximum fluctuation prediction result;
[0014] The upper limit of the confidence interval of the prediction result of the maximum short-term power fluctuation of the net load is taken as the system frequency regulation speed requirement.
[0015] The method for quantifying the frequency regulation speed demand of a power system according to the above technical solution of the present invention may also have the following additional technical features:
[0016] In the above technical solution, the input feature vector includes: load power characteristics, new energy power characteristics, net load power characteristics, meteorological factor characteristics and special date information;
[0017] Wherein, the special date information includes holiday type;
[0018] The output characteristic vector includes the maximum fluctuation value of the short-term power of the net load on the forecast day.
[0019] In the above technical solution, the screening of training samples includes:
[0020] The improved weighted Euclidean distance is used to quantify the matching degree between different samples and the predicted day samples:
[0021]
[0022] Among them, q represents the load power characteristics, new energy power characteristics, net load power characteristics and net load short-term power fluctuation characteristics on the day before the forecast; n represents the meteorological factor characteristics on the forecast day; r represents the holiday type on the forecast day; l represents the sample number; represents the weighted Euclidean distance of each type of feature; a and b are two w-dimensional sample data respectively, z represents the weight of each dimension feature, and the subscript is the data dimension number; d l The final Euclidean distance calculation result of the two samples;
[0023] The L samples with the smallest Euclidean distance are used as training samples.
[0024] In the above technical solution, the screening of sample feature vectors includes:
[0025] Quantify the matching degree of each dimension feature between the training sample and the sample to be predicted:
[0026]
[0027]
[0028] Among them, w l Represents the training sample weight constructed by the normalized Euclidean distance between the training sample and the sample to be predicted. The smaller the Euclidean distance, the greater the training sample weight, and the greater the impact on the quantization result of the input feature vector matching; They represent the maximum and minimum values of the Euclidean distance of the training samples after screening; L represents the total number of training samples after screening; p represents the input feature dimension; represents the Euclidean distance between different input feature vectors in the lth sample and the corresponding input feature vector on the forecast day; d p Indicates the matching degree of the p-th dimension feature between the final quantized training sample and the sample to be predicted;
[0029] Select the T feature vectors with the highest matching degree as the final input feature vectors.
[0030] In the above technical solution, the net load short-term power maximum fluctuation range prediction model includes:
[0031]
[0032] Among them, h i represents the adaptive bandwidth of the i-th sample point; h0 represents the global baseline bandwidth; λ represents the adjustment parameter; g(x) represents the density estimation at x; f(x) is the probability density function; K represents the kernel function; n represents the number of samples; x i represents the i-th sample point predicted by the net load short-term power maximum fluctuation point prediction model;
[0033]
[0034] Among them, the cumulative distribution function F(x) represents the probability that the random variable X is less than or equal to a specific value x;
[0035] The PPF function is obtained based on the inverse function of the cumulative distribution function F(x):
[0036] b=F -1 (r) Where b represents the quantile corresponding to the cumulative probability r, and r represents the cumulative probability.
[0037] In the above technical solution, the net load short-term power maximum fluctuation range prediction model includes:
[0038]
[0039] Where f(x) is the probability density function; h represents the fixed bandwidth; K represents the kernel function; n represents the number of samples; x i represents the i-th sample point predicted by the net load short-term power maximum fluctuation point prediction model;
[0040]
[0041] Among them, the cumulative distribution function F(x) represents the probability that the random variable X is less than or equal to a specific value x;
[0042] The PPF function is obtained based on the inverse function of the cumulative distribution function F(x):
[0043] b=F -1 (r)
[0044] Where b represents the quantile corresponding to the cumulative probability r, and r represents the cumulative probability.
[0045] In the above technical solution, a deep neural network is used to construct the net load short-term power maximum fluctuation point prediction model.
[0046] In the above technical solution, a data-driven model is used to construct the net load short-term power maximum fluctuation point prediction model;
[0047] The data-driven model includes an extreme learning machine, a decision tree, a long short-term memory network or a convolutional neural network.
[0048] The present invention also provides a computer device, including a processor and a memory, wherein a computer program is stored in the memory, and when the computer program is loaded and executed by the processor, a method for quantifying the frequency regulation speed demand of a power system as described in any one of the above technical solutions is implemented.
[0049] The present invention also provides a computer-readable storage medium storing a program, which, when loaded by a processor, implements the method for quantifying the frequency regulation speed demand of a power system as described in any one of the above technical solutions.
[0050] In summary, due to the adoption of the above technical features, the beneficial effects of the present invention are:
[0051] Based on the load, new energy and meteorological information actually available in the power grid, the present invention constructs a sample feature vector for predicting the maximum fluctuation of net load short-term power, and simultaneously screens the training samples and the sample feature vector, thereby establishing a prediction model for the maximum fluctuation point of net load short-term power, and realizing the accurate prediction of the maximum fluctuation of net load short-term power. In order to ensure the sufficiency of the frequency modulation speed demand, a prediction model for the maximum fluctuation interval of net load short-term power based on adaptive bandwidth kernel density estimation (ABKDE) is further established on the basis of the point prediction model, and the confidence interval of the prediction result of the maximum fluctuation of net load short-term power can be obtained. The upper limit of the interval is used as the maximum fluctuation of the short-term power of the net load of the system to ensure that it can cover the short-term maximum fluctuation of the actual net load. Through the method proposed by the present invention, the maximum fluctuation of the short-term power of the net load of the system can be accurately quantified, providing a powerful reference for the reasonable formulation of the day-ahead scheduling decision plan.
[0052] Additional aspects and advantages of the present invention will become apparent from the following description or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0053] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:
[0054] Figure 1 is a flow chart of a method for quantifying frequency regulation speed requirements of a power system according to an embodiment of the present invention;
[0055] Figure 2 It is a schematic diagram of the deep neural network structure used in the method for quantifying the frequency regulation speed demand of the power system according to an embodiment of the present invention. DETAILED DESCRIPTION
[0056] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.
[0057] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.
[0058] Refer to the following Figure 1 and Figure 2The following describes a method for quantifying frequency regulation speed requirements of a power system provided according to some embodiments of the present invention.
[0059] Some embodiments of the present application provide a method for quantifying frequency regulation speed requirements of a power system.
[0060] The first embodiment of the present invention proposes a method for quantifying frequency regulation speed demand of a power system, including steps S1-S5.
[0061] S1. Construct a sample feature vector for predicting the maximum short-term power fluctuation of net load based on the power grid operation data; wherein the power grid operation data includes load data, new energy output data, meteorological data and special date data; the maximum short-term power fluctuation of net load is defined as the maximum change value of net load per unit time within a fixed scheduling period; the sample feature vector includes an input feature vector and an output feature vector.
[0062] Specifically, the input feature vector should include the main influencing factors of the maximum fluctuation of the short-term power of the net load of the system. In some embodiments, the input feature vector includes: load power characteristics, new energy power characteristics, net load power characteristics, meteorological factor characteristics and special date information; wherein, the special date information includes holiday types, and it can be understood that special dates refer to specific dates that are different from regular dates, such as holidays, major event days, etc.; the output feature vector includes the maximum fluctuation value of the short-term power of the net load on the forecast day.
[0063] In a specific embodiment, the fixed scheduling period can be set to 15 minutes, and the unit duration can be set to 1 minute, then the maximum short-term power fluctuation of the net load means: the maximum change value of the net load in 1 minute within the 15-minute scheduling period. It should be noted that the fixed scheduling period and the unit duration can be flexibly set as needed.
[0064] In a specific embodiment, the actual operation data of the power grid load and new energy in a province of China from January 2023 to December 2023 are collected, wherein the net load data with high new energy penetration rate is obtained by proportionally expanding the new energy data. In order to consider the impact of meteorological factors on the maximum short-term power fluctuation of the net load of the system, various meteorological values of the region are also obtained. At the same time, the impact of special dates such as holidays on the prediction results is also considered, and the holiday type of the prediction day is marked.
[0065] Specifically, in order to predict the maximum short-term power fluctuation of the net load in the next day, an input feature vector with a granularity of 15 minutes is constructed based on the data with a fine granularity of 1 minute. The constructed input feature vector and output feature vector are shown in Table 1.
[0066]
[0067] Among them, the input feature vector includes meteorological factors, date, and historical fluctuation characteristics of new energy, load, and net load, totaling 2689 dimensions; the output feature vector is the maximum short-term power fluctuation of net load on the forecast day, totaling 96 dimensions.
[0068] S2. Screening the training samples and sample feature vectors according to the similarity between the training samples and the prediction day samples and the matching degree between the input feature vectors of the training samples and the prediction day samples.
[0069] Specifically, the prediction accuracy of the net load short-term power maximum fluctuation prediction model is closely related to the similarity between the training sample and the prediction day sample and the matching degree between the input features of the training sample and the prediction day sample. To this end, in step S2, the sample feature matching degree is first quantified and calculated, and samples with high feature matching degree are selected as training samples; then, the matching degree between the input features of the training sample and the prediction day sample is quantified, and features with high matching degree are selected as the final sample input features.
[0070] In some embodiments, the screening of training samples includes:
[0071] The improved weighted Euclidean distance is used to quantify the matching degree between different samples and the predicted day samples:
[0072]
[0073] Among them, q represents the load power characteristics, new energy power characteristics, net load power characteristics and net load short-term power fluctuation characteristics on the day before the forecast; n represents the meteorological factor characteristics on the forecast day; r represents the holiday type on the forecast day; l represents the sample number; represents the weighted Euclidean distance of each type of feature; a and b are sample data of two w dimensions respectively, z represents the weight of each dimension feature (obtained by Pearson correlation coefficient), and the subscript is the data dimension number; d l The final Euclidean distance calculation result of the two samples;
[0074] The L samples with the smallest Euclidean distance (highest matching degree) are used as training samples.
[0075] In the above training sample screening method, the weighted Euclidean distance of each type of feature is calculated separately, and then the final Euclidean distance calculation result between two samples is obtained by integrating the Euclidean distance of each type of feature. The reason for this is that the dimensions of features such as new energy, load, and meteorological factors are much higher than the dimensions of holiday types on the forecast day. Traditional Euclidean distance will be more inclined to the larger-dimensional fluctuation characteristics of new energy, load, and net load. In contrast, the holiday type characteristics on the forecast day are easily ignored. The above-mentioned improved Euclidean distance calculation method can effectively alleviate this problem.
[0076] In some embodiments, using data that is most similar to the training day sample data for training can improve the model training effect and save model training time. The dimension of the input feature vector constructed in this embodiment is 2689. There may be redundancy in these input features or features that do not match the prediction day sample data, which will mislead the model training process, increase the model training time, and reduce the model prediction accuracy. To this end, this embodiment further screens the input features of the training samples to ensure that the input features ultimately used for training can maximize the effect of model training. Specifically, the screening of sample feature vectors includes:
[0077] Quantify the matching degree of each dimension feature between the training sample and the sample to be predicted:
[0078]
[0079] Among them, w l It represents the training sample weight constructed by the Euclidean distance between the normalized training sample and the sample to be predicted. The smaller the Euclidean distance, the greater the training sample weight and the greater the impact on the quantization result of the input feature vector matching; They represent the maximum and minimum values of the Euclidean distance of the training samples after screening; L represents the total number of training samples after screening; p represents the input feature dimension; represents the Euclidean distance between different input feature vectors in the lth sample and the corresponding input feature vector on the forecast day; d p Indicates the matching degree of the p-th dimension feature between the final quantized training sample and the sample to be predicted;
[0080] Select the T feature vectors with the highest matching degree as the final input feature vectors.
[0081] Finally, L training samples with input features of T dimensions are formed, and the input feature dimension of the sample to be predicted is also reduced to T dimensions.
[0082] In some embodiments, the meteorological factors, new energy fluctuation characteristics, load fluctuation characteristics, and maximum short-term power fluctuation of net load contained in the training samples have different dimensions and large numerical differences, which is not conducive to the training of subsequent models. For this reason, the samples need to be normalized and preprocessed. The present disclosure uses the z-score method to normalize the input data and output data of the samples. The z-score standardization method uses the sample mean and standard deviation for normalization. The data processed by z-score satisfies the mean of 0 and the standard deviation of 1. The specific normalization method is:
[0083]
[0084] Among them, x* is the normalized sample value, x is the sample value to be normalized, μ is the sample mean, and σ is the sample standard deviation. It should be noted that sample preprocessing is usually performed before step S2.
[0085] S3. Construct a net load short-time power maximum fluctuation point prediction model, and use the screened training samples and sample feature vectors to train the net load short-time power maximum fluctuation point prediction model.
[0086] In some embodiments, DNN (deep neural network) is selected to construct a prediction model for the maximum fluctuation point of net load short-term power. Deep neural networks show good prediction performance due to their powerful nonlinear feature extraction capabilities, efficient processing of large-scale data, and robustness in noisy data environments, and the DNN calculation time can also meet the scheduling time. Therefore, the present invention selects DNN to construct a prediction model for the maximum fluctuation point of net load short-term power. It should be noted that the selection of which data-driven model to construct the point prediction model is not the focus of the present invention, and other data-driven models (extreme learning machine, decision tree, long short-term memory network, convolutional neural network, etc.) can also be used to predict the maximum fluctuation point of net load short-term power.
[0087] The trained DNN can be used to predict the specific value of the maximum short-term power fluctuation of the system net load in each scheduling period on the operating day.
[0088] In a specific embodiment, the deep neural network structure used in the present disclosure is as follows: Figure 2 As shown in the figure, it includes an input layer, multiple hidden layers and an output layer. At the same time, the layers are connected together in a fully connected manner. The data transfer formula between layers is:
[0089]
[0090] in, Represents the first d The output of the layer neurons, l d =1,2,…n d , input x is represented as a 0 ; To connect the first d-1 Layer and l d The weight matrix of the layer neurons; For the first d The output of the last layer of neurons (i.e. the final output of the neural network) is expressed as follows:
[0091]
[0092] Among them, function s(x) is the activation function. In the hidden layer, this disclosure selects LReLU (Leaky Rectified Linear Unit) as the activation function. The LReLU activation function is a piecewise function. When the input value x is less than or equal to zero, the output value is equal to 0.01x; when the input value x is greater than zero, the output value is equal to the input value x.
[0093] s(x)=max(0.01x,x)
[0094] In the output layer, a linear activation function is used, and its expression is as follows:
[0095] s(x)=x
[0096] The training of DNN uses labeled samples to adjust the parameters θ = {W, b} so that the loss function L o Minimum. The loss function used in this disclosure is the mean square error loss function, as shown below:
[0097]
[0098] Among them, M s is the number of samples; m is the sample number; is the output value of DNN; y is the actual value of the sample.
[0099] The present disclosure uses a learning algorithm that combines root mean square propagation and learning rate decay to train the DNN model. The root mean square propagation algorithm can keep the moving average of the square gradient of the parameter unchanged, so that the step size of each parameter is updated adaptively, thereby accelerating the convergence speed of the parameter. In addition, as the training progresses, the learning rate decay strategy can gradually reduce the learning rate, avoiding the algorithm from oscillating back and forth. The parameter update process of the learning algorithm used in the present disclosure is as follows:
[0100]
[0101]
[0102] In the formula, is the kth i The gradient of θ at the time of update; is at the kth i Update the parameters before the first update; is at the kth i The partial derivative of the loss function with respect to θ at the update; is at the kth i The moving average of the squared gradient of the parameter at the update; ρ is the decay rate, which is set to 0.9; ⊙ represents the Hadamard product; is at the kth i The amount of change in the parameter during the update; is at the kth i The learning rate for the update; δ is to prevent And set a very small constant; is at the kth i The updated parameters; η0 represents the initial learning rate, and its value is set to 0.001; ε represents the decay rate of the learning rate, and its value is set to 0.95.
[0103] S4. Input the predicted value of the maximum fluctuation point of the net load short-time power of the system output by the net load short-time power maximum fluctuation point prediction model into the net load short-time power maximum fluctuation interval prediction model to obtain the confidence interval of the net load short-time power maximum fluctuation prediction result.
[0104] Through steps S1-S3, a point prediction is made for the maximum short-term power fluctuation in the scheduling period. However, due to the existence of data-driven errors, the maximum short-term power fluctuation of the net load obtained by the point prediction may be greater than the actual maximum short-term power fluctuation of the net load, which in turn leads to insufficient reserved capacity for system auxiliary services, thereby reducing the system frequency performance. Therefore, in order to avoid the point prediction value being less than the actual value, the present disclosure selects the net load short-term power maximum fluctuation interval prediction model for interval prediction. If the interval prediction upper limit is selected as the frequency modulation speed requirement, the situation where the net load short-term power maximum fluctuation prediction value is less than the actual value can be avoided to a great extent.
[0105] In some embodiments, KDE (a non-parametric method for estimating probability density function) is selected to smooth each sample point and have an impact on the surrounding space. This impact is quantified by a kernel function. The KDE estimated density function is:
[0106]
[0107] Where f(x) is the probability density function; h represents the fixed bandwidth; K represents the kernel function; n represents the number of samples; x i It represents the i-th sample predicted by the prediction model of the maximum fluctuation point of net load short-term power.
[0108] However, although KDE can estimate the probability density of sample points to a certain extent, the fixed bandwidth h in KDE may lead to problems such as oversmoothing or overfitting when the data distribution is uneven. In order to solve this problem, in some embodiments, ABKDE with an adaptive bandwidth mechanism is adopted to change the bandwidth according to the local density of data points. That is, in areas with sparse data, the bandwidth is increased to reduce the variance of the estimate; and in areas with dense data, the bandwidth is reduced to reduce the deviation of the estimate. The ABKDE estimated density function is:
[0109]
[0110] Among them, h i represents the adaptive bandwidth of the i-th sample point; h0 represents the global baseline bandwidth; λ represents the adjustment parameter; g(x) represents the density estimation at x; f(x) is the probability density function; K represents the kernel function; n represents the number of samples; x i It represents the i-th sample point predicted by the prediction model of the maximum short-term power fluctuation point of the net load.
[0111] On this basis, in order to ensure that the predicted value is slightly larger than the actual value of the actual operation scheduling point, the present disclosure introduces the Cumulative Distribution Function (CDF) and the Percent Point Function (PPF).
[0112] The cumulative distribution function F(x) represents the probability that the random variable X is less than or equal to a specific value x, and its expression is:
[0113]
[0114] CDF is a non-decreasing function with a value between 0 and 1. The PPF function is obtained from the inverse function of the cumulative distribution function F(x):
[0115] b=F -1 (r)
[0116] Among them, b represents the quantile corresponding to the cumulative probability r, and r represents the cumulative probability. r is usually a value between 0 and 1. Taking the confidence interval of 95% as an example, r is 0.975; b represents the predicted value with a confidence interval of 95%.
[0117] S5. The upper limit value of the confidence interval of the prediction result of the maximum short-term fluctuation of the net load power is used as the system frequency regulation speed requirement.
[0118] Through the above steps S1-S5. The present disclosure screens the training sample data based on the improved Euclidean distance and obtains the weight, introduces it into the input feature screening composed of 2689-dimensional feature vectors, and proposes a DNN-based net load short-term power maximum fluctuation point prediction model. While ensuring that the scheduling time allows, the powerful nonlinear feature extraction capability of the DNN model is utilized to achieve point prediction of the maximum short-term power fluctuation. Then, combined with the ABKDE maximum short-term power fluctuation interval prediction method, the upper limit of the maximum fluctuation of the net load short-term power under a certain confidence level can be obtained. If the predicted maximum fluctuation of the net load short-term power is taken as the upper limit value, the true value can be greatly smaller than the predicted value, greatly improving the safety and economy of the power system. Therefore, the present disclosure can be widely used in the day-ahead scheduling decision of the power system, provide a reference for the quantification of the system frequency regulation speed demand, and ensure that the system scheduling decision plan can effectively respond to the net load power fluctuation.
[0119] In a specific embodiment, the actual operation data of a certain power grid (historical frequency regulation mileage and meteorological data, etc.) is used to verify the effectiveness of the proposed point prediction model and the interval-based prediction model. This embodiment adopts a sliding window prediction mode to perform load forecasting, that is, a corresponding load forecasting model is trained for each day to be predicted, and the training sample data is fixed. For example: If the data on February 11 is used as the sample to be predicted, the available data includes all data from February 1 to February 10 as input samples. If the data on February 12 is predicted, the available data is all data from February 2 to February 11, and so on.
[0120] The net load short-time power maximum fluctuation point prediction model is referred to as the DNN data-driven prediction model, and the net load short-time power maximum fluctuation interval prediction model is referred to as the interval probability prediction model ABKDE.
[0121] The training method comparison is mainly based on the following schemes M1-M6:
[0122] M1: DNN data-driven prediction model, which uses improved Euclidean distance to screen training samples and input features, and the 60 samples most similar to the input data of the prediction day are selected as training samples.
[0123] M2: A DNN data-driven prediction model that has not undergone training sample screening. It selects the 60 days adjacent to the day to be predicted as training data and performs feature screening.
[0124] M3: DNN data-driven prediction model without input feature screening, 60 samples most similar to the input data on the prediction day were selected as training samples.
[0125] M4: Add interval probability prediction model ABKDE based on M1.
[0126] M5: Add the interval probability prediction model ABKDE based on M2.
[0127] M6: Add interval probability prediction model ABKDE based on M3.
[0128] In order to verify the effectiveness of the net load short-term power maximum fluctuation interval prediction method M4 proposed in the present disclosure, this embodiment will use the evaluation indicators R2, MAE and RMSE to discuss the accuracy of the deterministic prediction model, and use MAPE, PICP and PINAW to discuss the interval probability prediction model.
[0129] Root Mean Square Error (RMSE):
[0130]
[0131] RMSE is a standardized measure of error. Larger errors will be penalized more heavily due to the square term. The smaller the value, the better.
[0132] Mean Absolute Percentage Error (MAPE):
[0133]
[0134] MAPE is the average value of the ratio of the absolute value of the prediction error to the true value. It is used to measure the accuracy of the prediction. The smaller the value, the better.
[0135] Prediction Interval Coverage Probability (PICP):
[0136]
[0137] Where, L n and U n are the upper and lower bounds of the nth predicted value respectively. I is the indicator function, which takes a value between 0% and 100%. The larger the value, the better.
[0138] Prediction Interval Normalized Average Width (PINAW):
[0139]
[0140] Where Y max ,Y minare the maximum and minimum values of the true value, respectively. PINAW is an indicator to measure the width of the prediction interval, which aims to balance the accuracy and reliability of the prediction. The smaller the value, the better.
[0141] This embodiment uses the point prediction model constructed by DNN to predict the maximum fluctuation demand of the short-term power of the net load of the system on February 11, 2023. The average RMSE and MAPE of M1-M3 on February 11, 2023 are shown in Table 2.
[0142] Table 2 Prediction accuracy of maximum short-term power fluctuation point of M1-M3 net load
[0143] Evaluation Criteria M1 M2 M3 RMSE 4.96 9.01 7.03 MAPE 3.37% 8.30% 6.21%
[0144] As can be seen from Table 2, the MAPE and RMSE of the comparison methods M1-M3 and M2-M3 are respectively greater than those of M1 after the training samples and input features are screened, which verifies the effectiveness of screening training samples and input features in improving the accuracy of the prediction model. It can be seen that the DNN prediction model combined with the input features and training sample screening can accurately predict the maximum short-term power fluctuation of the net load. Despite this, in the 96 scheduling period points throughout the day on February 11, 2023, method M1 still has nearly 50% of the scheduling period predictions for the maximum short-term power fluctuation of the net load that do not meet the requirements of being greater than the actual maximum short-term power fluctuation of the net load, and cannot effectively guarantee the safety of the system.
[0145] Therefore, in order to ensure that the predicted value of the maximum short-term power fluctuation of the net load in 96 scheduling periods is greater than the actual value as much as possible, the ABKDE interval probability prediction model is introduced. In order to reserve enough maximum short-term power fluctuation of the net load for the system, the present invention adopts the interval prediction upper limit as the frequency modulation speed requirement. The PINAW and PICP values of methods M4-M6 are shown in Table 3.
[0146] Table 3 Interval prediction accuracy of M4-M6
[0147] Evaluation indicators M4 M5 M6 PINAW 10.12% 13.96% 12.43% PICP 100% 84.38% 90.63%
[0148] As shown in Table 3, compared with methods M4-M6, the PICP of the frequency modulation speed demand of method M4 is 100%, and the PINAW value is smaller than that of methods M5-M6. It can be concluded that while method M4 ensures that the frequency modulation speed demand of the entire system on February 11 is greater than the actual value, the average frequency modulation speed prediction value at 96 scheduling times is also smaller than that of methods M5-M6. It can be seen that the accuracy of point prediction has an impact on the interval probability prediction. The higher the accuracy of point prediction, the more the frequency modulation speed prediction result obtained can ensure system safety and the average width is relatively narrow. On the other hand, the PICP value of methods M5-M6, due to the insufficient accuracy of point prediction, the probability that the upper limit value of the interval prediction is greater than the actual value in 96 scheduling periods does not reach 100%, resulting in insufficient system reserved frequency modulation capacity in some scheduling periods because the predicted maximum short-term power fluctuation does not cover the actual value, which in turn causes the system frequency to be lower than normal.
[0149] The above example results show that the interval prediction quantification method based on the DNN point prediction model and ABKDE proposed in this disclosure can ensure that the upper limit value of the obtained prediction interval can effectively cover the true value of the maximum short-term power fluctuation of the net load, and the system frequency regulation speed requirement set based on the upper limit of the prediction interval can ensure the adequacy of the system setting requirements.
[0150] Some embodiments of the present invention also provide a computer device, including a processor and a memory, wherein the memory stores a computer program, and when the computer program is loaded and executed by the processor, the method for quantifying the frequency regulation speed demand of the power system as described in any of the above embodiments is implemented.
[0151] Some embodiments of the present invention further provide a computer-readable storage medium storing a program, which, when loaded by a processor, implements the method for quantifying the frequency regulation speed demand of a power system as described in any of the above embodiments.
[0152] In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.
[0153] Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A method for quantifying frequency regulation speed requirements of a power system, characterized in that: include: Constructing a sample feature vector for predicting the maximum short-term power fluctuation of net load according to the power grid operation data; wherein the power grid operation data includes load data, new energy output data, meteorological data and special date data; the maximum short-term power fluctuation of net load is defined as the maximum change value of net load per unit time in a fixed scheduling period; the sample feature vector includes an input feature vector and an output feature vector; The training samples and sample feature vectors are screened according to the similarity between the training samples and the samples on the forecast day and the matching degree between the input feature vectors of the training samples and the samples on the forecast day; A net load short-term power maximum fluctuation point prediction model is constructed, and the net load short-term power maximum fluctuation point prediction model is trained using the screened training samples and sample feature vectors; The predicted value of the maximum fluctuation point of the net load short-time power of the system output by the net load short-time power maximum fluctuation point prediction model is input into the net load short-time power maximum fluctuation interval prediction model to obtain the confidence interval of the net load short-time power maximum fluctuation prediction result; The upper limit of the confidence interval of the prediction result of the maximum short-term power fluctuation of the net load is taken as the system frequency regulation speed requirement.
2. The method for quantifying the frequency regulation speed demand of a power system according to claim 1, characterized in that: The input feature vector includes: load power characteristics, new energy power characteristics, net load power characteristics, meteorological factor characteristics and special date information; Wherein, the special date information includes holiday type; The output characteristic vector includes the maximum fluctuation value of the short-term power of the net load on the forecast day.
3. The method for quantifying the frequency regulation speed demand of a power system according to claim 2, characterized in that: The screening of training samples includes: The improved weighted Euclidean distance is used to quantify the matching degree between different samples and the predicted day samples: Among them, q represents the load power characteristics, new energy power characteristics, net load power characteristics and net load short-term power fluctuation characteristics on the day before the forecast; n represents the meteorological factor characteristics on the forecast day; r represents the holiday type on the forecast day; l represents the sample number; represents the weighted Euclidean distance of each type of feature; a and b are two w-dimensional sample data respectively, z represents the weight of each dimension feature, and the subscript is the data dimension number; d l The final Euclidean distance calculation result of the two samples; The L samples with the smallest Euclidean distance are used as training samples.
4. The method for quantifying the frequency regulation speed demand of a power system according to claim 3, characterized in that: The screening of sample feature vectors includes: Quantify the matching degree of each dimension feature between the training sample and the sample to be predicted: Among them, w l It represents the training sample weight constructed by the Euclidean distance between the normalized training sample and the sample to be predicted. The smaller the Euclidean distance, the greater the training sample weight and the greater the impact on the quantization result of the input feature vector matching; They represent the maximum and minimum values of the Euclidean distance of the training samples after screening; L represents the total number of training samples after screening; p represents the input feature dimension; represents the Euclidean distance between different input feature vectors in the lth sample and the corresponding input feature vector on the forecast day; d p Indicates the matching degree of the p-th dimension feature between the final quantized training sample and the sample to be predicted; Select the T feature vectors with the highest matching degree as the final input feature vectors.
5. The method for quantifying the frequency regulation speed demand of a power system according to claim 1, characterized in that: The net load short-term power maximum fluctuation range prediction model includes: Among them, h i represents the adaptive bandwidth of the i-th sample point; h0 represents the global baseline bandwidth; λ represents the adjustment parameter; g(x) represents the density estimation at x; f(x) is the probability density function; K represents the kernel function; n represents the number of samples; x i represents the i-th sample point predicted by the net load short-term power maximum fluctuation point prediction model; Among them, the cumulative distribution function F(x) represents the probability that the random variable X is less than or equal to a specific value x; The PPF function is obtained based on the inverse function of the cumulative distribution function F(x): b=F -1 (r) Where b represents the quantile corresponding to the cumulative probability r, and r represents the cumulative probability.
6. The method for quantifying the frequency regulation speed demand of a power system according to claim 1, characterized in that: The net load short-term power maximum fluctuation range prediction model includes: Where f(x) is the probability density function; h represents the fixed bandwidth; K represents the kernel function; n represents the number of samples; x i represents the i-th sample point predicted by the net load short-term power maximum fluctuation point prediction model; Among them, the cumulative distribution function F(x) represents the probability that the random variable X is less than or equal to a specific value x; The PPF function is obtained based on the inverse function of the cumulative distribution function F(x): b=F -1 (r) Where b represents the quantile corresponding to the cumulative probability r, and r represents the cumulative probability.
7. The method for quantifying the frequency regulation speed demand of a power system according to claim 1, characterized in that: A deep neural network is used to construct the net load short-term power maximum fluctuation point prediction model.
8. The method for quantifying the frequency regulation speed demand of a power system according to claim 1, characterized in that: A data-driven model is used to construct a prediction model for the maximum fluctuation point of the net load short-term power; The data-driven model includes an extreme learning machine, a decision tree, a long short-term memory network or a convolutional neural network.
9. A computer device, characterized in that: It comprises a processor and a memory, wherein a computer program is stored in the memory, and when the computer program is loaded and executed by the processor, the method for quantifying the frequency regulation speed demand of the power system as described in any one of claims 1 to 8 is implemented.
10. A computer-readable storage medium, characterized in that: A program is stored, and when the program is loaded by a processor, the method for quantifying the frequency regulation speed demand of a power system as described in any one of claims 1 to 8 is implemented.
Citation Information
Patent Citations
Power distribution network block division method considering uncertainty and correlation of wind and light load power
CN115439000A
Electric power spot market clearing decision-making method embedded with frequency modulation mileage service and medium
CN116436027A
Frequency modulation capacity demand quantification method based on credible prediction of fluctuation extreme value in scheduling time period
CN118297334A
Cited By
Explanatable net load power probability prediction method and system
CN119917824A
