Improved wind and light field station level power supply capability evaluation and medium and short term prediction method and device

Abstract

The invention relates to a novel electric power system power supply capability evaluation and prediction technology, in particular to an improved wind and light field station level power supply capability evaluation and medium and short term prediction method and device. A novel comprehensive evaluation index system for the power supply capability of the power generation side of the power system is constructed from two dimensions of output reliability of renewable energy sources and adjustable capability of an energy storage system; in the aspect of power supply capability prediction, an empirical mode decomposition-kernel principal component analysis-long short-term memory neural network prediction algorithm EMD-KPCA-LSTM improved by a sliding window method is provided. The method can effectively overcome the limitation of a new energy station power supply capability evaluation method based on a traditional output concept, can more comprehensively reflect the actual power supply capability of the power generation side of a novel power system, can better process nonlinear and non-stationary data, captures the long and short term dependency relationship in time series data, and improves the power supply capability of the new energy station. The method has certain advantages in new energy power supply capability prediction.

Description

Technical Field

[0001] This invention belongs to the field of new power system power supply capacity assessment and prediction technology, and particularly relates to improved wind and solar power station-level power supply capacity assessment and short-to-medium term prediction methods and equipment. Background Technology

[0002] To support the large-scale, high-quality development of renewable energy, assessing and forecasting the power supply capacity of renewable energy power plants is crucial. Traditional power supply capacity assessment and output forecasting methods are insufficient to fully reflect the actual power supply capacity of new energy power plants in the development of new power systems. For new power systems with large-scale wind and solar integration, the research focus is generally on system absorption capacity, i.e., from the perspective of renewable energy power sources, requiring the system to fully absorb or absorb as much of the generated electricity as possible, with virtually no wind or solar curtailment. Therefore, it is necessary to construct a more suitable assessment index system for power plants and collection stations based on their characteristics. When there is a difference between the actual and planned output of wind and solar power plants, this system should effectively describe the curtailment situation, accurately assess the power supply capacity, and improve power supply reliability, adjustability, and curtailment rate through energy storage regulation at power plant collection stations.

[0003] Furthermore, current power supply capacity assessments primarily focus on distribution networks, with limited research on power supply capacity assessments at the substation and collection station levels. Focusing on the generation side of new power systems requires greater consideration of the output characteristics of renewable energy sources and the tracking and coordination capabilities of energy storage; existing forecasting methods for short- to medium-term timescales are insufficient.

[0004] A search of existing technical fields revealed Chinese patent application number CN202411901319.1, publication number CN119944622A, entitled "Method, System and Equipment for Predicting Maximum Power Supply Capacity of Energy Storage Based on TCN-LSTM". This patent proposes a method based on TCN-LSTM. A method, system, and equipment for predicting the maximum power supply capacity of energy storage terminals using LSTM. This method first uses historical data from each load node at the energy storage terminal and employs TCN. The LSTM model predicts the load power of each load node, then superimposes the predictions to obtain the total load power prediction. Finally, an iterative power flow method is used to determine the maximum power supply capacity of the energy storage end. However, the concept of maximum power supply capacity proposed in this invention is not comprehensive enough, as it only considers the impact of the energy storage end and does not take into account the volatility and uncertainty of the generation side. Therefore, the prediction method has certain limitations.

[0005] Therefore, how to better assess and forecast the power supply capacity of wind and solar power plants in the short and medium term has become a technical problem that the industry urgently needs to solve. Summary of the Invention

[0006] This invention provides an improved method and equipment for assessing and predicting the power supply capacity of wind and solar power plants in the short and medium term, in order to solve the defects of the existing technology, such as imperfect assessment indicators for the power supply capacity of wind and solar power plants and immature methods for predicting the power supply capacity in the short and medium term, thereby improving the accuracy of the prediction of the power supply capacity of wind and solar power plants in the short and medium term, enhancing the grid integration and absorption capacity of new energy, and optimizing the grid dispatch strategy.

[0007] This invention provides an improved method for assessing and predicting the short-to-medium term power supply capacity of wind and solar power plants, comprising: Establish a 1-15 day power supply capacity assessment index system for wind power / photovoltaic power plants; Establish a method for predicting the power supply capacity of wind power / photovoltaic power plants for 1-15 days.

[0008] According to an improved method for assessing and predicting the power supply capacity of wind and solar power plants in the short to medium term, the establishment of a 1-15 day wind / solar power plant-level power supply capacity assessment index system includes: S1.1. New energy power supply capacity indicators based on reliability assessment: The output characteristics of new energy sources are quantitatively analyzed based on the power generation rate index GR, the power output interruption rate index PIP, the volatility index PF, the average normal power supply time index ANPST, and the average normal power supply quantity index ANES. S1.2. Adjustable capability index: The error parameters between the planned and actual output curves of each power station as they approach the planned output curve, as well as the parameters of power curtailment and power shortage, are used as adjustable capacity indicators.

[0009] According to an improved method for assessing and predicting the power supply capacity of wind and solar power plants in the short to medium term, the meaning and calculation method of the new energy power supply capacity index based on reliability assessment include: (1) Power generation rate index:

[0010] In the formula: p ( t )for t The output value of new energy sources at all times; P N This refers to the rated output power of the new energy source. T This represents the total operating time of the new energy source; the closer the index value is to 1, the closer the overall power generation is to the rated full power output. (2) Output interruption rate index:

[0011] In the formula: T The total operating time of the new energy source. This refers to the set of moments during which the output of renewable energy is zero within the operating time limit; the index describes the degree of discontinuity in the output of renewable energy, and the closer it is to 0, the fewer the instances of zero output. (3) Volatility indicators:

[0012] In the formula: p ( t )for t The power output of new energy sources at all times; p ( t +1) is t The output power of the new energy source at time +1; this indicator is calculated by the difference between the previous and next time points to obtain the overall output power. T The fluctuation of renewable energy output over a period of time can effectively reflect the fluctuation of renewable energy output power; the unit is MW / min; the smaller the value, the more stable the renewable energy output power. (4) Average normal power supply time index:

[0013] In the formula: For new energy sources to meet actual and predicted values ​​by their own output or in conjunction with energy storage, there are no periods of power curtailment or shortage. T This is the simulation time limit; the closer the index value is to 1, the fewer the periods of power curtailment and shortage. (5) Average normal power supply index:

[0014] In the formula: T The simulation time limit; N The number of times that the new energy source provides normal power supply within the simulation time limit; E j For the first j The power supply of new energy sources during normal operation; the unit of the indicator is MWh / min, which represents the average power supply per minute during normal operation.

[0015] According to an improved method for assessing and predicting the power supply capacity of wind and solar power plants in the short to medium term, the calculation of the adjustable capacity index includes: 1) Calculation of error parameters between planned and actual results; Mean absolute error (MAE), root mean square error (RMSE), and maximum tracking error (MTE): The calculation method is as follows:

[0016]

[0017]

[0018] Where: m is the number of sampling points within the simulation period; For the first i The predicted value at each time point; For the first i The actual value at each moment; The rated power; the maximum tracking error refers to the maximum difference between the measured and predicted values; 2) Calculation of parameters for power curtailment and power shortage; The curtailment rate (ASR) is used to represent the curtailment situation, and the calculation formula is as follows:

[0019] In the formula: To discard power that is lacking; is the actual generated power; m is the number of sampling points in the simulated time period; the smaller this value, the less power curtailment or shortage there is.

[0020] According to an improved method for assessing and predicting the power supply capacity of wind and solar power plants in the short to medium term, the method for establishing a 1-15 day wind / solar power plant-level power supply capacity prediction method includes the following steps: S2.1. EMD decomposition; Complex signals are decomposed into a finite number of intrinsic mode functions (IMFs). Each IMF component contains local characteristic signals of the original signal at different time scales. The EMD decomposition method is based on the following assumptions: (1) The data has at least two extreme values, one maximum and one minimum; (2) The local time-domain characteristics of the data are uniquely determined by the time scale between the extreme points; (3) If the data has no extreme points but has inflection points, the extreme values ​​can be obtained by differentiating the data once or multiple times, and then the decomposition results can be obtained by integration. S2.2. Dimensionality reduction using the KPCA algorithm; The KPCA algorithm is used to map the original data to a high-dimensional feature space, which makes the linear separability of the data in the space stronger. Then, PCA is used to reduce the dimensionality of the data. The mathematical expression for the radial basis function (RBF) is as follows:

[0021] in, x i and x j Indicates the input sample. These are parameters that control the shape of the kernel function; S2.3. LSTM neural network prediction; In the LSTM neuron transmission mechanism, the recurrent body unit has three input information, including the hidden layer state of the previous time series. and cell state and the input data for the current time series. ; The gating units of LSTM include an input gate, a forget gate, and an output gate; together they determine the update of the cell state and the final hidden state output. The gating units generate values ​​between 0 and 1 through the sigmoid function, representing the degree of acceptance of new information, the degree of forgetting of old information, and the degree of exposure of the cell state to the output, respectively. The forget gate allows the model to selectively forget information of past cell states based on the current input. The closer the result is to 0, the greater the probability of forgetting; the closer it is to 1, the smaller the probability of forgetting. The input gate is responsible for filtering the important information in the current input and integrating it into the new candidate state.

[0022] According to an improved method for assessing and predicting the power supply capacity of wind and solar power plants in the short to medium term, the EMD decomposition process is as follows: (1) Find all the maximum points of the original data sequence X(t) and fit the upper envelope of the original data using the cubic spline interpolation function; (2) Find all the local minimum points and fit all the local minimum points to form the lower envelope of the data using a cubic spline interpolation function; (3) The mean of the upper and lower envelopes is denoted as m(t). Subtracting the mean envelope m(t) from the original data sequence X(t) yields a new data sequence h(t). (4) If there are still negative local maxima and positive local minima after subtracting the envelope average from the original data, then continue the filtering.

[0023] According to an improved method for assessing and predicting the power supply capacity of wind and solar power plants in the short to medium term, the steps of the KPCA algorithm are as follows: (1) Calculate the kernel function: Calculate the kernel matrix of the original dataset. K ; (2) Centralized kernel matrix: The kernel matrix is... K Centralization is performed by column or row to obtain the centralized kernel matrix. K c); (3) Solving for eigenvalues ​​and eigenvectors: For the centered kernel matrix K Perform eigenvalue decomposition to solve for its eigenvalues ​​and eigenvectors; (4) Select principal components: Select the eigenvectors corresponding to the k largest eigenvalues ​​as principal components to construct a new dimensionality reduction space; (5) Data projection: Project the original dataset X onto a new dimension-reduced space to obtain the dimension-reduced data matrix Y; (6) Inverse mapping: The inverse transformation is used to restore the dimensionality-reduced data to the original space, so as to perform tasks such as classification and visualization.

[0024] According to an improved method for assessing and predicting the power supply capacity of wind and solar power plants in the short to medium term, the LSTM neural network prediction steps are as follows: (1) Data preprocessing: normalization processing to make the data distributed between [0,1], which is beneficial for model training; (2) Divide the training set and the test set: The training set is used to train the model, and the test set is used to test the accuracy of the model; (3) Define the LSTM model: Define the input layer, output layer, and intermediate layer of the LSTM as needed; (4) Training the model: Specify the training batch size, training epochs, and validation set; (5) Prediction: Use the trained model to make predictions.

[0025] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the improved wind and solar power station-level power supply capacity assessment and short-to-medium term prediction method as described above.

[0026] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the improved wind and solar power station-level power supply capacity assessment and short-to-medium term prediction method as described above.

[0027] Compared with the prior art, the beneficial effects of the present invention are: The power supply capacity proposed in this invention is divided into two parts. The first part not only focuses on the output of new energy sources but also takes into account their volatility, intermittency, and uncertainty, proposing a series of evaluation indicators, and finally weighting them to obtain a power supply capacity indicator based on reliability. The second part considers the scenario where new energy power plants are equipped with energy storage. When the power plant's output curve is close to the dispatch curve of the dispatch center, it will charge the energy storage during periods of high new energy generation and discharge the energy storage during periods of low output to smooth the output and achieve a more stable power output. The adjustable capacity is defined as the adjustment range and adjustment speed of each power plant during the process of its actual output curve approaching the planned output curve.

[0028] In terms of power supply capacity prediction, this invention proposes a comprehensive prediction method that combines the sliding window method, EMD, KPCA, and LSTM time series prediction models. This method can effectively handle nonlinear and non-stationary data and capture the long-term and short-term dependencies in time series data. For the prediction of power supply capacity of wind and solar renewable energy power plants in the next 1-15 days, the proposed method outperforms traditional prediction algorithms in terms of prediction accuracy, robustness, and real-time adaptability. In summary, this invention effectively addresses the problems of incomplete power supply capacity assessment indicators for wind and solar power plants and the immaturity of short- and medium-term power supply capacity prediction methods.

[0029] This invention can effectively overcome the limitations of traditional power output concept-based assessment methods for the power supply capacity of new energy power plants. It can more comprehensively reflect the actual power supply capacity of the power generation side of the new power system. The proposed comprehensive prediction method can better handle nonlinear and non-stationary data and capture the long-term and short-term dependencies in time series data, thus having certain advantages in predicting the power supply capacity of new energy. Attached Figure Description

[0030] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0031] Figure 1 This is an improved power supply capability assessment system for wind and solar power plants provided in this embodiment of the invention; Figure 2 This is an improved wind and solar power station-level power supply capacity prediction process provided in the embodiments of the present invention; Figure 3 This is the EMD decomposition process provided in the embodiments of the present invention; Figure 4 This is a schematic diagram of the structure of the electronic device provided in an embodiment of the present invention. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0033] This embodiment addresses the shortcomings of current wind and solar power plant-level power supply capacity assessment indicators and the immaturity of short- and medium-term power supply capacity prediction methods. It constructs a new assessment system to comprehensively reflect the power supply capacity of the generation side of the new power system. Drawing on methods for predicting renewable energy output, it innovatively proposes an LSTM neural network prediction algorithm for EMD-KPCA, improved using the sliding window method.

[0034] This embodiment adopts the following technical solution for the improved EMD-KPCA-LSTM wind and solar power station power supply capacity assessment and short-to-medium term forecasting using a sliding window, including the following steps: S1: Establish a 1-15 day power supply capacity assessment index system for wind power / photovoltaic power plants; S2: Establish a method for predicting the power supply capacity of wind power / photovoltaic power plants for 1-15 days.

[0035] The specific steps of step S1 are as follows: S1.1 New Energy Power Supply Capacity Indicators Based on Reliability Assessment: Indicator system such as Figure 1 As shown, the values ​​of various indicators for each power station and the collection station before and after adding energy storage are obtained according to the proposed calculation method for the reliability indicators of new energy power stations. There are both very large and very small indicators; to comprehensively assess reliability, the data needs to be standardized. The power generation rate indicator is a very large indicator, ranging from 0 to 1, and can be used directly. The power output interruption rate indicator is a very small indicator, ranging from 0 to 1, and is maximized. The volatility indicator is a very small indicator, representing the average power change per minute, and is normalized and maximized. The average normal power supply time is a very large indicator, ranging from 0 to 1, and can be used directly. The average normal power supply is a very large indicator and needs to be normalized by dividing by the average value.

[0036] The subjective weighting method within the Analytic Hierarchy Process (AHP) was used to determine the weights of each indicator. After comprehensive consideration and expert advice, the judgment matrix is ​​as follows: E GR \ E PIP \ E PF \ E ANFST \ E ANES These represent the power generation rate, power output interruption rate, volatility, average normal power supply time, and average normal power supply volume, respectively.

[0037]

[0038] The weight vectors are calculated using both the arithmetic mean method and the eigenvalue method, and then the average weights are obtained by combining the two methods. The final comprehensive index result is the product of the comprehensive weight vector and each corresponding index, yielding the percentage improvement in reliability for each site and collection station before and after energy storage.

[0039] S1.2 Adjustability index based on power curtailment: The values ​​of various indicators for each site and collection station before and after adding energy storage were obtained according to the proposed method for calculating adjustable capacity indicators. These indicators include both very large and very small ones; therefore, to comprehensively assess reliability, the data needs to be standardized. E MAE and E RMSE Since it is a very small indicator and falls between 0 and 1, we use 1 for subtraction. E MTE According to 1-(maximum tracking error / P N ) to handle.

[0040] The subjective weighting method in the AHP approach was used to determine the weights of each indicator. After comprehensive consideration and expert advice, the judgment matrix is ​​as follows. E MAE \ E RMSE \ E MTE \ E ASR These represent the mean absolute error, root mean square error, maximum tracking error, and power curtailment rate, respectively.

[0041]

[0042] The weight vectors are calculated using both the arithmetic mean method and the eigenvalue method, and then the average of the two methods is obtained. The final comprehensive index result is the product of the comprehensive weight vector and each corresponding index, yielding the percentage increase in the adjustable capacity of each site and collection station before and after energy storage.

[0043] The importance of overall reliability and adjustability is still determined by the AHP method. After comprehensive consideration and expert advice, the judgment weight matrix is ​​as follows.

[0044]

[0045] The weight vectors are calculated using both the arithmetic mean method and the eigenvalue method, and then the weights are averaged by combining the two methods. The final calculated power supply capacity improvement for each power station and collection station before and after energy storage configuration allows us to determine the percentage increase in power supply capacity after energy storage is installed.

[0046] The specific steps of step S2 are as follows: S2: Establish a method for predicting the power supply capacity of wind power / photovoltaic power plants for 1-15 days.

[0047] Prediction process as follows Figure 2 As shown, the power supply capacity of a wind and solar power station is evaluated by selecting 15 consecutive days of operational data. The evaluation results and values ​​of each primary and secondary index are obtained. The 12-dimensional characteristic parameters of the obtained power supply capacity evaluation value are used as input to predict the power supply capacity value for the next 15 days.

[0048] S2.1: EMD Decomposition EMD decomposition process as follows: Figure 3 As shown, EMD (Empirical Mode Decomposition) is performed on the input 12-dimensional raw data to obtain the decomposition results under multiple features. The specific decomposition process is as follows: (1) Find all the maximum points of the original data sequence X(t) and fit the upper envelope of the original data using the cubic spline interpolation function.

[0049] (2) Find all the local minimum points and fit all the local minimum points to form the lower envelope of the data by using a cubic spline interpolation function.

[0050] (3) The mean of the upper and lower envelopes is denoted as m(t). Subtracting the mean envelope m(t) from the original data sequence X(t) yields a new data sequence h(t).

[0051] (4) If there are still negative local maxima and positive local minima after subtracting the envelope average from the original data, it means that this is not an intrinsic modulus function and needs to be further "filtered".

[0052] S2.2: KPCA Dimensionality Reduction KPCA was used to reduce the dimensionality of the IMF components obtained from EMD decomposition and extract the main features. The principal components were then selected based on their high contribution rates. The specific algorithm steps are as follows: (1) Calculate the kernel function: First, we need to calculate the kernel matrix K of the original dataset.

[0053] (2) Centralized kernel matrix: The kernel matrix K is centered by column or row to obtain the centralized kernel matrix Kc.

[0054] (3) Solve for eigenvalues ​​and eigenvectors: Perform eigenvalue decomposition on the centered kernel matrix K to solve for its eigenvalues ​​and eigenvectors.

[0055] (4) Select principal components: Select the eigenvectors corresponding to the top k largest eigenvalues ​​as principal components to construct a new dimensionality reduction space.

[0056] (5) Data projection: Project the original dataset X onto a new dimension-reduced space to obtain the dimension-reduced data matrix Y.

[0057] (6) Inverse mapping: The dimensionality-reduced data can be restored to the original space through inverse transformation, so as to perform tasks such as classification and visualization.

[0058] S2.3: LSTM Neural Network Prediction A sliding window method is introduced to construct a time series dataset. By building features using continuous points within a certain time step, the model can capture the dynamic changes in the training data over time, improving the utilization of the dataset. Secondly, by using data points within a weighted window, the impact of random noise on the prediction results is reduced, making the final prediction model more robust. The training results of the model with all step sizes ranging from 3 to 15 were tested, with the most ideal prediction accuracy achieved when the program step size was set to 10.

[0059] For the formed feature set, an LSTM model is constructed, containing two layers of LSTM units, each with 150 neurons. The bidirectional structure allows the model to learn in both directions of the time series, improving prediction accuracy. An early stopping callback function is introduced to prevent overfitting. After training, the model is used to predict power supply capacity for the next 15 days. The algorithm flow is as follows: (1) Data preprocessing: normalization processing to make the data distributed between [0,1], which is beneficial for model training.

[0060] (2) Divide the training set and the test set: The training set is used to train the model, and the test set is used to test the accuracy of the model.

[0061] (3) Define the LSTM model: Define the input layer, output layer and intermediate layer of LSTM as needed.

[0062] (4) Training model: Specify the training batch size, training epochs, and validation set.

[0063] (5) Prediction: Use the trained model to make predictions.

[0064] Figure 4 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 4 As shown, the electronic device may include a processor, a communications interface, a memory, and a communication bus, wherein the processor, communications interface, and memory communicate with each other via the communication bus. The processor can call logical instructions in the memory to execute an improved method for assessing and predicting the power supply capacity of wind and solar power plants in the short to medium term.

[0065] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, and can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0066] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the improved wind and solar power station-level power supply capacity assessment and short-to-medium term prediction methods provided by the methods described above.

[0067] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0068] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0069] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An improved wind and solar plant level supply capability assessment and short to medium term forecasting method, characterized in that, Comprise: S1. Establish 1-15 days wind power / photovoltaic power station level power supply capacity evaluation index system; S2. Establish 1-15 days wind power / photovoltaic power station level power supply capacity prediction method.

2. The improved wind power plant level supply capability assessment and short-term forecasting method according to claim 1, characterized in that, Establish 1-15 days wind power / photovoltaic power station level power supply capacity evaluation index system includes: S1.

1. New energy power supply capacity index based on reliability evaluation: According to the power generation rate index GR, the power output interruption rate index PIP, the fluctuation rate index PF, the average normal power supply time index ANPST and the average normal power supply energy index ANES, the output characteristics of new energy are quantitatively analyzed; S1.

2. Adjustable capacity index: The error parameters between the plan and the actual in the process of the actual output curve of each station close to the planned output curve and the power loss parameter are taken as the adjustable capacity index.

3. The improved wind power plant level supply capability assessment and short-term forecasting method according to claim 2, characterized in that, The meaning and calculation method of the new energy power supply capacity index based on reliability evaluation includes: S1.1.

1. Power generation rate index: In the formula: p t ) is t the output value of new energy at the moment; P N the rated output power of new energy; T the total time of new energy operation; the closer the index value is to 1, the closer the overall power generation is to the rated power full-load state;​ S1.1.

2. Power output interruption rate index: In the formula: T is the total time of new energy operation, is the set of time when the new energy output is 0 within the operation time limit; the index describes the discontinuity of the new energy output, and the closer to 0 indicates that the zero output condition is less; S1.1.

3. Fluctuation rate index: In the formula: p ( t )for t The power output of new energy sources at all times; p ( t +1) is t The output power of the new energy source at time +1; this indicator is calculated by the difference between the previous and next time points to obtain the overall output power. T The fluctuation of renewable energy output over a period of time can effectively reflect the fluctuation of renewable energy output power; the unit is MW / min; the smaller the value, the more stable the renewable energy output power. S1.1.

4. Average normal power supply time index: In the formula: is the actual value reached by the new energy alone or in cooperation with energy storage to meet the actual value to reach the predicted value, without power rejection or power shortage period; T is the simulation time limit; the closer the index value is to 1, the fewer the power rejection or power shortage periods. S1.1.

5. Average normal power supply energy index: In the formula: T is the simulation time limit; N is the number of times of normal power supply of new energy within the simulation time limit; E j is the first j power supply power when the new energy is normally powered; the unit of the index is MWh / min, indicating the average power supply per minute of normal power supply.

4. The improved wind power plant level supply capability assessment and short-term forecasting method of claim 2, wherein, The calculation of the adjustable capacity index includes: S1.2.

1. Error parameter calculation between plan and actual; Mean absolute error MAE, root mean square error RMSE and maximum tracking error MTE: The calculation method is as follows: wherein: m is the number of sampling points in the simulation period; is the predicted value for the i time instant; is the actual value for the i time instant; is the rated power; the maximum tracking error is the maximum difference between the measured and predicted values. S1.2.

2. Calculation of power loss parameter; Adopt the abandoned and short power rate ASR to represent the abandoned and short power situation, and the calculation formula is as follows: In the formula: is the rejected electric power; is the actual delivered power; m is the number of sampling points in the simulation period; the smaller the index value, the fewer the power rejection and power shortage.

5. The improved wind power plant level supply capability assessment and short-term forecasting method of claim 1, wherein, The establishment of 1-15 days wind power / photovoltaic power station level power supply capacity prediction method includes the following steps: S2.

1. EMD decomposition; The complex signal is decomposed into a limited number of intrinsic mode functions IMF, and each IMF component contains the local characteristic signal of different time scales of the original signal; EMD decomposition method is based on the following assumptions: (1) The data has at least two extreme values, one maximum value and one minimum value; (2) The local time domain characteristics of the data are determined by the time scale between the extreme points; (3) If the data has no extreme points but has inflection points, the extreme values are obtained by differentiating the data once or more times, and then the decomposition result is obtained by integration; S2.

2. Dimension reduction by KPCA algorithm; The KPCA algorithm is used to map the original data to a high-dimensional feature space, so that the linear separability of the data in the space is stronger, and then PCA is used for dimension reduction processing of the data; The mathematical expression of the radial basis function RBF is as follows: wherein, x i and x j denotes an input sample, is a parameter controlling the shape of the kernel function; S2.

3. LSTM neural network prediction; The LSTM neuron transmits in the following way: the recurrent unit has 3 inputs, including the hidden state of the previous time step and the cell state and the input data of the current time step ; The gating units of the LSTM include an input gate, a forget gate and an output gate, which jointly determine the update of the cell state and the final hidden state output. The gating units generate values between 0 and 1 through a sigmoid function, which respectively represent the degree of acceptance of new information, the degree of forgetting of old information and the degree of exposure of the cell state to the output. The forget gate allows the model to selectively forget the information of the past cell state according to the current input. The closer the result is to 0, the greater the forgetting probability, and the closer the result is to 1, the smaller the forgetting probability. The input gate is responsible for screening important information in the current input and integrating it into the new candidate state.

6. The improved wind power plant level supply capability assessment and short-term forecasting method according to claim 5, characterized in that, The EMD decomposition process is as follows: S2.1.

1. Find all the maximum points of the original data sequence X(t), and use a cubic spline interpolation function to fit to form the upper envelope line of the original data; S2.1.

2. Find all the minimum points, and fit all the minimum points through a cubic spline interpolation function to form the lower envelope line of the data; S2.1.

3. The mean of the upper envelope line and the lower envelope line is denoted as m(t), and the original data sequence X(t) is subtracted by the average envelope m(t) to obtain a new data sequence h(t); S2.1.

4. If there are still negative local maxima and positive local minima in the new data obtained by subtracting the envelope average from the original data, continue to screen.

7. The improved wind power plant level supply capability assessment and short-term forecasting method of claim 5, wherein, The steps of the KPCA algorithm are as follows: S2.2.

1. Computing the kernel function: Compute the kernel matrix of the original dataset K ; S2.2.

2. Centered kernel matrix: Center the kernel matrix K Centering by column or by row, resulting in a centered kernel matrix K c; S2.2.

3. Solving Eigenvalues and Eigenvectors: for the centered kernel matrix K performing eigen decomposition to solve its eigenvalues and eigenvectors; S2.2.

4. Select principal components: select the eigenvectors corresponding to the first k largest eigenvalues as principal components to construct a new reduced dimension space; S2.2.

5. Data projection: project the original data set X into the new reduced dimension space to obtain the reduced dimension data matrix Y; S2.2.

6. Back mapping: restore the reduced dimension data to the original space through inverse transformation, so as to perform classification and visualization tasks.

8. The improved wind power plant level supply capability assessment and short-term forecasting method of claim 5, wherein, The prediction steps of the LSTM neural network are as follows: S2.3.

1. Data preprocessing: normalization processing, so that the data is distributed between [0, 1], which is beneficial to model training; S2.3.

2. Divide the training set and the test set: the training set is used to train the model, and the test set is used to test the model accuracy; S2.3.

3. Define the LSTM model: define the input layer, output layer and intermediate layer of the LSTM as needed; S2.3.

4. Train the model: specify the training batch size, training rounds epochs and validation set; S2.3.

5. Prediction: use the trained model for prediction.

9. An electronic device comprising a memory, a processor, and a computer program stored on the memory, wherein the computer program comprises instructions that, when executed by the processor, cause the electronic device to perform the method of any one of claims 1-8. The processor executes the computer program to realize the steps of the method of any one of claims 1 to 8.

10. A computer readable storage medium having stored thereon computer programs / instructions, characterized in that, The computer program / instructions are executed by the processor to realize the steps of the method of any one of claims 1 to 8.

Images