Unit load setting bias method and system based on spot price clearing mode

Abstract

The invention provides a unit load setting bias method and system based on a spot price clearing mode, and belongs to the technical field of power control, and the method comprises the steps: processing obtained power market data and unit load associated data, and determining a second feature vector of each group of periodic data and the vector importance of each second feature vector; determining a corresponding feature vector queue based on the vector importance of each second feature vector of each group of periodic data; determining a prediction adjustment model based on the feature vector queue of each group of periodic data, and calculating a spot price trend value; and determining an optimal load value based on the spot price trend value and the prediction adjustment model, and adjusting the operation load of the single generator set based on the optimal load value. The operation load can be flexibly and accurately adjusted, the operation efficiency and economic benefits of the unit are improved, the market response capability is enhanced, the cost and energy utilization are optimized, the market risk is reduced, and the economy and reliability of power supply are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of power control, and in particular relates to a method and system for setting a load bias of a unit based on a spot clearing price mode. Background Art

[0002] The unit load setting bias method is a control strategy used to improve the dynamic response performance of thermal power unit load. This method mainly uses the dynamic characteristics of the main steam pressure deviation and the requirements of the load command change direction to determine the superposition of the turbine command dynamic bias through the dynamic bias model. This can solve the problem of slow load response and deviation between the unit load and the load command during dynamic tracking.

[0003] At present, since the production, circulation and consumption of electricity commodities are almost completed instantly and cannot be stored in large quantities, the supply and demand of electricity must be balanced in real time. In order to ensure the safety of power supply, the dispatching agency needs to accurately control it. Not only does the market need to have a good overall regulation mechanism, but the power plant (seller) needs to be able to adjust the operating status according to the transaction situation and adjust the unit load based on the spot clearing price.

[0004] Therefore, the present invention provides a method and system for setting a load bias of a unit based on a spot clearing price mode. Summary of the invention

[0005] The present invention provides a method and system for setting a bias for unit load under a spot-clearing price mode, which is used to determine a characteristic vector queue for each group of periodic data by processing power market data and unit load-related data, determine a prediction adjustment model based on all characteristic vector queues, and calculate a spot price trend value, determine an optimal load value based on the spot price trend value and the prediction adjustment model, and adjust the operating load of a single generator set. The operating load can be adjusted flexibly and accurately, the operating efficiency and economic benefits of the unit can be improved, the market responsiveness can be enhanced, the cost and energy utilization can be optimized, the market risk can be reduced, and the economy and reliability of power supply can be improved.

[0006] The present invention provides a method for setting a load bias of a unit based on a spot clearing price mode, comprising:

[0007] 101: Process the acquired power market data and unit load associated data to determine the second eigenvector of each group of periodic data and the vector importance of each second eigenvector;

[0008] 102: Determine a corresponding feature vector queue based on the vector importance of each second feature vector of each group of periodic data;

[0009] 103: Determine the forecast adjustment model based on the feature vector queue of each set of periodic data, and calculate the spot price trend value;

[0010] 104: Determine the optimal load value based on the spot price trend value and the forecast adjustment model, and adjust the operating load of the single generator set based on the optimal load value.

[0011] According to the method for setting a bias for unit load based on a spot clearing price mode provided by the present invention, the acquired power market data and unit load related data are processed, including:

[0012] Divide the power market data and unit load-related data into periods, and determine multiple groups of periodic data in periodic intervals, wherein the periodic interval is a single day, a single week, a single month, or a single quarter;

[0013] A first feature vector is determined for each input value in the periodic data, wherein the first feature vector includes a plurality of features.

[0014] According to the unit load setting bias method based on the spot clearing price mode provided by the present invention, the second eigenvector of each group of periodic data and the vector importance of each second eigenvector are determined, including:

[0015] Extracting each group of periodic data respectively, and calculating a corresponding second eigenvector based on the first eigenvector of each input value in the periodic data;

[0016]

[0017] Among them, x2 ij The second eigenvector representing the jth input value in the i-th set of periodic data, x1 ij represents the first eigenvector of the jth input value in the i-th group of periodic data, iN1 represents the number of input values ​​in the i-th group of periodic data, α represents the weight penalty factor during dimensionality reduction, and β represents the correction parameter;

[0018] Performing importance transformation on the second eigenvector of each input value in the periodic data, and calculating the vector importance of each second eigenvector;

[0019]

[0020] in, represents the vector importance of the second eigenvector of the j-th input value in the i-th set of periodic data, and λ represents the coupling coefficient.

[0021] According to the method for setting a load bias of a unit based on a spot clearing price mode provided by the present invention, a corresponding feature vector queue is determined based on the vector importance of each second feature vector of each group of periodic data, including:

[0022] Determine whether the vector importance of the second feature vector of each input value in each set of periodic data is within a preset feature threshold range;

[0023] If the vector importance of the second feature vector is not within the preset feature threshold interval, the second feature vector is determined to be an abnormal feature vector of periodic data; if the vector importance of the second feature vector is within the preset feature threshold interval, the second feature vector is determined to be a normal feature vector of periodic data;

[0024] Determine a normal set of corresponding periodic data based on all normal feature vectors of the periodic data;

[0025] Determine a third eigenvector corresponding to the periodic data based on the features appearing in all normal eigenvectors in the periodic data;

[0026] Determine a feature vector queue in the corresponding periodic data based on the third feature vector of the periodic data and the vector importance of each second feature vector;

[0027]

[0028] Among them, f i1 、f ik 、f iiN2 Respectively represent the features 1, k, iN2, and fI in the third eigenvector of the i-th group of periodic data i1 ,fI ik ,fI iiN2 They represent the feature importance of feature 1, feature k, and feature iN2 in the third feature vector of the i-th group of periodic data, iN2 represents the number of features in the third feature vector of the i-th group of periodic data, iN3 represents the number of features in the second feature vector of the i-th group of periodic data, iNn represents the number of normal feature vectors of the i-th group of periodic data, δ ij represents the adjustment parameter of the second eigenvector of the jth input value in the i-th group of periodic data, iN1-iNn represents the number of abnormal eigenvectors of the i-th group of periodic data, A i 5. The method for setting the load bias of a unit based on the spot clearing price mode according to claim 1, characterized in that the prediction adjustment model is determined based on the feature vector queue of each set of periodic data, comprising:

[0029] Selecting a features and corresponding feature importances from the feature vector queue of each group of periodic data as the initial clustering center of the periodic data;

[0030] Determine the prediction adjustment model parameters and construct an initial adjustment model, input the feature vector queue of each group of periodic data into the initial adjustment model respectively, and determine the cluster set of the periodic data based on the initial cluster center of the periodic data and the feature importance of each feature in the feature vector queue;

[0031] Based on the feature importance of each feature in the feature vector queue of the periodic data and the cluster set, the initial cluster center is updated, and the initial adjustment model is iteratively trained to determine the periodic prediction adjustment model of the periodic data;

[0032] A periodic forecast adjustment model based on all periodic data determines a forecast adjustment model.

[0033] According to the method for setting a load bias of a unit based on a spot clearing price mode provided by the present invention, the spot price trend value is calculated, including:

[0034] Input the feature vector queue of each set of periodic data into the prediction adjustment model to determine the output result corresponding to each set of periodic data;

[0035] Determine the relative error of the prediction adjustment model based on the feature vector queue of each set of periodic data and the corresponding output results;

[0036] Determine the spot price trend value based on the relative error of the forecast adjustment model, the corresponding output results in each set of periodic data, and the feature vector queue;

[0037]

[0038] Among them, PT represents the spot price trend value, θ ia represents the intra-class aggregation degree of the a-th cluster set of the i-th group of periodic data, σ represents the Gaussian surround scale, CS ia represents the ath cluster set of the i-th group of periodic data, V i represents the output result of the i-th group of periodic data, iN4 represents the number of cluster sets of the i-th group of periodic data, represents the relative error of the forecast adjustment model, P ib represents the spot price corresponding to the bth normal eigenvector in the i-th group of periodic data, and iNn represents the number of normal eigenvectors in the i-th group of periodic data.

[0039] According to the method for setting a load bias of a generator set in a spot clearing price mode provided by the present invention, an optimal load value is determined based on a spot price trend value and a prediction adjustment model, and the operating load of a single generator set is adjusted based on the optimal load value, including:

[0040] Determine a load pattern label of a single generator set based on the spot price trend value, and input the spot price trend value and the load pattern label into a prediction adjustment model;

[0041] The forecast adjustment model predicts the load peak and predicted transaction value of a single generator set, and generates a spot price trend curve and a spot transaction curve;

[0042] The optimal load value is determined based on the spot price trend curve and the spot trading curve, and the optimal load value is fed back to the forecast adjustment model. The forecast adjustment model adjusts the operating load of a single generator set based on the optimal load value.

[0043] The system for setting a load bias of a unit based on a spot clearing price mode provided by the present invention comprises:

[0044] Processing module: Processing the acquired power market data and unit load related data to determine the second eigenvector of each set of periodic data and the vector importance of each second eigenvector;

[0045] Determining module: determining a corresponding feature vector queue based on the vector importance of each second feature vector of each group of periodic data;

[0046] Calculation module: Determines the forecast adjustment model based on the feature vector queue of each set of periodic data and calculates the spot price trend value;

[0047] Adjustment module: Determine the optimal load value based on the spot price trend value and the forecast adjustment model, and adjust the operating load of a single generator set based on the optimal load value.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] By processing the power market data and the unit load-related data, the characteristic vector queue of each group of periodic data is determined, and the forecast adjustment model is determined according to all the characteristic vector queues, and the spot price trend value is calculated. The optimal load value is determined according to the spot price trend value and the forecast adjustment model, and the operating load of a single generator set is adjusted. The operating load can be adjusted flexibly and accurately, the unit operating efficiency and economic benefits can be improved, the market responsiveness can be enhanced, the cost and energy utilization can be optimized, the market risk can be reduced, and the economy and reliability of power supply can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0051] Figure 1 It is a flow chart of a method for setting a load bias of a unit based on a spot clearing price mode provided by an embodiment of the present invention;

[0052] Figure 2 It is a structural schematic diagram of a unit load setting bias system based on a spot clearing price mode provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] Embodiment 1:

[0055] The embodiment of the present invention provides a method for setting a load bias of a unit based on a spot clearing price mode, such as Figure 1 As shown, including:

[0056] 101: Process the acquired power market data and unit load associated data to determine the second eigenvector of each group of periodic data and the vector importance of each second eigenvector;

[0057] 102: Determine a corresponding feature vector queue based on the vector importance of each second feature vector of each group of periodic data;

[0058] 103: Determine the forecast adjustment model based on the feature vector queue of each set of periodic data, and calculate the spot price trend value;

[0059] 104: Determine the optimal load value based on the spot price trend value and the forecast adjustment model, and adjust the operating load of the single generator set based on the optimal load value.

[0060] In this embodiment, the electricity market data includes but is not limited to spot prices, historical prices, listing information, transaction information, and transaction constraint data.

[0061] In this embodiment, the unit load-related data includes but is not limited to the actual power, vibration index, angle, temperature and torque of the generator set.

[0062] In this embodiment, the second eigenvector represents a vector obtained by performing dimensionality reduction processing on the first eigenvector of the input value of the corresponding periodic data.

[0063] In this embodiment, the vector importance indicates the importance of the corresponding second eigenvector in the periodic data.

[0064] In this embodiment, the feature vector array represents all features included in the corresponding periodic data and the feature importances corresponding to the features.

[0065] In this embodiment, the spot price trend indicates the possible changes in future market prices.

[0066] In this embodiment, the optimal load value refers to the power generation load setting value that can optimize the operation of the power system and the economic benefits under the given power market data and unit load related data conditions.

[0067] The beneficial effects of the above technical solution are as follows: by processing electricity market data and unit load-related data, the characteristic vector queue of each group of periodic data is determined, the prediction adjustment model is determined according to all characteristic vector queues, and the spot price trend value is calculated. The optimal load value is determined according to the spot price trend value and the prediction adjustment model, and the operating load of a single generator set is adjusted. The operating load can be adjusted flexibly and accurately, the unit operating efficiency and economic benefits can be improved, the market responsiveness can be enhanced, the cost and energy utilization can be optimized, the market risk can be reduced, and the economy and reliability of power supply can be improved.

[0068] Embodiment 2:

[0069] The embodiment of the present invention provides a method for setting a bias for a unit load under a spot clearing price mode, and processes the acquired power market data and unit load related data, including:

[0070] Divide the power market data and unit load-related data into periods, and determine multiple groups of periodic data in periodic intervals, wherein the periodic interval is a single day, a single week, a single month, or a single quarter;

[0071] A first feature vector is determined for each input value in the periodic data, wherein the first feature vector includes a plurality of features.

[0072] In this embodiment, the features included in the first feature vector may be spot price, historical price, listing information, transaction information, transaction constraint data, actual power of the generator set, vibration index, angle, temperature and torque.

[0073] In this embodiment, each input value in the periodic data corresponds to a first eigenvector.

[0074] In this embodiment, the first eigenvector represents the power market data and the unit load related data of the input value in the corresponding periodic data.

[0075] The beneficial effects of the above technical solution are as follows: by dividing the power market and unit load related data into periods and determining the first eigenvector of each input value in each set of periodic data, the power market and unit load related data can be preliminarily processed to ensure data consistency, reduce the system data processing load, and improve unit operation efficiency.

[0076] Embodiment 3:

[0077] The embodiment of the present invention provides a method for setting a load bias of a unit based on a spot clearing price mode, which determines a second eigenvector of each set of periodic data and a vector importance of each second eigenvector, including:

[0078] Extracting each group of periodic data respectively, and calculating a corresponding second eigenvector based on the first eigenvector of each input value in the periodic data;

[0079]

[0080] Among them, x2 ij The second eigenvector representing the jth input value in the i-th set of periodic data, x1 ij represents the first eigenvector of the jth input value in the i-th group of periodic data, iN1 represents the number of input values ​​in the i-th group of periodic data, α represents the weight penalty factor during dimensionality reduction, and β represents the correction parameter;

[0081] Performing importance transformation on the second eigenvector of each input value in the periodic data, and calculating the vector importance of each second eigenvector;

[0082]

[0083] in, represents the vector importance of the second eigenvector of the j-th input value in the i-th set of periodic data, and λ represents the coupling coefficient.

[0084] In this embodiment, each input value in the periodic data corresponds to a second eigenvector.

[0085] In this embodiment, the number of features included in the second feature vector of each input value is smaller than the number of features included in the corresponding first feature vector.

[0086] In this embodiment, Represents the correction value of the jth input value in the i-th set of periodic data.

[0087] In this embodiment, each second eigenvector in the periodic data corresponds to a vector importance.

[0088] The beneficial effects of the above technical solution are as follows: by determining the second eigenvector of each set of periodic data and the vector importance of each second eigenvector, the power market and unit load related data can be comprehensively analyzed to improve the modeling efficiency of the prediction and adjustment model and further improve the unit operation efficiency.

[0089] Embodiment 4:

[0090] The embodiment of the present invention provides a method for setting a load bias of a unit based on a spot clearing price mode, and determines a corresponding feature vector queue based on the vector importance of each second feature vector of each group of periodic data, including:

[0091] Determine whether the vector importance of the second feature vector of each input value in each set of periodic data is within a preset feature threshold range;

[0092] If the vector importance of the second feature vector is not within the preset feature threshold interval, the second feature vector is determined to be an abnormal feature vector of periodic data; if the vector importance of the second feature vector is within the preset feature threshold interval, the second feature vector is determined to be a normal feature vector of periodic data;

[0093] Determine a normal set of corresponding periodic data based on all normal feature vectors of the periodic data;

[0094] Determine a third eigenvector corresponding to the periodic data based on the features appearing in all normal eigenvectors in the periodic data;

[0095] Determine a feature vector queue in the corresponding periodic data based on the third feature vector of the periodic data and the vector importance of each second feature vector;

[0096]

[0097] Among them, f i1 、f ik 、f iiN2 Respectively represent the features 1, k, iN2, and fI in the third eigenvector of the i-th group of periodic data i1 ,fI ik ,fI iiN2 They represent the feature importance of feature 1, feature k, and feature iN2 in the third feature vector of the i-th group of periodic data, iN2 represents the number of features in the third feature vector of the i-th group of periodic data, iN3 represents the number of features in the second feature vector of the i-th group of periodic data, iNn represents the number of normal feature vectors of the i-th group of periodic data, δ ij represents the adjustment parameter of the second eigenvector of the jth input value in the i-th group of periodic data, iN1-iNn represents the number of abnormal eigenvectors of the i-th group of periodic data, A i Represents the normal set of the i-th group of period data.

[0098] In this embodiment, the preset characteristic threshold interval represents the normal range interval of the second characteristic vector of the periodic data.

[0099] In this embodiment, the normal feature vector represents a second feature vector whose vector importance is within a preset feature threshold range.

[0100] In this embodiment, each periodic data corresponds to a normal set, and the normal set includes all second feature vectors in the corresponding periodic data that are within a preset feature threshold interval.

[0101] In this embodiment, the third feature vector represents the features contained in all the second feature vectors in the normal set of periodic data.

[0102] In this embodiment, each periodic data corresponds to a third eigenvector.

[0103] In this embodiment, the feature importance in the feature vector array represents the feature importance of the corresponding feature in the periodic data.

[0104] The beneficial effects of the above technical solution are as follows: by determining the corresponding eigenvector queue according to the vector importance of each second eigenvector of each group of periodic data, the load of system processing data can be reduced, data basis for determining the prediction adjustment model can be improved, energy utilization can be optimized, unit operation efficiency and economic benefits can be improved, and the economy and reliability of power supply can be improved.

[0105] Embodiment 5:

[0106] The embodiment of the present invention provides a method for setting a load bias of a unit under a spot clearing price mode, and determines a prediction adjustment model based on a feature vector queue of each group of periodic data, including:

[0107] Selecting a features and corresponding feature importances from the feature vector queue of each group of periodic data as the initial clustering center of the periodic data;

[0108] Determine the prediction adjustment model parameters and construct an initial adjustment model, input the feature vector queue of each group of periodic data into the initial adjustment model respectively, and determine the cluster set of the periodic data based on the initial cluster center of the periodic data and the feature importance of each feature in the feature vector queue;

[0109] Based on the feature importance of each feature in the feature vector queue of the periodic data and the cluster set, the initial cluster center is updated, and the initial adjustment model is iteratively trained to determine the periodic prediction adjustment model of the periodic data;

[0110] A periodic forecast adjustment model based on all periodic data determines a forecast adjustment model.

[0111] In this embodiment, a features are selected from the feature vector queue of each group of periodic data, and the value of a is determined according to the number of features in the third feature vector of the periodic data.

[0112] In this embodiment, the number of initial cluster centers is in the range of 2≤a≤K, the number of model iterations is 500-1000, and the iteration error ε=10 -5 .

[0113] In this embodiment, the initial adjustment model is iteratively trained, and one cycle prediction adjustment model is generated every A iterations.

[0114] In this embodiment, the comprehensive clustering results of the periodic prediction adjustment models after each A iteration training are evaluated based on the first evaluation index, and the periodic prediction adjustment model with the smallest comprehensive clustering result is selected as the periodic prediction adjustment model of the periodic data.

[0115] In this embodiment, the periodic prediction adjustment model of each group of periodic data is evaluated based on the second evaluation index to determine the prediction adjustment model.

[0116] In this embodiment, the initial adjustment model combines generalized regression network, support vector machine and deep neural network to perform load forecasting.

[0117] In this embodiment, the first evaluation index and the second evaluation index may be an XBI index and an FSI index. The smaller the XBI index and the FSI index are, the more convergent and accurate the prediction adjustment model is.

[0118] In this embodiment, the prediction adjustment model includes at least one group of fully connected layers, LSTM layers and predictors. The LSTM layers are connected to the fully connected layers and the predictors respectively, and the fully connected layers are used to perform dimensionality reduction processing on the feature vector queue with a 7-dimensional period. Each group of the fully connected layers includes two neurons, and the neurons are composed of a 7*1 matrix.

[0119] The beneficial effects of the above technical solution are as follows: by determining the prediction adjustment model based on the characteristic vector queue of each set of periodic data, the accuracy of the spot price trend curve, the spot trading curve and the optimal load value of a single generator set can be ensured, the operating load can be adjusted flexibly and accurately, the market responsiveness can be enhanced, and the market risk can be reduced.

[0120] Embodiment 6:

[0121] The embodiment of the present invention provides a method for setting a load bias of a unit based on a spot clearing price mode, and calculates a spot price trend value, including:

[0122] Input the feature vector queue of each set of periodic data into the prediction adjustment model to determine the output result corresponding to each set of periodic data;

[0123] Determine the relative error of the prediction adjustment model based on the feature vector queue of each set of periodic data and the corresponding output results;

[0124] Determine the spot price trend value based on the relative error of the forecast adjustment model, the corresponding output results in each set of periodic data, and the feature vector queue;

[0125]

[0126] Among them, PT represents the spot price trend value, θ ia represents the intra-class aggregation degree of the a-th cluster set of the i-th group of periodic data, σ represents the Gaussian surround scale, CS ia represents the ath cluster set of the i-th group of periodic data, V i represents the output result of the i-th group of periodic data, iN4 represents the number of cluster sets of the i-th group of periodic data, represents the relative error of the forecast adjustment model, P ib represents the spot price corresponding to the bth normal eigenvector in the i-th group of periodic data, and iNn represents the number of normal eigenvectors in the i-th group of periodic data.

[0127] In this embodiment, each set of periodic data corresponds to an output result.

[0128] In this embodiment, the intra-class aggregation degree represents a measure of the similarity between features in the same cluster set.

[0129] In this embodiment, the output result of the i-th group of periodic data represents the influence of all features in the feature vector queue of the corresponding periodic data on the price trend.

[0130] In this embodiment, the Gaussian surround scale represents the distribution of all features included in the third eigenvector of the periodic data.

[0131] In this embodiment, the relative error represents the prediction accuracy of the prediction adjustment model on each set of periodic data.

[0132] In this embodiment, the spot price trend value represents the expected change trend of future electricity market prices.

[0133] The beneficial effects of the above technical solution are as follows: by calculating the spot price trend value based on the characteristic vector queue of each set of periodic data and the predictive adjustment model, the accuracy of the optimal load value evaluation of a single generator set can be ensured, the operating load can be adjusted flexibly and quickly, the market responsiveness can be enhanced, the cost and energy utilization can be optimized, and the operating efficiency and economic benefits of the unit can be improved.

[0134] Embodiment 7:

[0135] The embodiment of the present invention provides a method for setting a load bias of a generator set under a spot clearing price mode, determining an optimal load value based on a spot price trend value and a prediction adjustment model, and adjusting the operating load of a single generator set based on the optimal load value, including:

[0136] Determine a load pattern label of a single generator set based on the spot price trend value, and input the spot price trend value and the load pattern label into a prediction adjustment model;

[0137] The forecast adjustment model predicts the load peak and predicted transaction value of a single generator set, and generates a spot price trend curve and a spot transaction curve;

[0138] The optimal load value is determined based on the spot price trend curve and the spot trading curve, and the optimal load value is fed back to the forecast adjustment model. The forecast adjustment model adjusts the operating load of a single generator set based on the optimal load value.

[0139] In this embodiment, the load mode label indicates the load behavior mode of a single generator set, which may be high load, low load or balanced load, etc. In this embodiment, the load peak value indicates the maximum load value of a single generator set predicted by the prediction adjustment model, reflecting the highest load level that the generator set may need to reach under the conditions of power market data and unit load-related data. In this embodiment, the predicted transaction value indicates the power transaction volume and transaction amount forecast that the generator set may carry out under the predicted market price trend.

[0140] In this embodiment, the spot price trend curve represents the predicted trend of future market prices over time.

[0141] In this embodiment, the spot transaction curve represents the predicted future market transaction volume and the temporal variation trend of the transaction amount.

[0142] The beneficial effects of the above technical solution are as follows: the optimal load value is determined according to the spot price trend value and the forecast adjustment model, and the operating load of a single generator set is adjusted based on the optimal load value. The operating load can be adjusted flexibly and accurately, the operating efficiency and economic benefits of the unit can be improved, the market responsiveness can be enhanced, the cost and energy utilization can be optimized, the market risk can be reduced, and the economy and reliability of power supply can be improved.

[0143] Embodiment 8:

[0144] An embodiment of the present invention provides a system for setting a load bias of a unit based on a spot clearing price mode, comprising:

[0145] Processing module: Processing the acquired power market data and unit load related data to determine the second eigenvector of each set of periodic data and the vector importance of each second eigenvector;

[0146] Determining module: determining a corresponding feature vector queue based on the vector importance of each second feature vector of each group of periodic data;

[0147] Calculation module: Determines the forecast adjustment model based on the feature vector queue of each set of periodic data and calculates the spot price trend value;

[0148] Adjustment module: Determine the optimal load value based on the spot price trend value and the forecast adjustment model, and adjust the operating load of a single generator set based on the optimal load value.

[0149] The beneficial effects of the above technical solution are as follows: by processing electricity market data and unit load-related data, the characteristic vector queue of each group of periodic data is determined, the prediction adjustment model is determined according to all characteristic vector queues, and the spot price trend value is calculated. The optimal load value is determined according to the spot price trend value and the prediction adjustment model, and the operating load of a single generator set is adjusted. The operating load can be adjusted flexibly and accurately, the unit operating efficiency and economic benefits can be improved, the market responsiveness can be enhanced, the cost and energy utilization can be optimized, the market risk can be reduced, and the economy and reliability of power supply can be improved.

Claims

1. A method for setting a load bias for a unit based on a spot clearing price mode, characterized in that: include: 101: Process the acquired power market data and unit load associated data to determine the second eigenvector of each group of periodic data and the vector importance of each second eigenvector; 102: Determine a corresponding feature vector queue based on the vector importance of each second feature vector of each group of periodic data; 103: Determine the forecast adjustment model based on the feature vector queue of each set of periodic data, and calculate the spot price trend value; 104: Determine the optimal load value based on the spot price trend value and the forecast adjustment model, and adjust the operating load of the single generator set based on the optimal load value.

2. The method for setting a load bias of a unit based on a spot clearing price mode according to claim 1, characterized in that: Process the acquired power market data and unit load related data, including: Divide the power market data and unit load-related data into periods, and determine multiple groups of periodic data in periodic intervals, wherein the periodic interval is a single day, a single week, a single month, or a single quarter; A first feature vector is determined for each input value in the periodic data, wherein the first feature vector includes a plurality of features.

3. The method for setting a load bias of a unit based on a spot clearing price mode according to claim 2, characterized in that: Determining the second eigenvector of each set of periodic data and the vector importance of each second eigenvector includes: Extracting each group of periodic data respectively, and calculating a corresponding second eigenvector based on the first eigenvector of each input value in the periodic data; Among them, x2 ij The second eigenvector representing the jth input value in the i-th set of periodic data, x1 ij represents the first eigenvector of the jth input value in the i-th group of periodic data, iN1 represents the number of input values ​​in the i-th group of periodic data, α represents the weight penalty factor during dimensionality reduction, and β represents the correction parameter; Performing importance transformation on the second eigenvector of each input value in the periodic data, and calculating the vector importance of each second eigenvector; in, represents the vector importance of the second eigenvector of the j-th input value in the i-th set of periodic data, and λ represents the coupling coefficient.

4. The method for setting a load bias of a unit based on a spot clearing price mode according to claim 3, characterized in that: Determining a corresponding feature vector queue based on the vector importance of each second feature vector of each group of periodic data includes: Determine whether the vector importance of the second feature vector of each input value in each set of periodic data is within a preset feature threshold range; If the vector importance of the second feature vector is not within the preset feature threshold interval, the second feature vector is determined to be an abnormal feature vector of periodic data; if the vector importance of the second feature vector is within the preset feature threshold interval, the second feature vector is determined to be a normal feature vector of periodic data; Determine a normal set of corresponding periodic data based on all normal feature vectors of the periodic data; Determine a third eigenvector corresponding to the periodic data based on the features appearing in all normal eigenvectors in the periodic data; Determine a feature vector queue in the corresponding periodic data based on the third feature vector of the periodic data and the vector importance of each second feature vector; Among them, f i1 、f ik 、f iiN2 Respectively represent the features 1, k, iN2, and fI in the third eigenvector of the i-th group of periodic data i1 ,fI ik ,fI iiN2 They represent the feature importance of feature 1, feature k, and feature iN2 in the third feature vector of the i-th group of periodic data, iN2 represents the number of features in the third feature vector of the i-th group of periodic data, iN3 represents the number of features in the second feature vector of the i-th group of periodic data, iNn represents the number of normal feature vectors of the i-th group of periodic data, δ ij represents the adjustment parameter of the second eigenvector of the jth input value in the i-th group of periodic data, iN1-iNn represents the number of abnormal eigenvectors of the i-th group of periodic data, A i Represents the normal set of the i-th group of period data.

5. The method for setting a load bias of a unit based on a spot clearing price mode according to claim 1, characterized in that: The prediction adjustment model is determined based on the feature vector queue of each set of periodic data, including: Selecting a features and corresponding feature importances from the feature vector queue of each group of periodic data as the initial clustering center of the periodic data; Determine the prediction adjustment model parameters and construct an initial adjustment model, input the feature vector queue of each group of periodic data into the initial adjustment model respectively, and determine the cluster set of the periodic data based on the initial cluster center of the periodic data and the feature importance of each feature in the feature vector queue; Based on the feature importance of each feature in the feature vector queue of the periodic data and the cluster set, the initial cluster center is updated, and the initial adjustment model is iteratively trained to determine the periodic prediction adjustment model of the periodic data; A periodic forecast adjustment model based on all periodic data determines a forecast adjustment model.

6. The method for setting a load bias of a unit based on a spot clearing price mode according to claim 4, characterized in that: Calculate spot price trend values, including: Input the feature vector queue of each set of periodic data into the prediction adjustment model to determine the output result corresponding to each set of periodic data; Determine the relative error of the prediction adjustment model based on the feature vector queue of each set of periodic data and the corresponding output results; Determine the spot price trend value based on the relative error of the forecast adjustment model, the corresponding output results in each set of periodic data, and the feature vector queue; Among them, PT represents the spot price trend value, θ ia represents the intra-class aggregation degree of the a-th cluster set of the i-th group of periodic data, σ represents the Gaussian surround scale, CS ia represents the ath cluster set of the i-th group of periodic data, V i represents the output result of the i-th group of periodic data, iN4 represents the number of cluster sets of the i-th group of periodic data, represents the relative error of the forecast adjustment model, P ib represents the spot price corresponding to the bth normal eigenvector in the i-th group of periodic data, and iNn represents the number of normal eigenvectors in the i-th group of periodic data.

7. The method for setting a load bias of a unit based on a spot clearing price mode according to claim 6, characterized in that: Determine the optimal load value based on the spot price trend value and the forecast adjustment model, and adjust the operating load of a single generator set based on the optimal load value, including: Determine a load pattern label of a single generator set based on the spot price trend value, and input the spot price trend value and the load pattern label into a prediction adjustment model; The forecast adjustment model predicts the load peak and predicted transaction value of a single generator set, and generates a spot price trend curve and a spot transaction curve; The optimal load value is determined based on the spot price trend curve and the spot trading curve, and the optimal load value is fed back to the forecast adjustment model. The forecast adjustment model adjusts the operating load of a single generator set based on the optimal load value.

8. Based on the unit load setting bias system under the spot clearing price mode, it is characterized by: include: Processing module: Processing the acquired power market data and unit load related data to determine the second eigenvector of each set of periodic data and the vector importance of each second eigenvector; Determining module: determining a corresponding feature vector queue based on the vector importance of each second feature vector of each group of periodic data; Calculation module: Determines the forecast adjustment model based on the feature vector queue of each set of periodic data and calculates the spot price trend value; Adjustment module: Determine the optimal load value based on the spot price trend value and the forecast adjustment model, and adjust the operating load of a single generator set based on the optimal load value.

Images