Multivariable generating capacity regression prediction method, system and device based on CPO-BiTCN-BiGRU and medium
Through the CPO-BiTCN-BiGRU model combined with the crown porcupine algorithm, the photovoltaic power generation prediction is optimized, which solves the problem of low medium- and long-term prediction accuracy, and achieves more accurate and reliable power generation prediction, which is suitable for grid planning and management of photovoltaic power stations.
Patent Information
- Application Number
- CN202411034097.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-30
- Publication Date
- 2024-11-29
AI Technical Summary
The existing photovoltaic power generation prediction methods have low prediction accuracy on medium and long-term scales and rely on limited historical data, making it difficult to accurately predict photovoltaic power generation, affecting power grid planning and scheduling management.
Using a multivariate power generation regression prediction method based on CPO-BiTCN-BiGRU, the time series data of meteorology, irradiance, wind speed, precipitation probability and dust/particle matter are obtained, and a two-way time domain convolution network and a two-way gating cycle unit are used. The initial model is constructed and combined with the optimization of the crown porcupine algorithm, a more accurate power generation prediction curve is generated.
It improves the accuracy and stability of power generation prediction, can better capture feature information and patterns in the time series, reduce manual parameter adjustment, and enhance the performance and reliability of the model.
Smart Images

Figure BDA0004970614090000041 
Figure BDA0004970614090000081 
Figure BDA0004970614090000101
Abstract
Description
Technical Field
[0001] The present invention belongs to new energy photovoltaic power generation technology, and specifically relates to a multivariate power generation regression prediction method, system, equipment and medium based on CPO-BiTCN-BiGRU. Background Art
[0002] my country's photovoltaic industry is developing rapidly, and the newly added photovoltaic power generation capacity continues to grow. The importance of photovoltaic power generation in the energy structure continues to be highlighted; however, the inherent intermittent fluctuation characteristics of photovoltaic power generation pose a huge challenge to the safe and stable operation of the power system. Compared with short-term predictions, medium- and long-term predictions face problems such as low accuracy of weather forecasts on a long-term scale and limited historical data samples. However, the accuracy of medium- and long-term predictions is crucial for grid planning optimization, dispatching management, and improving the absorption capacity.
[0003] At present, photovoltaic power generation prediction methods can be divided into physical methods, statistical methods and artificial intelligence methods according to their internal logic. Among them: physical methods do not need to rely on historical data, but require detailed and accurate monitoring of the meteorological environment around the photovoltaic power station, and are highly dependent on numerical weather forecasts; statistical methods are more suitable for short-term and ultra-short-term prediction scenarios; with the continuous advancement of artificial intelligence technology, photovoltaic prediction models based on machine learning are more competitive and can extract key features that affect solar radiation from complex and changeable high-dimensional data, but the prediction accuracy on medium and long-term scales still needs to be improved; at the same time, the basic data for prediction is too single. Summary of the invention
[0004] The purpose of the present invention is to provide a multivariate power generation regression prediction method, system, device and medium based on CPO-Bi TCN-Bi GRU, which solves the defect of low power generation prediction accuracy on medium and long-term scales in existing photovoltaic power generation prediction methods.
[0005] In order to achieve the above object, the technical solution adopted by the present invention is:
[0006] The present invention provides a multivariate power generation regression prediction method based on CPO-Bi TCN-Bi GRU, comprising the following steps:
[0007] Step 1, obtaining a time series matrix of the target electric field within a future set period, the time series matrix including a time subsequence of meteorological temperature prediction data, a time subsequence of irradiance prediction data, a time subsequence of wind speed prediction data, a time subsequence of precipitation probability prediction data, and a time subsequence of dust / particulate matter prediction data;
[0008] Step 2, using the obtained time series matrix as the input of the regression prediction model to obtain the power generation prediction curve within the future set period;
[0009] Wherein, the construction method of the regression prediction model is:
[0010] The initial model is constructed using a bidirectional time-domain convolutional network and a bidirectional gated recurrent unit;
[0011] The historical power generation data of the target electric field was combined with the crested porcupine algorithm to optimize the initial model and obtain a regression prediction model.
[0012] Preferably, in step 2, the obtained time series matrix is used as the input of the regression prediction model to obtain the power generation prediction curve within the future set period. The specific method is:
[0013] S21, setting the initial weight of each subsequence in the time series matrix;
[0014] S22, using the initial weights and combining the bidirectional time-domain convolutional network to perform multi-dimensional time-domain convolution processing on the obtained time series matrix to obtain a convolution operation result;
[0015] S23, combining the obtained convolution operation result with the bidirectional gated recurrent unit to perform a reverse convolution operation to obtain a reverse convolution operation result;
[0016] S24, cross-feature accumulation of the obtained convolution operation result and the inverse convolution operation result to obtain a time series used for the final prediction of power generation;
[0017] S25, using the obtained time series used for the final predicted power generation to calculate the power generation prediction curve within the future set time period, to obtain the power generation prediction curve.
[0018] Preferably, in S22, the initial weights are used in combination with a bidirectional time-domain convolutional network to perform multi-dimensional time-domain convolution processing on the obtained time series matrix to obtain a convolution operation result. The specific method is:
[0019] S221, performing a time domain flip operation on the time series matrix to obtain a time domain flip operation matrix;
[0020] S222, performing a first-layer multi-dimensional time-domain convolution calculation on the time series matrix and the time-domain flip operation matrix, respectively, to obtain a first-layer time-domain convolution calculation cumulative result corresponding to the time series matrix and a first-layer time-domain convolution calculation cumulative result corresponding to the time-domain flip operation matrix, respectively;
[0021] S223, perform k-1 layers of convolution operations on the accumulated results of the first layer of time domain convolution calculations corresponding to the time series matrix and the accumulated results of the first layer of time domain convolution calculations corresponding to the time domain flip operation matrix, respectively, to obtain the convolution operation results corresponding to the time series matrix and the convolution operation results corresponding to the time domain flip operation matrix.
[0022] Preferably, in S23, the obtained convolution operation result is combined with the bidirectional gated recurrent unit to perform a reverse convolution operation to obtain a reverse convolution operation result, and the specific method is:
[0023] The convolution operation results corresponding to the obtained time series matrix and the convolution operation results corresponding to the time domain flip operation matrix are respectively subjected to inverse convolution operation to obtain the inverse convolution operation results corresponding to the time series matrix and the inverse convolution operation results corresponding to the time domain flip operation matrix.
[0024] Preferably, the expression of the power generation prediction curve in the future set period is:
[0025]
[0026] Among them, p i is the i-th point in the power generation prediction curve within the future set time period; g, h1, h2, h3, and h4 are all model importance functions; is an operator, indicating a positive effect; is an operator, indicating the reverse effect; W l_pv_i , W l_ws_i , W l_temp_i , W l_rainp_i , W l_dust_i are all elements in the time series used for the final prediction of power generation; m is the total number of points in the power generation prediction curve within the future set time period.
[0027] Preferably, in step 3, the initial model is optimized using the historical power generation data of the target electric field in combination with the crested porcupine algorithm to obtain a regression prediction model, and the specific method is:
[0028] Obtain the historical actual power generation data of the target electric field and the time series matrix within the set time period, the time series matrix including the meteorological temperature prediction data time subsequence, the irradiance prediction data time subsequence, the wind speed prediction data time subsequence, the precipitation probability prediction data time subsequence, and the dust / particulate matter prediction data time subsequence;
[0029] The time series array is used as the input of the initial model to obtain the power generation forecast curve within the set time period;
[0030] The initial model is optimized based on the obtained power generation prediction curve and the historical actual power generation data within the set time period combined with the crested porcupine algorithm to obtain a regression prediction model.
[0031] A multivariable power generation regression prediction system based on CPO-BiTCN-BiGRU, comprising:
[0032] A data acquisition unit is used to acquire a time series matrix within a future set period of the target electric field, wherein the time series matrix includes a meteorological temperature prediction data time subsequence, an irradiance prediction data time subsequence, a wind speed prediction data time subsequence, a precipitation probability prediction data time subsequence, and a dust / particulate matter prediction data time subsequence;
[0033] A model building module is used to construct an initial model using a bidirectional time-domain convolutional network and a bidirectional gated recurrent unit; and to optimize the initial model using the historical power generation data of the target electric field combined with the crested porcupine algorithm to obtain a regression prediction model;
[0034] The power generation prediction module is used to use the obtained time series matrix as the input of the regression prediction model to obtain the power generation prediction curve within the future set time period.
[0035] A computer device comprising:
[0036] a processor suitable for executing a computer program;
[0037] A computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, the method described is performed.
[0038] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described above is implemented.
[0039] A computer program product comprises a computer program, wherein the computer program implements the method when executed by a processor.
[0040] Compared with the prior art, the present invention has the following beneficial effects:
[0041] The present invention provides a multivariate power generation regression prediction method based on CPO-Bi TCN-Bi GRU. The bidirectional time-domain convolutional network and the bidirectional gated recurrent unit are combined to construct an initial model, which is then optimized using the CP (Crown Porcupine Optimization Algorithm), so that the final regression prediction model has the following advantages:
[0042] 1) It can capture the characteristic information in the sequence more comprehensively and accurately;
[0043] 2) It can simultaneously consider the previous and next information of the sequence data, which helps to better understand and predict the patterns and trends in the sequence and improve the expressiveness of the model;
[0044] 3) It can improve the performance and stability of the model, making the time series prediction results more accurate and reliable.
[0045] 4) It can better process information of different time scales in time series;
[0046] 5) The initial model has many parameters. CPO is used to optimize the initial model to avoid the limitations and blindness of manual parameter adjustment, improve its prediction accuracy, and reduce manual parameter adjustment.
[0047] In summary, the present invention is applicable to the prediction of power generation of photovoltaic power stations and has broad application prospects and social benefits. BRIEF DESCRIPTION OF THE DRAWINGS
[0048] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION
[0049] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.
[0050] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, wholes, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or combinations thereof.
[0051] It should also be understood that the term “and / or” used in the specification and appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.
[0052] As used in the specification and appended claims of this application, the term "if" can be interpreted as "when" or "uponce" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "uponce it is determined" or "in response to determining" or "uponce [described condition or event] is detected" or "in response to detecting [described condition or event]", depending on the context.
[0053] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.
[0054] References to "one embodiment" or "some embodiments" etc. described in the specification of this application mean that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.
[0055] Example 1
[0056] The present embodiment provides a multivariate power generation regression prediction method based on CPO-BiTCN-BiGRU, comprising the following steps:
[0057] Step 1: Set up a regression prediction model, specifically the input layer L of the model combined with BiTCN (bidirectional temporal convolutional network) and BiGRU (bidirectional gated recurrent unit) in With the output layer L out ,in:
[0058] Input layer L in The weather forecast data matrix W for the target PV power station in the next 7 days;
[0059] Output layer L out The power generation prediction curve P of the target photovoltaic power station in the next 7 days;
[0060] The weather forecast data matrix W for the next 7 days is expressed as follows:
[0061]
[0062] Among them, time1…time m is a time tag with an interval of 15 minutes; temp1…temp m is the meteorological temperature forecast data corresponding to the time label; pv1…pv m is the irradiance prediction data corresponding to the time label; ws1…ws m is the wind speed forecast data corresponding to the time label; rainp1…rainp m is the precipitation probability prediction data corresponding to the time label; dust1…dust m Dust / particle prediction data corresponding to the time label;
[0063] It can be seen that except for the first column of time labels, the other five columns can be regarded as time series, which can be recorded as:
[0064] W=[W temp ,W pv ,W ws ,W rainp ,W dust ]
[0065] Among them, W temp ,W pv ,W ws ,W rainp ,W dust They are respectively the time series of meteorological temperature forecast data, irradiance forecast data, wind speed forecast data, precipitation probability forecast data, and dust / particle forecast data;
[0066] The power generation forecast curve P for the next 7 days is as follows:
[0067] P=[p1,p2,…,p i ,…,p m ]
[0068] Among them, p1, p2, …, p i ,…,p m They are the predicted power generation values at every 15-minute interval in the next 7 days;
[0069] Step 2: Set up a regression prediction model, specifically a BiTCN (bidirectional temporal convolutional network) and a BiGRU (bidirectional gated recurrent unit) combined with the model’s intermediate hidden layer L hide ;
[0070] The middle hidden layer L hide It is divided into k layers, each layer has a model parameter set [C1,…,C k ] and the activation function set [f1,…,f k ]; where C1,…,C k are the model parameters of each layer, f1,…,f k is the activation function for each layer of k hidden layers;
[0071] The activation function of each layer of the k hidden layers is the Relu function.
[0072] Step 3: After completing steps 1 and 2, set the specific execution process of the "BiTCN+BiGRU" algorithm as follows:
[0073] S31, set the initial weight vector λ of each column matrix in the input layer weather forecast data matrix W for the next 7 days, then:
[0074] λ=[λ temp ,λ pv ,λ ws ,λ rainp ,λdust ]
[0075] λ temp +λ pv +λ ws +λ rainp +λ dust =1
[0076] Among them, λ temp ,λ pv ,λ ws ,λ rainp ,λ dust They are the initial weights of meteorological temperature forecast data, irradiance forecast data, wind speed forecast data, precipitation probability forecast data, and dust / particle forecast data;
[0077] S32, performing a time domain flip operation on the input layer weather forecast data matrix W for the next 7 days to obtain a matrix W′; according to step 1, the matrix W′ after time domain flip can also be expressed as:
[0078] W′=[W′ temp ,W′ pv ,W′ ws ,W′ rainp ,W′ dust ]
[0079] S33, for the time series W in step 1 temp ,W pv ,W ws ,W rainp ,W dust The time series W′ obtained in step S32 temp ,W′ pv ,W′ ws ,W′ rainp ,W′ dust The first layer of multi-dimensional time domain convolution calculations are performed respectively to obtain Conv1 and Conv1′, where:
[0080] Conv1 = λ conv
[0081] λ=[λ temp ,λ pv ,λ ws ,λ rainp ,λ dust ]
[0082] conv=[conv temp ,conv pv ,conv ws ,conv rainp ,conv dust ]
[0083] In the above formula, Conv1 is the time series W temp ,W pv ,W ws ,W rainp ,W dust The accumulated result after executing the first layer of multi-dimensional time-domain convolution calculation is obtained, λ is the initial weight vector in S31, conv is the convolution vector, conv temp ,conv pv ,conv ws ,conv rainp ,conv dust These are the convolution calculation results for each time series;
[0084] The convolution calculation method of each time series is as follows:
[0085]
[0086] With conv temp Take calculation as an example, Relu is the activation function, W temp is the convolution operation independent variable, and C1 is the model parameter of this layer;
[0087] Similarly, we can get
[0088] Conv1′=λ·conv′
[0089] λ=[λ temp ,λ pv ,λ ws ,λ rainp ,λ dust ]
[0090] conv′=[conv′ temp ,conv′ pv ,conv′ ws ,conv′ rainp ,conv′ dust ]
[0091] In the above formula, Conv1′ is the time series W′ temp ,W′ pv ,W′ ws ,W′ rainp ,W′ dust The accumulated result after executing the first layer of multi-dimensional time domain convolution calculation is obtained, λ is the initial weight vector in S31, conv′ is the convolution vector, conv′ temp ,conv′ pv ,conv′ ws ,conv′ rainp ,conv′ dustThese are the convolution calculation results for each time series;
[0092] The convolution calculation method of each time series is as follows:
[0093]
[0094] conv′ temp Take calculation as an example, Relu is the activation function, W′ temp is the convolution operation independent variable, and C1 is the model parameter of this layer;
[0095] S34, for Conv1 and Conv1′ obtained in step S33, subsequent k-1 layers of convolution operations are performed to obtain Convk and Convk′, as follows:
[0096]
[0097] And so on.
[0098]
[0099] Step 4, perform a reverse convolution operation on Convk and Convk′ obtained in step S34 to obtain:
[0100] W k =[W k_temp ,W k_pv ,W k_ws ,W k_rainp ,W k_dust ]
[0101] W k ′=[W′ k_temp ,W′ k_pv ,W′ k_ws ,W′ k_rainp ,W′ k_dust ]
[0102] Step 5, combine the time series obtained in step 3 and step 4:
[0103] W=[W temp ,W pv ,W ws ,W rainp ,W dust ]
[0104] W′=[W′ temp ,W′ pv ,W′ ws ,W′ rainp ,W′ dust ]
[0105] W k =[Wk_temp ,W k_pv ,W k_ws ,W k_rainp ,W k_dust ]
[0106] W k ′=[W′ k_temp ,W′ k_pv ,W′ k_ws ,W′ k_rainp ,W′ k_dust ]
[0107] Perform cross-feature accumulation to obtain the time series W used for the final prediction of power generation l ,
[0108] W l =[W l_temp ,W l_pv ,W l_ws ,W l_rainp ,W l_dust ]
[0109] Step 6: Use the time series W obtained in step 5 l The power generation forecast curve P for the next 7 days is calculated as follows:
[0110]
[0111] In the above formula, p i is the i-th point in the power generation forecast curve for the next 7 days, g, h1, h2, h3, h4 are model importance functions, and the operator Indicates a positive effect, i.e. an increase in power generation; operator Indicates the reverse effect, that is, the power generation decreases; W l_pv_i , W l_ws_i , W l_temp_i , W l_rainp_i , W l_dust_i W l =[W l_temp ,W l_pv ,W l_ws ,W l_rainp ,W l_dust ] is the i-th element in ; m is the total number of points in the power generation prediction curve within the future set time period.
[0112] Then we can get the calculated result, the power generation forecast curve for the next 7 days P = [p1, p2, ..., p i ,…,p m ]
[0113] Step 7: Divide the power generation forecast curve for the next 7 days obtained in step 6 into 7 sub-vectors, namely:
[0114] P=[P1,P2,…,P7]
[0115] Among them, P1, P2, ..., P7 are the power generation prediction sub-curves for the first day, the second day, and the seventh day in the future respectively;
[0116] Step 8: Use the CP (Crowned Porcupine) algorithm for training and parameter optimization, as follows:
[0117] S81, assuming that the above steps are run from Day 0, the power generation forecast curve for the next 7 days starting from Day 0 can be calculated:
[0118] P Day0 =[P Day0+1 ,P Day0+2 ,…,P Day0+7 ]
[0119] Similarly, we can get,
[0120] P Day1 =[P Day1+1 ,P Day1+2 ,…,P Day1+7 ]
[0121] P Day2 =[P Day2+1 ,P Day2+2 ,…,P Day2+7 ]
[0122] P Day3 =[P Day3+1 ,P Day3+2 ,…,P Day3+7 ]
[0123] P Day4 =[P Day4+1 ,P Day4+2 ,…,P Day4+7 ]
[0124] P Day5 =[P Day5+1 ,P Day5+2 ,…,P Day5+7 ]
[0125] P Day6 =[P Day6+1 ,P Day6+2 ,…,P Day6+7 ]
[0126] Among them, P Day1 The power generation forecast curve for the next 7 days is calculated from Day 1. And so on. According to each forecast curve, there is the following table:
[0127] Table S81 Power Generation Forecast
[0128]
[0129]
[0130] S82, on Day 0+7, with the help of the actual power generation data of the photovoltaic power station, collect the actual daily power generation curve of the previous 7 days and record it as:
[0131] P′ Day0 =[P′ Day0+1 ,P′ Day0+2 ,…,P′ Day0+7 ]
[0132] Among them, P′ Day0+1 ,P′ Day0+2 ,…,P′ Day0+7 They are respectively the measured power generation curves for the next 7 days starting from Day 0;
[0133] S83, convert table S81 into table S83, using the following method,
[0134] Table S82 Output results and weight table
[0135] Output <![CDATA[out1]]> <![CDATA[out2]]> <![CDATA[out3]]> <![CDATA[out4]]> <![CDATA[out5]]> <![CDATA[out6]]> <![CDATA[out7]]> Weight <![CDATA[θ1]]> <![CDATA[θ2]]> <![CDATA[θ3]]> <![CDATA[θ4]]> <![CDATA[θ5]]> <![CDATA[θ6]]> <![CDATA[θ7]]>
[0136] in,
[0137]
[0138] By analogy, we have
[0139]
[0140] Among them, β u is the CP algorithm model parameter;
[0141] θ1+θ2+θ3+θ4+θ5+θ6+θ7=1
[0142] S84, setting a multi-layer optimization strategy according to the CP (Crowned Porcupine) algorithm, as follows:
[0143] S841,
[0144]
[0145] No optimization required;
[0146] S842,
[0147]
[0148] Adopt the first-layer optimization strategy in the CP algorithm and adjust the model importance functions such as g, h1, h2, h3, and h4 in step 6 until S841 is satisfied;
[0149] S843,
[0150]
[0151] Adopting the second-layer optimization strategy in the CP algorithm, on the basis of completing the adjustment of the model importance function in S842, further optimize the initial weight vector λ of each column matrix in the meteorological forecast data matrix W of the input layer for the next 7 days in S31 until S841 is satisfied;
[0152] S844,
[0153]
[0154] The third-layer optimization strategy in the CP algorithm is adopted. After adjusting the model importance functions such as g, h1, h2, h3, and h4 in step 6, and optimizing the initial weight vector λ of each column matrix in the meteorological forecast data matrix W of the input layer for the next 7 days in S31, the model parameter set [C1,…,C k ] until S841 is satisfied;
[0155] S845,
[0156]
[0157] The fourth-layer optimization strategy in the CP algorithm is adopted to adjust the model importance functions such as g, h1, h2, h3, and h4 in step 6, optimize the initial weight vector λ of each column matrix in the meteorological forecast data matrix W of the input layer in S31 for the next 7 days, and optimize the model parameter set [C1,…,C k ], further increase the number of hidden layers in step 2 until S841 is met;
[0158] Among them, π1, π2, π3, and π4 are system judgment parameters;
[0159] The optimization method of each parameter corresponds to the four-layer defense logic of the CP algorithm, which is progressive;
[0160] The above steps complete the optimization of the prediction model based on the CP algorithm.
[0161] Example 2
[0162] This embodiment provides a multivariable power generation regression prediction system based on CPO-BiTCN-BiGRU, including:
[0163] A data acquisition unit is used to acquire a time series matrix within a future set period of the target electric field, wherein the time series matrix includes a meteorological temperature prediction data time subsequence, an irradiance prediction data time subsequence, a wind speed prediction data time subsequence, a precipitation probability prediction data time subsequence, and a dust / particulate matter prediction data time subsequence;
[0164] A model building module is used to construct an initial model using a bidirectional time-domain convolutional network and a bidirectional gated recurrent unit; and to optimize the initial model using the historical power generation data of the target electric field combined with the crested porcupine algorithm to obtain a regression prediction model;
[0165] The power generation prediction module is used to use the obtained time series matrix as the input of the regression prediction model to obtain the power generation prediction curve within the future set time period.
[0166] Example 3
[0167] This embodiment 3 provides a computer device, including: a memory for storing a computer program; and a processor for implementing the steps of a computer method when executing the computer program.
[0168] When the processor executes the computer program, the steps of the above computer method are implemented.
[0169] Alternatively, the processor implements the functions of each module in the above system when executing the computer program,
[0170] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, and a cloud server. The computer device may include, but is not limited to, a processor and a memory. Those skilled in the art will appreciate that the above are examples of computer devices and do not constitute a limitation on computer devices. The computer device may include more components than the above, or a combination of certain components, or different components. For example, the computer device may also include input and output devices, network access devices, buses, etc.
[0171] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the computer device, and uses various interfaces and lines to connect various parts of the entire computer device.
[0172] The memory may be used to store the computer programs and / or modules, and the processor implements various functions of the computer device by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory.
[0173] The memory may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (SmartMediaCard, SMC), a secure digital (SecureDigital, SD) card, a flash card (FlashCard), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.
[0174] Example 4
[0175] This embodiment 4 also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps of the method described are implemented.
[0176] If the module / unit integrated in the computer system is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium.
[0177] Based on such understanding, the present invention implements all or part of the processes in the above method, and can also be completed by instructing related hardware through a computer program, and the computer program can be stored in a computer-readable storage medium, and when the computer program is executed by a processor, the steps of the above computer method can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or preset intermediate form, etc.
[0178] The computer-readable storage medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium, etc.
[0179] It should be noted that the content contained in the computer-readable storage medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable storage media do not include electrical carrier signals and telecommunication signals.
[0180] Example 5
[0181] This embodiment 5 provides a computer product, which includes a computer program, and the computer program is stored in a computer-readable storage medium; the processor of the computer device reads the computer program from the computer-readable storage medium, and the processor executes the computer program, so that the computer device can execute the method in embodiment 1, which will not be repeated here.
[0182] It should be noted that a person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing related hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods.
[0183] Compared with the prior art, the present invention has the following beneficial effects:
[0184] The multivariate power generation regression prediction model based on CPO-BiTCN-BiGRU provided by the present invention combines BiTCN (bidirectional time-domain convolutional network) and BiGRU (bidirectional gated recurrent unit) and is optimized by CP (Crown Porcupine Optimization Algorithm), which has the following advantages:
[0185] 1) BiTCN can effectively capture local patterns and long-term dependencies in time series, while BiGRU is better at capturing long-term dependencies in sequences. By combining the two, the feature information in the sequence can be captured more comprehensively and accurately;
[0186] 2) Both BiTCN and BiGRU have a bidirectional structure, which can consider the previous and next information of the sequence data at the same time, which helps to better understand and predict the patterns and trends in the sequence and improve the expressiveness of the model;
[0187] 3) BiTCN and BiGRU each have certain advantages in processing sequence data. Combining them can make up for each other's shortcomings, improve the performance and stability of the model, and make the time series prediction results more accurate and reliable.
[0188] 4) BiTCN is suitable for capturing features of shorter time scales, while BiGRU can handle long-term dependencies. Combining the two can better process information of different time scales in time series;
[0189] 5) The initial BiTCN-BiGRU model has many parameters. The BiTCN-BiGRU model is optimized based on CPO to avoid the limitations and blindness of manual parameter adjustment, improve its prediction accuracy, and reduce manual parameter adjustment.
[0190] In summary, the present invention is applicable to the prediction of power generation of photovoltaic power stations and has broad application prospects and social benefits.
[0191] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.
Claims
1. A multivariate power generation regression prediction method based on CPO-BiTCN-BiGRU, characterized in that: The following steps are involved: Step 1, obtaining a time series matrix of the target electric field within a future set period, the time series matrix including a time subsequence of meteorological temperature prediction data, a time subsequence of irradiance prediction data, a time subsequence of wind speed prediction data, a time subsequence of precipitation probability prediction data, and a time subsequence of dust / particulate matter prediction data; Step 2, using the obtained time series matrix as the input of the regression prediction model to obtain the power generation prediction curve within the future set period; Wherein, the construction method of the regression prediction model is: The initial model is constructed using a bidirectional time-domain convolutional network and a bidirectional gated recurrent unit; The historical power generation data of the target electric field was combined with the crested porcupine algorithm to optimize the initial model and obtain a regression prediction model.
2. The multivariate power generation regression prediction method based on CPO-BiTCN-BiGRU according to claim 1 is characterized in that: In step 2, the obtained time series matrix is used as the input of the regression prediction model to obtain the power generation prediction curve within the future set period. The specific method is: S21, setting the initial weight of each subsequence in the time series matrix; S22, using the initial weights and combining the bidirectional time-domain convolutional network to perform multi-dimensional time-domain convolution processing on the obtained time series matrix to obtain a convolution operation result; S23, combining the obtained convolution operation result with the bidirectional gated recurrent unit to perform a reverse convolution operation to obtain a reverse convolution operation result; S24, cross-feature accumulation of the obtained convolution operation result and the inverse convolution operation result to obtain a time series used for the final prediction of power generation; S25, using the obtained time series used for the final predicted power generation to calculate the power generation prediction curve within the future set time period, to obtain the power generation prediction curve.
3. The multivariate power generation regression prediction method based on CPO-BiTCN-BiGRU according to claim 2 is characterized in that: In S22, the initial weights are used in combination with the bidirectional time-domain convolutional network to perform multi-dimensional time-domain convolution processing on the obtained time series matrix to obtain the convolution operation result. The specific method is: S221, performing a time domain flip operation on the time series matrix to obtain a time domain flip operation matrix; S222, performing a first-layer multi-dimensional time-domain convolution calculation on the time series matrix and the time-domain flip operation matrix, respectively, to obtain a first-layer time-domain convolution calculation cumulative result corresponding to the time series matrix and a first-layer time-domain convolution calculation cumulative result corresponding to the time-domain flip operation matrix, respectively; S223, perform k-1 layers of convolution operations on the accumulated results of the first layer of time domain convolution calculations corresponding to the time series matrix and the accumulated results of the first layer of time domain convolution calculations corresponding to the time domain flip operation matrix, respectively, to obtain the convolution operation results corresponding to the time series matrix and the convolution operation results corresponding to the time domain flip operation matrix.
4. The multivariate power generation regression prediction method based on CPO-BiTCN-BiGRU according to claim 3 is characterized in that: In S23, the obtained convolution operation result is combined with the bidirectional gated recurrent unit to perform a reverse convolution operation to obtain a reverse convolution operation result. The specific method is: The convolution operation results corresponding to the obtained time series matrix and the convolution operation results corresponding to the time domain flip operation matrix are respectively subjected to inverse convolution operation to obtain the inverse convolution operation results corresponding to the time series matrix and the inverse convolution operation results corresponding to the time domain flip operation matrix.
5. The multivariate power generation regression prediction method based on CPO-BiTCN-BiGRU according to claim 1 is characterized in that: The expression of the power generation prediction curve in the future set period is: 1≤i≤m Among them, p i is the i-th point in the power generation prediction curve within the future set time period; g, h1, h2, h3, and h4 are all model importance functions; is an operator, indicating a positive effect; is an operator, indicating the reverse effect; W l_pv_i , W l_ws_i , W l_temp_i , W l_rainp_i , W l_dust_i are all elements in the time series used for the final prediction of power generation; m is the total number of points in the power generation prediction curve within the future set time period.
6. The multivariate power generation regression prediction method based on CPO-BiTCN-BiGRU according to claim 1 is characterized in that: In step 3, the historical power generation data of the target electric field is combined with the crested porcupine algorithm to optimize the initial model and obtain a regression prediction model. The specific method is: Obtain the historical actual power generation data of the target electric field and the time series matrix within the set time period, the time series matrix including the meteorological temperature prediction data time subsequence, the irradiance prediction data time subsequence, the wind speed prediction data time subsequence, the precipitation probability prediction data time subsequence, and the dust / particulate matter prediction data time subsequence; The time series array is used as the input of the initial model to obtain the power generation forecast curve within the set time period; The initial model is optimized based on the obtained power generation prediction curve and the historical actual power generation data within the set time period combined with the crested porcupine algorithm to obtain a regression prediction model.
7. A multivariable power generation regression prediction system based on CPO-BiTCN-BiGRU, characterized in that: include: A data acquisition unit is used to acquire a time series matrix within a future set period of the target electric field, wherein the time series matrix includes a meteorological temperature prediction data time subsequence, an irradiance prediction data time subsequence, a wind speed prediction data time subsequence, a precipitation probability prediction data time subsequence, and a dust / particulate matter prediction data time subsequence; A model building module is used to construct an initial model using a bidirectional time-domain convolutional network and a bidirectional gated recurrent unit; and to optimize the initial model using the historical power generation data of the target electric field combined with the crested porcupine algorithm to obtain a regression prediction model; The power generation prediction module is used to use the obtained time series matrix as the input of the regression prediction model to obtain the power generation prediction curve within the future set time period.
8. A computer device, characterized in that: include: a processor suitable for executing a computer program; A computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, and when the computer program is executed by the processor, the method according to any one of claims 1 to 6 is performed.
9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.
10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.
Citation Information
Cited By
Wind turbine generator and fault early warning method and device thereof, storage medium and program product
CN119686929A
Wind turbine and fault early warning method and device thereof, storage medium and program product
CN119686929B
Short-term photovoltaic power prediction method for optimizing bidirectional gating circulation unit based on improved crown porcupine algorithm
CN120090163A
Online monitoring and early warning method for bus capacitor of electric drive system
CN120722094A
