Deep learning model analysis method and system for new energy power generation output prediction

Abstract

The invention discloses a deep learning model analysis method and system for new energy power generation output prediction. The method comprises the following steps: constructing a new energy power generation output prediction model based on deep learning, and training the prediction model; performing input feature simplification on the trained prediction model, and extracting key influence factors; contribution degree calculation is carried out on the key influence factors, and the influence degree of the key influence factors on the prediction result is analyzed; carrying out meteorological type division on the output data sample based on a clearness index; transforming a regression layer of the prediction model into a logical reasoning layer; and performing parameter fine tuning on the logical reasoning layer of the prediction model based on the output data samples divided by the meteorological types, and analyzing the output logic of the prediction model. According to the method, the credibility level of the new energy power generation output prediction deep learning model can be improved, electric power operators are assisted in understanding the prediction reasoning logic of the deep learning model, and the safety and stability of the electric power system for absorbing new energy power generation are improved.

Description

Technical Field

[0001] The present invention relates to a deep learning model analysis method and system for renewable energy power generation output prediction, belonging to the technical field of power system big data analysis. Background Art

[0002] The penetration rate of renewable energy power generation in new power systems is increasing year by year, which makes the randomness and volatility of renewable energy power generation the main factors affecting the stability of new power systems. Therefore, the development of high-precision renewable energy power generation output forecasting technology is of great significance for improving the safe and stable operation of power systems and promoting green and low-carbon energy transformation.

[0003] Regarding the prediction technology of renewable energy power generation output, the models constructed by the current mainstream methods are mainly divided into simple statistical models, traditional machine learning models and deep learning models. Among them, the deep learning model based on deep neural network can better fit high-dimensional nonlinear problems, so it performs well in various multi-time scale renewable energy power generation output predictions. However, the improvement of the prediction accuracy of the deep learning model comes at the cost of the model becoming deeper and more complex, and the interpretability of the prediction model is extremely poor. Especially when extreme prediction error events occur, the deep learning model with poor interpretability is difficult to provide effective guidance for the adjustment of renewable energy power generation and consumption strategies. Therefore, it is urgent to develop analytical methods for deep learning models for renewable energy power generation output prediction. Summary of the invention

[0004] Purpose of the invention: In view of the shortcomings of the current power system big data analysis technology field in the renewable energy power generation prediction method, the present invention proposes a deep learning model analysis method and system for renewable energy power generation output prediction, which improves the performance of the deep learning model for renewable energy power generation output prediction by analyzing the key influencing factors and output logic of the deep learning model, and helps to guide the power system dispatching staff to adjust the renewable energy power generation and consumption strategy in a timely manner, thereby increasing the power system's renewable energy consumption capacity and enhancing the competitiveness of power grid companies.

[0005] Technical solution: In order to achieve the above-mentioned invention object, the deep learning model analysis method for renewable energy power generation output prediction proposed in the present invention comprises the following steps:

[0006] Step 1: Construct a new energy power generation output prediction model based on deep learning, and train the prediction model based on the new energy power generation output data sample. The input features of the prediction model and data samples include the new energy power generation historical time series, meteorological historical time series, meteorological forecast and timestamp. The formula is expressed as:

[0007]

[0008] In the formula, is the output forecast value of renewable energy power generation from time t+1 to t+L; f pred is a deep learning model; x pv is the historical time series of renewable energy power generation, contained in the real number domain of dimension D×1 x w is the meteorological historical time series, contained in the real number domain range of dimension D×W x wp For weather forecast, it is contained in the real number domain of dimension L×W x ts is a timestamp, contained in the real number domain range of dimension L×S L is the length of the output forecast value; D is the length of the historical time series; W is the number of meteorological attributes; S is the number of timestamp dimensions;

[0009] Step 2: Simplify the input features of the trained prediction model and extract the key influencing factors Among them, 1+W+S is the number of key influencing factors, is the range of the real number domain of dimension 1+W+S;

[0010] Step 3: Use the Shapley value summation interpretation algorithm to calculate the contribution value of the key influencing factors and analyze the impact of the key influencing factors on the prediction results. The analytical formula is expressed as:

[0011]

[0012] x j ∈x,1≤i≤L,1≤j≤(1+W+S)

[0013] In the formula, is the output forecast value of renewable energy power generation at time t+i, φ0 is the contribution bias value, x j is the jth key influencing factor, g i The key influencing factors Contribution value of

[0014] Step 4: Classify the meteorological types of the renewable energy power generation output data samples based on the clear sky index;

[0015] Step 5: Transform the regression layer of the new energy power generation output prediction model into a logical reasoning layer, fine-tune the parameters of the logical reasoning layer of the prediction model based on the new energy power generation output data samples divided by meteorological types, and parse out the output logic of the prediction model. The analytical formula is expressed as:

[0016]

[0017] In the formula, C t+iis the meteorological type of the data sample at time t+i, p(C t+i = k) is the probability that the meteorological type of the data sample at time t+i belongs to the kth category, h k,t+i is the probability weight that the meteorological type of the data sample at time t+i belongs to the kth category.

[0018] As a preferred method, in step 2, the input features of the trained prediction model are simplified to extract the key influencing factors. The specific implementation steps are as follows:

[0019] Construct an input feature approximation function for the trained prediction model, and the formula is expressed as:

[0020]

[0021] In the formula, x wp,t+i is the weather forecast at time t+i, contained in the real number domain of dimension 1×W x ts,t+i is the timestamp at time t+i, contained in the real number domain range of dimension 1×S Φ1 is the approximate function of the historical time series of renewable energy power generation, which is included in the real number domain. Φ2 is an approximate function of the weather, which is contained in the real number domain of dimension W. Φ3 is an approximate function of the timestamp, which is contained in the real number domain of dimension S.

[0022] Based on the goal of minimizing the prediction error and fitting error, the input feature approximate function is solved, and the solved input feature approximate function is merged according to the column vector to obtain the simplified and extracted key influencing factors. The formula is expressed as follows:

[0023]

[0024] In the formula, f concat is the column vector merging function, and x is the key influencing factor.

[0025] As a preference, the objective function formula for solving the input feature approximation function is:

[0026]

[0027] Where P t+i is the actual value of the output of renewable energy power generation at time (t+i), is the inverse function of Φ1, is the inverse function of Φ2, is the inverse function of Φ3, is the sum of squares of errors.

[0028] Preferably, in step 3, the contribution value of the key influencing factor is calculated using the Shapley value sum interpretation algorithm, and the calculation formula of the algorithm is:

[0029]

[0030] In the formula, x j is the jth key influencing factor, g i The key influencing factors The contribution value of x\{x j} does not contain x j , S is the subset of key influencing factors included in x\{x j}, S∪{x j} is a subset S containing x j The set after, || is the length of the set, ! is the factorial, x S is the key impact factor in S, For S∪{x j}, For input x S The deep learning model For input Deep learning model.

[0031] As a preferred embodiment, in step 4, the meteorological type classification of the renewable energy power generation output data samples is performed based on the clear sky index index, and the formula is expressed as follows:

[0032]

[0033] In the formula, C t is the meteorological type of the data sample at time t, k is the meteorological type number, K is the total number of meteorological types, if(C t = k) is the conditional judgment that the meteorological type of the data sample at time t belongs to the kth category, P t a is the actual value of renewable energy power generation output of the data sample at time t, P t clear is the clear sky value of renewable energy power generation output of the data sample at time t, The clear sky index.

[0034] Preferably, in step 5, the regression layer of the new energy power generation output prediction model is transformed into a logical reasoning layer, and the specific implementation steps are as follows:

[0035] The number of output nodes of the regression layer of the new energy power generation output prediction model is set to the total number of meteorological types K, and a K-category classification function is constructed. The formula is expressed as: p(C t+i = = k) = f class (H D×W D×K )[k]; 1≤k≤K, where C t+i is the meteorological type of the data sample at time (t+i), p(C t+i = k) is the probability that the meteorological type of the data sample at time (t+i) belongs to the kth category, f class is the classification function, H D is the input node of the regression layer, D is the number of input nodes, W D×K is the weight of the classification function, [k] is the output node number of the classification function;

[0036] The weight W of the classification function D×K The weight vector in is set as the terminal node of the logical reasoning layer, and the formula is expressed as: i =w k ∈W D×K ; 1≤i≤K, where n i is the i-th terminal node of the logical reasoning layer, w k W D×K The kth weight vector in ;

[0037] Calculate the inference node of the previous layer of the terminal node until ∑ j K(j)=1, the last n j It is the head end node of the logical reasoning layer;

[0038] Copy the K-category classification function, randomly initialize the weights of the classification function, and construct the classification weight function. The formula is expressed as: 1≤k≤K, where h k,t+i is the probability weight of the meteorological type of the data sample at time t+i belonging to the kth category, is the classification weight function, H D is the input node of the regression layer, is the weight of the randomly initialized classification function, and [k] is the output node number of the classification function.

[0039] As a preference, the inference node of the previous layer of the terminal node is calculated, and the formula is expressed as:

[0040]

[0041] Where n j is the jth inference node in the previous layer of the terminal node, and K(j) is the number of terminal nodes under the jth inference node.

[0042] The present invention also provides a deep learning model analysis system for new energy power generation output prediction, comprising:

[0043] The model building and training module is used to build a new energy power generation output prediction model based on deep learning, and train the prediction model based on new energy power generation output data samples. The input features of the prediction model and data samples include the new energy power generation historical time series, meteorological historical time series, meteorological forecast and timestamp. The formula is expressed as follows:

[0044]

[0045] In the formula, is the output forecast value of renewable energy power generation from time t+1 to t+L; f pred is a deep learning model; x pv is the historical time series of renewable energy power generation, contained in the real number domain of dimension D×1 x w is the meteorological historical time series, contained in the real number domain range of dimension D×W x wp For weather forecast, it is contained in the real number domain of dimension L×W x ts is a timestamp, contained in the real number domain range of dimension L×S L is the length of the output forecast value; D is the length of the historical time series; W is the number of meteorological attributes; S is the number of timestamp dimensions;

[0046] Input feature simplification module, used to simplify the input features of the trained prediction model and extract key influencing factors Among them, 1+W+S is the number of key influencing factors, is the range of the real number domain of dimension 1+W+S;

[0047] The impact degree analysis module is used to calculate the contribution value of key impact factors using the Shapley value summation interpretation algorithm, and analyze the impact degree of key impact factors on the prediction results. The analytical formula is expressed as:

[0048]

[0049] x j ∈x,1≤i≤L,1≤j≤(1+W+S)

[0050] In the formula, is the output forecast value of renewable energy power generation at time t+i, φ0 is the contribution bias value, x j is the jth key influencing factor, g i The key influencing factors Contribution value of

[0051] The meteorological type classification module is used to classify the meteorological types of the renewable energy power generation output data samples based on the clear sky index index;

[0052] The model output parsing module is used to transform the regression layer of the new energy power generation output prediction model into a logical reasoning layer, fine-tune the parameters of the logical reasoning layer of the prediction model based on the new energy power generation output data samples divided by meteorological types, and parse out the output logic of the prediction model. The parsing formula is expressed as:

[0053]

[0054] In the formula, C t+i is the meteorological type of the data sample at time t+i, p(C t+i = k) is the probability that the meteorological type of the data sample at time t+i belongs to the kth category, h k,t+i is the probability weight that the meteorological type of the data sample at time t+i belongs to the kth category.

[0055] The present invention also provides a computer device, comprising: one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and when the programs are executed by the processors, the steps of the deep learning model analysis method for predicting new energy power generation output as described above are implemented.

[0056] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the deep learning model analysis method for predicting the output of new energy power generation as described above.

[0057] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages:

[0058] (1) The present invention constructs a new energy power generation output prediction model based on deep learning and trains the prediction model; simplifies the input features of the trained prediction model and extracts the key influencing factors; calculates the contribution of the key influencing factors and analyzes the degree of influence of the key influencing factors on the prediction results; classifies the output data samples into meteorological types based on the clear sky index indicator; transforms the regression layer of the prediction model into a logical reasoning layer; fine-tunes the parameters of the logical reasoning layer of the prediction model based on the output data samples classified by meteorological types, and analyzes the output logic of the prediction model. Compared with the existing model analysis method, the method of the present invention can better adapt to the deep learning model and has higher accuracy in the prediction of new energy power generation output;

[0059] (2) The method of the present invention can meet the forecasting needs of various types of renewable energy power generation output based on numerical meteorological forecasts, and improve the interpretability and reliability of renewable energy power generation output forecasting models;

[0060] (3) The model analysis results of the method of the present invention can provide guidance and data support for dispatchers of new power systems, thereby improving the power system's renewable energy generation and consumption capacity and improving the economic efficiency of the power system's operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 Schematic diagram of the principle of the method of the present invention.

[0062] Figure 2 This is a diagram showing the analysis results of key influencing factors of the method of the present invention.

[0063] Figure 3 This is a diagram of the output logic analysis result of the method of the present invention. DETAILED DESCRIPTION

[0064] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0065] like Figure 1 As shown, the present invention proposes a deep learning model analysis method for renewable energy power generation output prediction, which specifically includes the following steps:

[0066] Step 1: Construct a new energy power generation output prediction model based on deep learning, and train the prediction model based on the new energy power generation output data sample. The input features of the prediction model and data samples include the new energy power generation historical time series, meteorological historical time series, meteorological forecast and timestamp. The formula is expressed as:

[0067]

[0068] In the formula, is the output forecast value of renewable energy power generation from time (t+1) to (t+L), f pred is a deep learning model, x pv is the historical time series of renewable energy generation, contained in the real number domain of dimension D×1 x w is the meteorological historical time series, contained in the real number domain range of dimension D×W x wp For weather forecast, it is contained in the real number domain of dimension L×W x ts A timestamp, contained in the real number domain of dimension L×S L is the length of the output forecast value, D is the length of the historical time series, W is the number of meteorological attributes, and S is the number of timestamp dimensions;

[0069] Step 2: Simplify the input features of the trained prediction model and extract the key influencing factors Among them, (1+W+S) is the number of key influencing factors, is the range of the real number domain of dimension (1+W+S)

[0070] Step 3: Use the Shapley value summation interpretation algorithm to calculate the contribution value of the key influencing factors and analyze the impact of the key influencing factors on the prediction results. The analytical formula is expressed as:

[0071]

[0072] x j ∈x,1≤i≤L,1≤j≤(1+W+S)

[0073] In the formula, is the output forecast value of renewable energy power generation at time (t+i), φ0 is the contribution bias value, x j is the jth key influencing factor, g i The key influencing factors Contribution value of

[0074] Step 4: Classify the meteorological types of the renewable energy power generation output data samples based on the clear sky index index. The formula is:

[0075]

[0076] In the formula, C t is the meteorological type of the data sample at time t, k is the meteorological type number, K is the total number of meteorological types, if(C t = k) is the conditional judgment that the meteorological type of the data sample at time t belongs to the kth category, P t a is the actual value of renewable energy power generation output of the data sample at time t, P t clear is the clear sky value of renewable energy power generation output of the data sample at time t, is the clear sky index;

[0077] Step 5: Transform the regression layer of the new energy power generation output prediction model into a logical reasoning layer, fine-tune the parameters of the logical reasoning layer of the prediction model based on the new energy power generation output data samples divided by meteorological types, and parse out the output logic of the prediction model. The analytical formula is expressed as:

[0078]

[0079] In the formula, C t+i is the meteorological type of the data sample at time (t+i), p(C t+i = k) is the probability that the meteorological type of the data sample at time (t+i) belongs to the kth category, h k,t+i is the probability weight that the meteorological type of the data sample at time (t+i) belongs to the kth category.

[0080] The following is a detailed description of the specific implementation process of using the method in the present invention to perform deep learning model analysis for renewable energy power generation output prediction, in conjunction with specific embodiments. The present invention selects distributed photovoltaic data from a county in Jiangsu Province in 2018 to extract a 24-hour renewable energy power generation output forecast. Therefore, the length of the output forecast value is selected = 24, the length of the historical time series = 168, the number of meteorological attributes = 8 (temperature, humidity, wind speed, wind direction, cloud coverage, total radiation, direct radiation, scattered radiation), and the number of timestamp dimensions = 2 (hours, days). Based on the above data, the specific implementation steps of the method of the present invention are as follows:

[0081] First, a new energy power generation output prediction model based on deep learning is constructed. The prediction model is trained based on new energy power generation output data samples. The input features of the prediction model and data samples include the new energy power generation historical time series, meteorological historical time series, meteorological forecast and timestamp. The formula is expressed as follows:

[0082]

[0083] In the formula, is the output forecast value of renewable energy power generation from time (t+1) to (t+L); f pred is a deep learning model; x pv is the historical time series of renewable energy power generation, contained in the real number domain of dimension D×1 x w is the meteorological historical time series, contained in the real number domain range of dimension D×W x wp For weather forecast, it is contained in the real number domain of dimension L×W x ts is a timestamp, contained in the real number domain range of dimension L×S L is the length of the output forecast value; D is the length of the historical time series; W is the number of meteorological attributes; S is the number of timestamp dimensions;

[0084] Secondly, the input features of the trained prediction model are simplified and the input feature approximation function is constructed. The formula is expressed as:

[0085]

[0086] In the formula, x wp,t+i is the weather forecast at time (t+i), contained in the real number domain of dimension 1×W x ts,t+i is the timestamp at time (t+i), contained in the real number domain range of dimension 1×S Φ1 is the approximate function of the historical time series of renewable energy power generation, which is included in the real number domain. Φ2 is an approximate function of the weather, which is contained in the real number domain of dimension W. Φ3 is an approximate function of the timestamp, which is contained in the real number domain of dimension S. Based on the goal of minimizing the prediction error and fitting error, the input feature approximation function is solved. The objective function formula is:

[0087]

[0088] Where P t+i is the actual value of the output of renewable energy power generation at time (t+i), is the inverse function of Φ1, is the inverse function of Φ2, is the inverse function of Φ3, is the sum of squares of the errors; the solved input feature approximate functions are merged by column vectors to obtain the simplified and extracted key influencing factors, which are expressed as follows:

[0089]

[0090] In the formula, f concat is the column vector merging function, x is the key influencing factor; the key influencing factors are extracted by simplification Among them, (1+W+S) is the number of key influencing factors, is the range of the real number domain of dimension (1+W+S);

[0091] Next, the Shapley value sum interpretation algorithm is used to calculate the contribution value of the key influencing factors. The calculation formula of the algorithm is:

[0092]

[0093] In the formula, x j is the jth key influencing factor, g i The key influencing factors The contribution value of x\{x j} does not contain x j , S is the subset of key influencing factors included in x\{x j}, S∪{x j} is a subset S containing x j The set after, || is the length of the set, ! is the factorial, x S is the key impact factor in S, For S∪{x j}, For input x S The deep learning model For input The deep learning model is used to analyze the influence of key influencing factors on the prediction results. The analytical formula is expressed as:

[0094]

[0095] x j ∈x,1≤i≤L,1≤j≤(1+W+S)

[0096] In the formula, is the output forecast value of renewable energy power generation at time (t+i), φ0 is the contribution bias value, x j is the jth key influencing factor, g i The key influencing factors The contribution value of Figure 2 The result diagram of the key influencing factor analysis of the deep learning model for the prediction of renewable energy power generation output in the embodiment of the present invention, wherein the deep learning model predicts that a huge photovoltaic power ramp will occur at 12:30 noon on the same day. After the key influencing factor analysis of the present invention, the prediction result is mainly affected by the cloud coverage rate in the weather forecast, and the credibility is relatively high; after comparing with the true value afterwards, the prediction value of the deep learning model at this moment is indeed more accurate, which is consistent with the analysis result of the present invention. Therefore, the method of the present invention can improve the reliability of the prediction of the deep learning model by analyzing the degree of influence of the key influencing factors on the prediction results.

[0097] Then, based on the clear sky index, the meteorological types of the renewable energy power generation output data samples are divided into the following formula:

[0098]

[0099] In the formula, C t is the meteorological type of the data sample at time t, k is the meteorological type number, K is the total number of meteorological types, and in the embodiment of the present invention, the meteorological types are selected as sunny, cloudy and rainy, and the value K=3; if (C t = k) is the conditional judgment that the meteorological type of the data sample at time t belongs to the kth category, P t a is the actual value of renewable energy power generation output of the data sample at time t, P t clear is the clear sky value of renewable energy power generation output of the data sample at time t, The clear sky index.

[0100] Finally, the regression layer of the renewable energy power generation output prediction model is transformed into a logical reasoning layer. The number of output nodes of the regression layer of the renewable energy power generation output prediction model is set to the total number of meteorological types K, and a K-category classification function is constructed. The formula is expressed as follows:

[0101] p(Ct+i = = k) = f class (H D ×W D×K )[k]; 1≤k≤K

[0102] In the formula, C t+i is the meteorological type of the data sample at time (t+i), p(C t+i = k) is the probability that the meteorological type of the data sample at time (t+i) belongs to the kth category, f class is the classification function, H D is the input node of the regression layer, D is the number of input nodes, W D×K is the weight of the classification function, [k] is the output node number of the classification function; the weight of the classification function W D×K The weight vector in is set as the terminal node of the logical reasoning layer, and the formula is expressed as:

[0103] n i =w k ∈W D×K ; 1≤i≤K

[0104] Where n i is the i-th terminal node of the logical reasoning layer, w k W D×K The kth weight vector in ; calculate the inference node of the previous layer of the terminal node, the formula is expressed as:

[0105]

[0106] Where n j is the jth inference node in the previous layer of the terminal node, K(j) is the number of terminal nodes under the jth inference node; repeat the above steps until ∑ j K(j)=1, the last n j It is the head node of the logical reasoning layer; it copies the K-category classification function, randomly initializes the weight of the classification function, and constructs the classification weight function. The formula is expressed as follows:

[0107]

[0108] In the formula, h k,t+i is the probability weight of the meteorological type of the data sample at time (t+i) belonging to the kth category, is the classification weight function, H D is the input node of the regression layer, is the weight of the randomly initialized classification function, [k] is the output node number of the classification function; based on the renewable energy power generation output data samples divided by meteorological types, the parameters of the logical reasoning layer of the prediction model are fine-tuned, and the output logic of the prediction model is parsed. The parsing formula is expressed as:

[0109]

[0110] In the formula, C t+i is the meteorological type of the data sample at time (t+i), p(C t+i = k) is the probability that the meteorological type of the data sample at time (t+i) belongs to the kth category, h k,t+i is the probability weight of the meteorological type of the data sample at time (t+i) belonging to the kth category, such as Figure 3 The result diagram of the output logic analysis of the deep learning model for the prediction of new energy power generation output, in which the deep learning model predicts that a huge photovoltaic power ramp will occur at 12:30 noon on the same day. After the output logic analysis of the present invention, the predicted value has a probability of more than 80% for cloudy or rainy days at the current moment, and the credibility is relatively high, but the probability of cloudy turning to overcast or continuing to remain cloudy is less than 50%, and the credibility is relatively low; after comparing with the true value afterwards, the deep learning model predicted the value at this moment more accurately, and the prediction error at subsequent moments increased significantly, which is consistent with the analysis results of the present invention. Therefore, the method of the present invention can intuitively display the credibility level of the prediction by parsing the logic of the prediction output, thereby improving the interpretability of the deep learning model. Therefore, the deep model analysis method of the present invention can provide guidance and data support for the dispatchers of new power systems to adjust the new energy power generation and consumption strategy.

[0111] In summary, the deep learning model analysis method for renewable energy power generation output prediction of the present invention can better adapt to the deep learning model, has higher accuracy in renewable energy power generation output prediction, can meet various renewable energy power generation output prediction needs based on numerical meteorological forecasts, improve the interpretability and reliability of renewable energy power generation output prediction models, and provide guidance and data support for dispatchers of new power systems, thereby improving the renewable energy power generation absorption capacity of the power system and improving the economy of the power system operation.

[0112] Based on the same technical concept as the method embodiment, the present invention also provides a deep learning model analysis system for new energy power generation output prediction, including:

[0113] The model building and training module is used to build a new energy power generation output prediction model based on deep learning, and train the prediction model based on new energy power generation output data samples. The input features of the prediction model and data samples include the new energy power generation historical time series, meteorological historical time series, meteorological forecast and timestamp. The formula is expressed as follows:

[0114]

[0115] In the formula, is the output forecast value of renewable energy power generation from time t+1 to t+L; f pred is a deep learning model; x pv is the historical time series of renewable energy generation, contained in the real number domain of dimension D×1 x w is the meteorological historical time series, contained in the real number domain range of dimension D×W x wp For weather forecast, it is contained in the real number domain of dimension L×W x ts is a timestamp, contained in the real number domain range of dimension L×S L is the length of the output forecast value; D is the length of the historical time series; W is the number of meteorological attributes; S is the number of timestamp dimensions;

[0116] Input feature simplification module, used to simplify the input features of the trained prediction model and extract key influencing factors Among them, 1+W+S is the number of key influencing factors, is the range of the real number domain of dimension 1+W+S;

[0117] The impact degree analysis module is used to calculate the contribution value of key impact factors using the Shapley value summation interpretation algorithm, and analyze the impact degree of key impact factors on the prediction results. The analytical formula is expressed as:

[0118]

[0119] x j ∈x,1≤i≤L,1≤j≤(1+W+S)

[0120] In the formula, is the output forecast value of renewable energy power generation at time t+i, φ0 is the contribution bias value, x j is the jth key influencing factor, g i The key influencing factors Contribution value of

[0121] The meteorological type classification module is used to classify the meteorological types of the renewable energy power generation output data samples based on the clear sky index index;

[0122] The model output parsing module is used to transform the regression layer of the new energy power generation output prediction model into a logical reasoning layer, fine-tune the parameters of the logical reasoning layer of the prediction model based on the new energy power generation output data samples divided by meteorological types, and parse out the output logic of the prediction model. The parsing formula is expressed as:

[0123]

[0124] In the formula, Ct+i is the meteorological type of the data sample at time t+i, p(C t+i = k) is the probability that the meteorological type of the data sample at time t+i belongs to the kth category, h k,t+i is the probability weight that the meteorological type of the data sample at time t+i belongs to the kth category.

[0125] It should be understood that the deep learning model analysis system for renewable energy power generation output prediction in the embodiment of the present invention can implement all the technical solutions in the above-mentioned method embodiment, and the functions of its various functional modules can be specifically implemented according to the methods in the above-mentioned method embodiments. The specific implementation process can refer to the relevant description in the above-mentioned embodiments, which will not be repeated here.

[0126] The present invention also provides a computer device, comprising: one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and when the programs are executed by the processors, the steps of the deep learning model analysis method for predicting new energy power generation output as described above are implemented.

[0127] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the deep learning model analysis method for predicting the output of new energy power generation as described above.

[0128] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, devices (systems), computer equipment or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0129] The present invention is described with reference to a flowchart of a method according to an embodiment of the present invention. It should be understood that each process in the flowchart and a combination of processes in the flowchart can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the process in the flowchart. Figure 1 A device that specifies functions in a process or multiple processes.

[0130] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A function specified in a process or multiple processes.

[0131] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 The steps of a specified function in a process or multiple processes.

Claims

1. A deep learning model analysis method for renewable energy power generation output prediction, characterized in that: The following steps are involved: Step 1: Construct a new energy power generation output prediction model based on deep learning, and train the prediction model based on the new energy power generation output data sample. The input features of the prediction model and data samples include the new energy power generation historical time series, meteorological historical time series, meteorological forecast and timestamp. The formula is expressed as: In the formula, is the output forecast value of renewable energy power generation from time t+1 to t+L; f pred is a deep learning model; x pv is the historical time series of renewable energy power generation, contained in the real number domain of dimension D×1 x w is the meteorological historical time series, contained in the real number domain range of dimension D×W x wp For weather forecast, it is contained in the real number domain of dimension L×W x ts is a timestamp, contained in the real number domain range of dimension L×S L is the length of the output forecast value; D is the length of the historical time series; W is the number of meteorological attributes; S is the number of timestamp dimensions; Step 2: Simplify the input features of the trained prediction model and extract the key influencing factors Among them, 1+W+S is the number of key influencing factors, is the range of the real number domain of dimension 1+W+S; Step 3: Use the Shapley value summation interpretation algorithm to calculate the contribution value of the key influencing factors and analyze the impact of the key influencing factors on the prediction results. The analytical formula is expressed as: In the formula, is the output forecast value of renewable energy power generation at time t+i, φ0 is the contribution bias value, x j is the jth key influencing factor, g i The key influencing factors Contribution value of Step 4: Classify the meteorological types of the renewable energy power generation output data samples based on the clear sky index; Step 5: Transform the regression layer of the new energy power generation output prediction model into a logical reasoning layer, fine-tune the parameters of the logical reasoning layer of the prediction model based on the new energy power generation output data samples divided by meteorological types, and parse out the output logic of the prediction model. The analytical formula is expressed as: In the formula, C t+i is the meteorological type of the data sample at time t+i, p(C t+i = k) is the probability that the meteorological type of the data sample at time t+i belongs to the kth category, h k,t+i is the probability weight that the meteorological type of the data sample at time t+i belongs to the kth category.

2. The method according to claim 1, characterized in that Step 2: Simplify the input features of the trained prediction model to extract the key influencing factors The specific implementation steps are as follows: Construct an input feature approximation function for the trained prediction model, and the formula is expressed as: In the formula, x wp,t+i is the weather forecast at time t+i, contained in the real number domain of dimension 1×W x ts,t+i is the timestamp at time t+i, contained in the real number domain range of dimension 1×S Φ1 is the approximate function of the historical time series of renewable energy power generation, which is included in the real number domain. Φ2 is an approximate function of the weather, which is contained in the real number domain of dimension W. Φ3 is an approximate function of the timestamp, which is contained in the real number domain of dimension S. Based on the goal of minimizing the prediction error and fitting error, the input feature approximate function is solved, and the solved input feature approximate function is merged according to the column vector to obtain the simplified and extracted key influencing factors. The formula is expressed as follows: In the formula, f concat is the column vector merging function, and x is the key influencing factor.

3. The method according to claim 2, characterized in that The objective function formula for solving the input feature approximation function is: Where P t+i is the actual value of the output of renewable energy power generation at time (t+i), is the inverse function of Φ1, is the inverse function of Φ2, is the inverse function of Φ3, is the sum of squares of errors.

4. The method according to claim 1, characterized in that: Step 3 uses the Shapley value sum interpretation algorithm to calculate the contribution value of the key influencing factor. The calculation formula of the algorithm is: In the formula, x j is the jth key influencing factor, g i The key influencing factors The contribution value of x\{x j } does not contain x j , S is the subset of key influencing factors included in x\{x j }, S∪{x j } is a subset S containing x j The set after, || is the length of the set, ! is the factorial, x S is the key impact factor in S, For S∪{x j }, For input x S The deep learning model For input Deep learning model.

5. The method according to claim 1, characterized in that In step 4, the meteorological type classification of the renewable energy power generation output data samples is performed based on the clear sky index index, and the formula is expressed as follows: In the formula, C t is the meteorological type of the data sample at time t, k is the meteorological type number, K is the total number of meteorological types, if(C t = k) is the conditional judgment that the meteorological type of the data sample at time t belongs to the kth category, P t a is the actual value of renewable energy power generation output of the data sample at time t, is the clear sky value of renewable energy power generation output of the data sample at time t, The clear sky index.

6. The method according to claim 1, characterized in that Step 5 describes the transformation of the regression layer of the new energy power generation output prediction model into a logical reasoning layer. The specific implementation steps are as follows: The number of output nodes of the regression layer of the new energy power generation output prediction model is set to the total number of meteorological types K, and a K-category classification function is constructed. The formula is expressed as: p(C t+i = = k) = f class (H D ×W D×K )[k]; 1≤k≤K, where C t+i is the meteorological type of the data sample at time (t+i), p(C t+i = k) is the probability that the meteorological type of the data sample at time t+i belongs to the kth category, f class is the classification function, H D is the input node of the regression layer, D is the number of input nodes, W D×K is the weight of the classification function, [k] is the output node number of the classification function; The weight W of the classification function D×K The weight vector in is set as the terminal node of the logical reasoning layer, and the formula is expressed as: i =w k ∈W D×K ; 1≤i≤K, where n i is the i-th terminal node of the logical reasoning layer, w k W D×K The kth weight vector in ; Calculate the inference node of the previous layer of the terminal node until ∑ j K(j)=1, the last n j It is the head end node of the logical reasoning layer; Copy the K-category classification function, randomly initialize the weights of the classification function, and construct the classification weight function. The formula is expressed as: 1≤k≤K, where h k,t+i is the probability weight of the meteorological type of the data sample at time (t+i) belonging to the kth category, is the classification weight function, H D is the input node of the regression layer, is the weight of the randomly initialized classification function, and [k] is the output node number of the classification function.

7. The method according to claim 6, characterized in that Calculate the inference node of the previous layer of the terminal node. The formula is expressed as: Where n j is the jth inference node in the previous layer of the terminal node, and K(j) is the number of terminal nodes under the jth inference node.

8. A deep learning model analysis system for renewable energy power generation output prediction, characterized in that: include: The model building and training module is used to build a new energy power generation output prediction model based on deep learning, and train the prediction model based on new energy power generation output data samples. The input features of the prediction model and data samples include the new energy power generation historical time series, meteorological historical time series, meteorological forecast and timestamp. The formula is expressed as follows: In the formula, is the output forecast value of renewable energy power generation from time t+1 to t+L; f pred is a deep learning model; x pv is the historical time series of renewable energy power generation, contained in the real number domain of dimension D×1 x w is the meteorological historical time series, contained in the real number domain range of dimension D×W x wp For weather forecast, it is contained in the real number domain of dimension L×W x ts is a timestamp, contained in the real number domain range of dimension L×S L is the length of the output forecast value; D is the length of the historical time series; W is the number of meteorological attributes; S is the number of timestamp dimensions; Input feature simplification module, used to simplify the input features of the trained prediction model and extract key influencing factors Among them, 1+W+S is the number of key influencing factors, is the range of the real number domain of dimension 1+W+S; The impact degree analysis module is used to calculate the contribution value of key impact factors using the Shapley value summation interpretation algorithm, and analyze the impact degree of key impact factors on the prediction results. The analytical formula is expressed as: In the formula, is the output forecast value of renewable energy power generation at time t+i, φ0 is the contribution bias value, x j is the jth key influencing factor, g i The key influencing factors Contribution value of The meteorological type classification module is used to classify the meteorological types of the renewable energy power generation output data samples based on the clear sky index index; The model output parsing module is used to transform the regression layer of the new energy power generation output prediction model into a logical reasoning layer, fine-tune the parameters of the logical reasoning layer of the prediction model based on the new energy power generation output data samples divided by meteorological types, and parse out the output logic of the prediction model. The parsing formula is expressed as: In the formula, C t+i is the meteorological type of the data sample at time t+i, p(C t+i = k) is the probability that the meteorological type of the data sample at time t+i belongs to the kth category, h k,t+i is the probability weight that the meteorological type of the data sample at time t+i belongs to the kth category.

9. A computer device, characterized in that: include: one or more processors; Memory; And one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and when the programs are executed by the processors, the steps of the deep learning model analysis method for new energy power generation output prediction as described in any one of claims 1-7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the deep learning model analysis method for new energy power generation output prediction as described in any one of claims 1 to 7 are implemented.

Images