Energy system operation optimization method and system based on photo-thermal energy storage
Through the multi-interval uncertainty set and CNN-MI-BILSTM model, combined with the photothermal energy storage system, the conservative problem of traditional photothermal systems in load uncertainty processing is solved, real-time stable control and economical when load fluctuations are achieved Sex optimization.
Patent Information
- Application Number
- CN202411007727.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-25
- Publication Date
- 2024-11-15
AI Technical Summary
Traditional photothermal systems have conservative problems in handling load uncertainty, resulting in a decrease in economic efficiency of optimized scheduling solutions and it is difficult to accurately quantify the impact of load uncertainty.
Multi-interval uncertainty set is used to describe load uncertainty, combined with the convolutional mutual information method, the load prediction is carried out, and the uncertain load model is constructed, and the photothermal energy storage effect is established. Run the optimization model.
The stability and economics of the system are improved, and the conservatism of optimization is improved through multi-interval uncertainty sets, real-time stable control and economical optimal output during load fluctuations are achieved.
Smart Images

Figure CN118966418A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of solar thermal power generation grid-connected technology, and specifically to an energy system operation optimization method and system based on solar thermal energy storage. Background Art
[0002] In the study of wind and solar thermal system operation optimization, the uncertainty of load is taken into account. This feature makes it different from traditional operation optimization. Load uncertainty is introduced into operation optimization, so that the unit output at any time should meet the load demand, improve system stability, and improve real-time control capabilities. However, how to accurately quantify uncertainty factors, traditional methods face severe challenges. How to consider the impact of strong uncertainty is an urgent problem to be solved.
[0003] At present, there are three main ways to model the uncertainty of loads and the uncertainty variables in distribution networks: probability model, fuzzy number model and interval number model. In the traditional single uncertainty set, the worst scenario obtained by optimization is usually taken at the boundary of the interval. However, the probability of actually obtaining this situation is extremely small, which makes the optimization scheduling scheme obtained using a single interval uncertainty set often too conservative, reducing the economic efficiency. Summary of the invention
[0004] In order to solve the above problems, the present invention proposes an energy system operation optimization method and system based on solar thermal energy storage, which can solve the problem of treating the load as non-volatile when load fluctuations occur, causing serious accidents. Multi-interval uncertainty sets are used to describe load uncertainty. The load prediction model is used to predict the load, which can not only ensure the real-time stability of the power system, but also improve conservatism and improve economy.
[0005] In order to achieve the above object, the present invention is implemented by the following technical solutions:
[0006] The energy system operation optimization method based on solar thermal energy storage of the present invention comprises the following operations:
[0007] Collect weather data and load data, build a load forecasting model based on the collected historical weather data and historical load data, and use the weather data, historical load data and load forecasting model to obtain load forecast values, and obtain forecast errors based on the load forecast values and the collected actual load data;
[0008] Based on the prediction error, the probability distribution function is used to construct the uncertainty set divided into multiple intervals. The uncertainty of the load is described by the uncertainty set, and the uncertain load model is obtained.
[0009] An operation optimization model is constructed based on the effect of solar thermal energy storage and the uncertain load obtained from the uncertain load model, and the system operation optimization strategy is obtained using the operation optimization model.
[0010] A further improvement of the present invention is that the weather data includes temperature and solar radiation.
[0011] A further improvement of the present invention is that the load forecasting model is a convolution mutual information method bidirectional long short-term memory neural network model CNN-MI-BILSTM model, and the specific operations of constructing the load forecasting model include:
[0012] Construct a convolutional long short-term memory neural network CNN-LSTM network. The CNN-LSTM network structure includes an input layer, a convolutional neural network layer, a long short-term memory neural network layer, and an output layer. The convolutional neural network layer is represented by a CNN layer, and the long short-term memory neural network layer is represented by an LSTM layer. Historical weather data and historical load data are input through the input layer.
[0013] Introduce the mutual information method to update the data output by CNN;
[0014] Improvements to the LSTM layer, including the introduction of a bias optimization parameter c t , forward and backward prediction based on LSTM layer.
[0015] A further improvement of the present invention is to introduce the mutual information method to update the data output by the CNN, and the specific operations include:
[0016] Determine the input feature data set and obtain the input feature matrix X corresponding to the load data set output by CNN t , input feature matrix X t The expression is:
[0017]
[0018] For the input feature matrix X t After normalization, we can obtain the importance value fluctuation matrix M of the input features that change over time, which is expressed as follows:
[0019]
[0020] Among them, o′ t is the input feature matrix X at time t C The output load data obtained, x t,z Represented as the input feature matrix X t The input data of the zth row and tth column in M(x t,z , o t ') is the correlation between the input data and the corresponding column output load data, which is represented by the M value. The stronger the correlation between the two variables, the larger the M value. When the two variables are independent of each other, the M value is 0;
[0021] Updated input feature matrix Xt The expression of ′ is:
[0022] X t ′=X t ×M.
[0023] A further improvement of the present invention is that the LSTM layer is improved, including introducing a bias optimization parameter c t , based on the LSTM layer forward and backward prediction, the specific operations include:
[0024] Using the updated input feature matrix X t 'Improve the LSTM layer and get:
[0025] f t =σ(W fx x t ′+W fh h t-1 +b f )
[0026] i t =σ(W ix x t ′+W ih h t-1 +b i )
[0027]
[0028] o t =Sigmoid(W ox x t ′+W oh h t-1 +b o )
[0029] Among them, x t ′ is the input of the forget gate, h t-1 is the intermediate output, S t-1 is the state memory unit, f t 、i t , h t , o t are the states of the forget gate, input gate, input node, intermediate output, state unit and output gate respectively, W fx , W ix , W gx , W ox They are respectively the forget gate, input gate, input node, output gate and forget gate input x t ′’s matrix weight, W fh , W ih , W gh , Woh They are respectively the forget gate, input gate, input node, output gate and intermediate output h t-1 The matrix weights to be multiplied, b f 、b i 、b g 、b o They are the bias items of the forget gate, input gate, input node, and output gate, respectively, indicating the bitwise multiplication of vector elements, and φ indicates the change of tanh function;
[0030] Introducing bias optimization parameter c t , and make forward and backward predictions:
[0031]
[0032] in, are the load prediction values of the forward and backward LSTM units at time t, respectively, t-1 , o t+1 are the load forecast values at time t-1 and t+1 respectively, For the forward and backward calculation of LSTM, a t 、b t are the forward and backward output weights respectively.
[0033] A further improvement of the present invention is that based on the prediction error, a probability distribution function is used to construct an uncertainty set divided into multiple intervals, and the load uncertainty is described by the uncertainty set to obtain an uncertain load model. The specific operations include:
[0034] Based on the prediction error, the probability distribution function is used to construct the uncertainty set divided into multiple intervals, the time period is and The total number of time periods divided between b intervals, the time of each interval Depending on the deviation and probability distribution of the corresponding interval, the expression of the uncertain load model is obtained as follows:
[0035]
[0036] in, For uncertain load power, is the load forecast value, are the upper and lower error limits corresponding to time period b, For variables, by introducing variables Limit the power of uncertain load to within the scope;
[0037] The constraints are:
[0038]
[0039] in, is the maximum deviation rate allowed between the deviation value and the true value, N b 、N t Represent the total number of intervals and the total time respectively.
[0040] A further improvement of the present invention is that an operation optimization model is constructed based on the solar thermal energy storage effect and the uncertain load obtained from the uncertain load model, and the expression is:
[0041]
[0042] in:
[0043]
[0044] P tes (t) = H tes (t)η1
[0045]
[0046] Where C′ is the total cost of day-ahead scheduling, C1′, C2′, and C3′ are daily operation costs, start-up and shutdown costs, and equipment maintenance costs, respectively. T is the scheduling period. h(C′)=0, g(C′)≤0 are the equality constraints and non-equality constraints satisfied by day-ahead scheduling. P w (t), P p (t), P pb (t), P tes (t) are the power of wind turbine, photovoltaic, solar thermal power generation system and solar thermal energy storage system at the tth moment, c w 、c p 、c pb are the unit operating costs of wind turbines, photovoltaic and solar thermal power generation systems, c tes is the unit operating cost of the CSP energy storage system, H tes (t) is the charge / discharge heat of the CSP system at time t, η1 is the thermal-electric conversion efficiency of the CSP system, C2 is the start-up and shutdown cost of wind power, photovoltaic power generation, and CSP power generation, SU W ,SU P ,SU PB are the startup costs of wind power, photovoltaic power generation, and solar thermal power generation, SD W , SD P , SD PB are the shutdown costs of wind power, photovoltaic power generation, and solar thermal power generation, are binary variables for wind turbine, photovoltaic, solar thermal power generation system, and solar thermal energy storage system started at time t, are binary variables indicating that the wind turbine, photovoltaic, solar thermal power generation system, and solar thermal energy storage system stop at time t. When the value is 1, the energy block starts at time t. When the value is 1, the energy block stops at time t, SU TES Start-up cost of heat-to-electric conversion for CSP storage, SD TES The startup cost of charging the CSP system, OM WT OM PV OM PB OM TES They are the maintenance cost coefficients of wind turbines, photovoltaic, solar thermal power generation systems, and solar thermal energy storage systems;
[0047] The constraints for running the optimization model are:
[0048]
[0049]
[0050] 0≤P g,t ≤P g,max
[0051]
[0052] in, are the rated power of wind power and photovoltaic units respectively, are the lower limits of the ramp rates of wind, photovoltaic and solar thermal power generation systems, is the upper limit of the ramp rate of wind, photovoltaic and solar thermal power generation systems. is the maximum output power of the CSP system, is the binary variable of the CSP system started at time t, P is the binary variable for starting and stopping each unit; g,t is the power of the tie line, P g,max is the maximum allowable power of the tie line, For uncertain load power, To charge and release heat for solar thermal energy storage, is a binary variable, The heat storage unit stores heat. is a binary variable, When the heat storage unit releases heat, are the maximum heat storage and heat release powers of the thermal energy unit, is the heat output of the CSP system at time t, H csp Absorb total heat for the concentrated solar collector system.
[0053] The energy system operation optimization system based on solar thermal energy storage of the present invention comprises:
[0054] The load forecasting module is used to collect weather data and load data, build a load forecasting model based on the collected historical weather data and historical load data, and obtain a load forecast value using the weather data, historical load data and the load forecasting model, and obtain a forecast error based on the load forecast value and the collected actual load data;
[0055] The uncertain load model building module is used to construct an uncertain set divided into multiple intervals based on the prediction error using the probability distribution function, and to describe the load uncertainty through the uncertain set to obtain the uncertain load model;
[0056] The operation optimization model module is used to construct an operation optimization model based on the effect of solar thermal energy storage and the uncertain load obtained from the uncertain load model, and use the operation optimization model to obtain the system operation optimization strategy.
[0057] A further improvement of the present invention is that the operations performed by the load forecasting module include:
[0058] Construct a convolutional long short-term memory neural network CNN-LSTM network. The CNN-LSTM network structure includes an input layer, a convolutional neural network layer, a long short-term memory neural network layer, and an output layer. The convolutional neural network layer is represented by a CNN layer, and the long short-term memory neural network layer is represented by an LSTM layer. Historical weather data and historical load data are input through the input layer.
[0059] Introduce the mutual information method to update the data output by the CNN layer;
[0060] Improvements to the LSTM layer, including the introduction of a bias optimization parameter c t , forward and backward prediction based on LSTM layer.
[0061] A further improvement of the present invention is to introduce the mutual information method to update the data output by the CNN, and the specific operations include:
[0062] Determine the input feature data set and obtain the input feature matrix X corresponding to the load data set output by CNN t , input feature matrix X t The expression is:
[0063]
[0064] For the input feature matrix X t After normalization, we can obtain the importance value fluctuation matrix M of the input features that change over time, which is expressed as follows:
[0065]
[0066] Among them, o′ t is the input feature matrix X at time t C The output load data obtained, x t,z Represented as the input feature matrix X t The input data of the zth row and tth column in M(x t,z , o′ t ) is the correlation between the input data and the corresponding column output load data, which is represented by the M value. The stronger the correlation between the two variables, the larger the M value. When the two variables are independent of each other, the M value is 0.
[0067] Updated input feature matrix X t The expression of ′ is:
[0068]
[0069] A further improvement of the present invention is that the LSTM layer is improved, including introducing a bias optimization parameter c t , based on the LSTM layer forward and backward prediction, the specific operations include:
[0070] Using the updated input feature matrix X t 'Improve the LSTM layer and get:
[0071] f t =σ(W fx x t ′+W fh h t-1 +b f )
[0072] i t =σ(W ix x t ′+W ih h t-1 +b i )
[0073]
[0074] o t =Sigmoid(W ox x t ′+W oh h t-1 +b o )
[0075] Among them, x t ′ is the input of the forget gate, h t-1 is the intermediate output, S t-1 is the state memory unit, f t 、i t , h t , o t are the states of the forget gate, input gate, input node, intermediate output, state unit and output gate respectively, W fx , W ix , W gx , W ox They are respectively the forget gate, input gate, input node, output gate and forget gate input x t ′’s matrix weight, W fh , W ih , W gh , W oh They are respectively the forget gate, input gate, input node, output gate and intermediate output h t-1 The matrix weights to be multiplied, b f 、b i 、b g 、b o They are the bias items of the forget gate, input gate, input node, and output gate, respectively, indicating the bitwise multiplication of vector elements, and φ indicates the change of tanh function;
[0076] Introducing bias optimization parameter c t , and make forward and backward predictions:
[0077]
[0078]
[0079] in, are the load prediction values of the forward and backward LSTM units at time t, respectively, t-1 , o t+1 are the load forecast values at time t-1 and t+1 respectively, For the forward and backward calculation of LSTM, a t 、b t are the forward and backward output weights respectively.
[0080] A further improvement of the present invention is that the specific operations of the uncertain load model building module include:
[0081] Based on historical forecast errors, the probability distribution function is used to construct an uncertainty set divided into multiple intervals, and the time period is and The total number of time periods divided between b intervals, the time of each interval Depending on the deviation and probability distribution of the corresponding interval, the expression of the uncertain load model is obtained as follows:
[0082]
[0083] in, is the load forecast value, are the upper and lower error limits corresponding to time period b, For variables, by introducing variables Limit the power of uncertain load to within the scope;
[0084] The constraints are:
[0085]
[0086] in, is the maximum deviation rate allowed between the deviation value and the true value, N b 、N t Represent the total number of intervals and the total time respectively.
[0087] A further improvement of the present invention is that an operation optimization model is constructed based on the solar thermal energy storage effect and the uncertain load obtained from the uncertain load model, and the expression is:
[0088]
[0089] in:
[0090]
[0091] P tes (t) = H tes (t)η1
[0092]
[0093] Where C′ is the total cost of day-ahead scheduling, C1′, C2′, and C3′ are daily operation costs, start-up and shutdown costs, and equipment maintenance costs, respectively. T is the scheduling period. h(C′)=0, g(C′)≤0 are the equality constraints and non-equality constraints satisfied by day-ahead scheduling. P w (t), P p (t), P pb (t), P tes (t) are the power of wind turbine, photovoltaic, solar thermal power generation system and solar thermal energy storage system at the tth moment, c w 、c p 、c pb are the unit operating costs of wind turbines, photovoltaic and solar thermal power generation systems, c tes is the unit operating cost of the CSP energy storage system, H tes (t) is the charge / discharge heat of the CSP system at time t, η1 is the thermal-electric conversion efficiency of the CSP system, C2 is the start-up and shutdown cost of wind power, photovoltaic and CSP power generation, SU W ,SUP ,SU PB are the startup costs of wind power, photovoltaic power generation, and solar thermal power generation, SD W , SD P , SD PB are the shutdown costs of wind power, photovoltaic power generation, and solar thermal power generation, are binary variables for wind turbine, photovoltaic, solar thermal power generation system, and solar thermal energy storage system started at time t, are binary variables indicating that the wind turbine, photovoltaic, solar thermal power generation system, and solar thermal energy storage system stop at time t. When the value is 1, the energy block starts at time t. When the value is 1, the energy block stops at time t, SU TES Start-up cost of heat-to-electric conversion for CSP storage, SD TES The startup cost of charging the CSP system, OM WT OM PV OM PB ,OM TES They are the maintenance cost coefficients of wind turbines, photovoltaic, solar thermal power generation systems, and solar thermal energy storage systems;
[0094] The constraints for running the optimization model are:
[0095]
[0096] 0≤P g,t ≤P g,max
[0097]
[0098] in, are the rated power of wind power and photovoltaic units respectively; are the lower limits of the ramp rates of wind, photovoltaic and solar thermal power generation systems, is the upper limit of the ramp rate of wind, photovoltaic and solar thermal power generation systems. is the maximum output power of the CSP system, is the binary variable of the CSP system started at time t, P is the binary variable for starting and stopping each unit; g,t is the power of the tie line, P g,max is the maximum allowable power of the tie line, For uncertain load power, To charge and release heat for solar thermal energy storage, is a binary variable, The heat storage unit stores heat. is a binary variable, When the heat storage unit releases heat, are the maximum heat storage and heat release powers of the thermal energy unit, is the output heat of the CSP system at time t; H csp Absorb total heat for the concentrated solar collector system.
[0099] A computer storage medium of the present invention stores a program for optimizing the operation of an energy system based on solar thermal energy storage, and when the program for optimizing the operation of an energy system based on solar thermal energy storage is executed by at least one processor, the steps of the above-mentioned method for optimizing the operation of an energy system based on solar thermal energy storage are implemented.
[0100] The beneficial effects of the present invention are as follows: the present invention introduces a multi-interval uncertainty set to describe load uncertainty, and improves the conservative problem of robust optimization according to upper and lower error limits and probability distribution.
[0101] Based on the CNN-LSTM model, the present invention not only considers the fluctuation of important values in the feature vector and introduces the mutual information method, but also improves the prediction efficiency by introducing the bidirectional method, which can realize the extraction of fluctuation of important values of data, and also realizes the bidirectional flow of data, realizes short-term load prediction, and mines the intrinsic connection between long-term data, which can avoid the adverse effects of non-continuous features on prediction accuracy and ensure prediction accuracy. The present invention considers the solar thermal energy storage system and uncertain load, establishes an operation optimization model, and optimizes the output power of each module including wind power generation, photovoltaic power generation, and solar thermal power generation to ensure optimal economic performance. BRIEF DESCRIPTION OF THE DRAWINGS
[0102] Figure 1 is a flow chart of the method of the present invention;
[0103] Figure 2 is a structural diagram of a load forecasting model in an embodiment of the present invention;
[0104] Figure 3 It is a multi-interval probability distribution diagram of the uncertain load in the embodiment of the present invention. DETAILED DESCRIPTION
[0105] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.
[0106] like Figure 1As shown, the energy system operation optimization method based on solar thermal energy storage in this embodiment establishes an operation optimization model considering load uncertainty and solar thermal energy storage for a comprehensive energy system composed of wind power generation, photovoltaic power generation, and solar thermal power generation, and includes the following operation steps:
[0107] Step 1: Collect weather data and load data, build a load forecasting model, and use weather data, historical load data and load forecasting model to forecast load data. The load forecasting model is a convolutional mutual information bidirectional long short-term memory neural network model, namely the CNN-MI-BILSTM model, with the following structure: Figure 2 As shown in the figure, the mutual information method MI and the bidirectional method BI are added to the convolutional long short-term memory neural network, namely the CNN-LSTM network. The CNN network (CNN layer) realizes the preprocessing of the data, while the MI-BILSTM network realizes the accurate prediction of the load.
[0108] The CNN-LSTM network structure consists of an input layer, a CNN layer, an LSTM layer, and an output layer. The historical weather data and historical load data are input, and the dynamic change information of the historical load data is obtained through the CNN layer, useful features are extracted, and the output data feature vector is time-series, which is then sent to the LSTM layer to learn the change rules within the feature vector output by the CNN layer, capture the time dependency relationship between the data, and finally output the load forecast data. The specific operations of constructing the CNN layer and the LSTM layer are as follows:
[0109] CNN layer: mainly includes convolution layer and pooling layer, which uses the convolution layer to extract static features of input data through sliding window operation of convolution kernel, and then uses the scale invariance of key features through pooling layer to reduce the dimension of extracted features, so that key features are further highlighted, and the complexity of the network is reduced through parameter sharing. Through convolution and pooling layers, key features are extracted from input data, and the dimension of features can be reduced;
[0110] The process of CNN layer is expressed as:
[0111]
[0112] P1=max(C1)+b2 (2)
[0113]
[0114] P2=max(C2)+b4 (4)
[0115] X C =Sigmoid(P2×W3+b5) (5)
[0116] Among them, C1, C2 and P1, P2 are the outputs of convolution layer 1, convolution layer 2 and pooling layer 1, pooling layer 2 respectively, W1, W2, W3 are the weight matrices of each CNN layer, b1, b2, b3, b4 and b5 are deviation values, For convolution operation, the activation function sigmoid is used in the fully connected layer.
[0117] The output feature vector of the CNN layer is X c , the output length of the CNN layer is t, and the output feature vector X c Note: X C =[x1…x i-1 ,x i …,x t ] T .
[0118] The basic unit model of the LSTM layer is as follows:
[0119] f t =σ(W fx x t +W fh h t-1 +b f ) (6)
[0120] i t =σ(W ix x t +W ih h t-1 +b i ) (7)
[0121] g t =φ(W gx x t +W gh h t-1 +b g ) (8)
[0122] S t =g t i t +S t-1 f t (9)
[0123] h t =φ(S t ) t (10)
[0124] o t =Sigmoid(W ox x t +W oh h t-1 +b o ) (11)
[0125] Among them, x t is the input of the forget gate, h t-1 is the intermediate output, S t-1 is the state memory unit, f t 、i t , g t 、h t , S t , o t are the states of the forget gate, input gate, input node, intermediate output, state unit and output gate respectively, W fx , W ix , W gx , W ox They are respectively the forget gate, input gate, input node, output gate and forget gate input x t ′, the matrix weight, W fh , W ih , W gh , W oh They are respectively the forget gate, input gate, input node, output gate and intermediate output h t-1 The matrix weights to be multiplied, b f , b i , b g , b o They are the bias items of the forget gate, input gate, input node, and output gate respectively, indicating the bitwise multiplication of vector elements, and φ indicates the change of tanh function.
[0126] Output layer: The output data of the LSTM layer output gate is used as the load forecast data result:
[0127]
[0128] in, Load forecast data.
[0129] The mutual information method is introduced to update the data output by the CNN layer:
[0130] Determine the input feature data set and obtain the input feature matrix X corresponding to the load data set output by the above CNN layer t :
[0131]
[0132] For the input feature matrix X t After normalization, we can obtain the importance value fluctuation matrix M of the input features that change over time:
[0133]
[0134] Among them, o′ tis the input feature matrix X at time t C The output load data obtained, x t,z Represented as the input feature matrix X t The input data of the zth row and tth column in M(x t,z , o′ t ) is the correlation between the input data and the corresponding column output load data, which is represented by the M value. The stronger the correlation between the two variables, the larger the M value. When the two variables are independent of each other, the M value is 0.
[0135] The importance value fluctuation matrix M of the input features that change over time contains the importance information of the input features in different dimensions. The horizontal time dimension comparison can reflect the importance value fluctuation of the same type of input features at different times, while the vertical type dimension comparison can reflect the importance difference of different types of input features at the same time.
[0136] X t ′=X t ×M (15)
[0137] Among them, X t ′ is the updated input feature matrix.
[0138] By improving expressions (6)-(11), we can obtain:
[0139] f t =σ(W fx x t ′+W fh h t-1 +b f ) (16)
[0140] i t =σ(W ix x t ′+W ih h t-1 +b i ) (17)
[0141]
[0142] o t =Sigmoid(W ox x t ′+W oh h t-1 +b o ) (twenty one)
[0143] Among them, x t ′ is the input of the forget gate, They are the states of input node and state unit respectively.
[0144] Introducing bias optimization parameter c t , the bidirectional long short-term memory neural network model, namely the BILSTM model, performs forward and backward predictions based on the LSTM layer:
[0145]
[0146] in, are the load prediction values of the forward and backward LSTM units at time t, respectively, t-1 , o t+1 are the load forecast values at time t-1 and t+1 respectively, For the forward and backward calculation of LSTM, a t 、b t are the forward and backward output weights respectively.
[0147] Step 2: Describe the load uncertainty through the uncertainty set to obtain the uncertain load model. That is, describe the load uncertainty through the uncertainty set, and obtain the probability of each interval period and the upper and lower limit errors based on the historical load data. At the same time, take the load forecast value obtained by the load forecasting model as input to obtain the load uncertainty model. Specifically include:
[0148] This embodiment uses a multi-interval-divided uncertainty set to further consider the actual distribution of load power under the condition of ensuring uncertainty, improving the conservative problem of the traditional single uncertainty set. This embodiment uses the prediction error based on historical load data and uses the probability distribution function to construct a multi-interval-divided uncertainty set. is and The total number of time periods divided between b intervals, the time of each interval Depends on the deviation ratio of the corresponding interval and the probability distribution coefficient Deviation and probability distribution can be obtained from historical load data, and multi-interval probability distribution diagrams can be drawn, such as Figure 3 shown.
[0149]
[0150] in, are the upper and lower limits of the prediction deviation in interval b, are the upper and lower limits of the prediction deviation corresponding to a single interval, Expressed as the probability coefficient of interval b, is the probability coefficient of interval b+1, expressions (25)-(26) can be realized by A single interval is divided according to the upper and lower error limits and the probability distribution.
[0151] The uncertain load model is expressed as:
[0152]
[0153] in, For uncertain load power, is the load forecast value, are the upper and lower error limits corresponding to time period b, For variables, by introducing variables Limit the power of uncertain load to within the range.
[0154] The constraints are:
[0155]
[0156] in, is the maximum deviation rate allowed between the deviation value and the true value. Expression (28) assumes that only one deviation value will appear in any period from a spatial perspective. Expression (29) ensures that the total uncertainty time period of each interval does not exceed Expression (30) controls the deviation range between the actual output and the load forecast value.
[0157] Step 3: construct an operation optimization model based on the solar thermal energy storage function and the uncertain load obtained from the uncertain load model, and use the operation optimization model to obtain the system operation optimization strategy. Specifically include:
[0158] Based on the operating relationship between independent thermoelectric conversion of solar thermal energy storage and solar thermal power generation system, a mathematical model for the operation of solar thermal modules is constructed.
[0159] The energy for solar thermal power generation and energy storage charging comes from the concentrated solar collector system, so the concentrated solar collector system is modeled first. The CSP power station uses the concentrated solar collector system to convert the reflected light energy of the mirror field into thermal energy. The resulting thermal power is as follows:
[0160] P csp (t) = η2H csp =η2SD t (31)
[0161] Among them, P csp (t) is the thermal power of the concentrating and collecting device at time t, η2 is the photothermal conversion efficiency, H csp is the total heat absorbed by the concentrating solar collector system, S is the mirror area of the CSP power station, and D t is the direct radiation index of light at time t.
[0162] The concentrated solar collector can not only convert the heat obtained to generate electricity directly when there is a load demand, but also store energy when the load demand is low:
[0163] P csp (t) = P pb (t)+P tes (t) (32)
[0164] Among them, P pb (t) is the power of the CSP system at the tth moment, P tes (t) is the power of the solar thermal energy storage system at the tth moment.
[0165] Solar thermal power generation is defined as energy transferred only by a concentrated solar collector system and is modeled as follows:
[0166] P pb (t) = η d P csp (t) (33)
[0167] Among them, η d is the thermoelectric conversion efficiency.
[0168] The solar thermal energy storage system releases heat when the load demand is high and absorbs heat when the load demand is low. Both heat storage and heat release will result in heat loss. The characteristics of the two are expressed as follows:
[0169]
[0170] in, and is the heat storage and release power of TES, P HT (t) represents the heat transfer power at time t, η c and η f are TES heat charging and heat release loss rates respectively.
[0171] The solar thermal energy storage system stores heat and generates heat energy loss, which can be expressed as:
[0172]
[0173] When the load demand is high, the stored heat is converted into electrical energy supply:
[0174]
[0175] Based on the role of solar thermal energy storage and uncertain load, an operation optimization model is constructed. The expression of the operation optimization model is:
[0176]
[0177] in:
[0178]
[0179] P tes (t) = H tes(t)η1
[0180]
[0181] Among them, C′ is the total cost of day-ahead scheduling, C1′, C2′, and C3′ are daily operation costs, start-stop costs, and equipment maintenance costs, respectively. T is the scheduling period, i.e., 24 h. h(C′)=0, g(C′)≤0 are the equality constraints and non-equality constraints satisfied by day-ahead scheduling, i.e., the constraints of the operation optimization model. P w (t), P p (t), P pb (t), P tes (t) are the power of wind turbine, photovoltaic, solar thermal power generation system and solar thermal energy storage system at the tth moment, c w 、c p 、c pb are the unit operating costs of wind turbines, photovoltaic and solar thermal power generation systems, c tes is the unit operating cost of the CSP energy storage system, H tes (t) is the charge / discharge heat of the CSP system at time t, η1 is the thermal-electric conversion efficiency of the CSP system, C2 represents the start-up and shutdown costs of wind power, photovoltaic power generation, and CSP power generation, SU W ,SU P ,SU PB are the startup costs of wind power, photovoltaic power generation, and solar thermal power generation, SD W , SD P , SD PB are the shutdown costs of wind power, photovoltaic power generation, and solar thermal power generation, are binary variables for wind turbine, photovoltaic, solar thermal power generation system, and solar thermal energy storage system started at time t, are binary variables indicating that the wind turbine, photovoltaic, solar thermal power generation system, and solar thermal energy storage system stop at time t. When the value is 1, the energy block starts at time t. When the value is 1, the energy block stops at time t, SU TES Start-up cost of heat-to-electric conversion for CSP storage, SD TES The startup cost of charging the CSP system, OM WT OM PV OM PB OM TES They are the maintenance cost coefficients of wind turbines, photovoltaics, solar thermal power generation systems, and solar thermal energy storage systems respectively.
[0182] The constraints for running the optimization model are:
[0183]
[0184] 0≤P g,t ≤P g,max
[0185]
[0186]
[0187] in, are the rated power of wind power and photovoltaic units respectively, are the lower limits of the ramp rates of wind, photovoltaic and solar thermal power generation systems, is the upper limit of the ramp rate of wind, photovoltaic and solar thermal power generation systems. is the maximum output power of the CSP system, is the binary variable of the CSP system started at time t, is the binary variable for starting and stopping each unit, P g,t is the power of the tie line, P g,max is the maximum allowable power of the tie line, For uncertain load power, To charge and release heat for solar thermal energy storage, is a binary variable, The heat storage unit stores heat. is a binary variable, When the heat storage unit releases heat, are the maximum heat storage and heat release powers of the thermal energy unit, is the output heat of the CSP system at time t; H csp Absorb total heat for the concentrated solar collector system.
[0188] The operation optimization model of this embodiment not only realizes the prediction and control of load fluctuations, but also uses the solar thermal energy storage system as an energy storage module to optimize the operation of each unit, thereby enhancing the reliability and safety of the operation of the solar thermal power generation system.
[0189] The above method is implemented by the energy system operation optimization system based on solar thermal energy storage of this embodiment, and the system includes:
[0190] The load forecasting module is used to collect weather data and load data, build a load forecasting model, and use weather data, historical load data and load forecasting models to forecast load data;
[0191] An uncertain load model building module is used to describe load uncertainty through multi-interval uncertainty sets to obtain an uncertain load model;
[0192] The operation optimization model module is used to build an operation optimization model based on the effect of solar thermal energy storage and uncertain load.
[0193] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as those generally understood by those skilled in the art in the art to which the present invention belongs. It should also be understood that terms such as those defined in common dictionaries should be understood to have the same meaning as in the context of the prior art, and will not be interpreted with an idealized or overly formal meaning unless defined similarly here.
[0194] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. An energy system operation optimization method based on solar thermal energy storage, characterized by: The following operations are included: Collect weather data and load data, build a load forecasting model based on the collected historical weather data and historical load data, and use the weather data, historical load data and load forecasting model to obtain load forecast values, and obtain forecast errors based on the load forecast values and the collected actual load data; Based on the prediction error, the probability distribution function is used to construct the uncertainty set divided into multiple intervals. The uncertainty of the load is described by the uncertainty set, and the uncertain load model is obtained. An operation optimization model is constructed based on the effect of solar thermal energy storage and the uncertain load obtained from the uncertain load model, and the system operation optimization strategy is obtained using the operation optimization model.
2. The energy system operation optimization method based on solar thermal energy storage according to claim 1 is characterized in that: The weather data includes temperature and solar radiation.
3. The energy system operation optimization method based on solar thermal energy storage according to claim 1 is characterized in that: The load forecasting model is a convolution mutual information method bidirectional long short-term memory neural network model CNN-MI-BILSTM model. The specific operation of constructing the load forecasting model includes: constructing a convolution long short-term memory neural network CNN-LSTM network, the CNN-LSTM network structure includes an input layer, a convolution neural network layer, a long short-term memory neural network layer and an output layer, wherein the convolution neural network layer is represented by a CNN layer, the long short-term memory neural network layer is represented by an LSTM layer, and historical weather data and historical load data are input through the input layer; Introduce the mutual information method to update the data output by the CNN layer; Improvements to the LSTM layer, including the introduction of a bias optimization parameter c t , forward and backward prediction based on LSTM layer.
4. The energy system operation optimization method based on solar thermal energy storage according to claim 3 is characterized in that: The mutual information method is introduced to update the data output by the CNN layer, and the specific operations include: Determine the input feature data set and obtain the input feature matrix X corresponding to the load data set output by the CNN layer t , input feature matrix X t The expression is: For the input feature matrix X t After normalization, we can obtain the importance value fluctuation matrix M of the input features that change over time, which is expressed as follows: Among them, o' t is the input feature matrix X at time t C The output load data obtained, x t,z Represented as the input feature matrix X t The input data of the zth row and tth column in M(x t,z , o' t ) is the correlation between the input data and the corresponding column output load data, which is represented by the M value. The stronger the correlation between the two variables, the larger the M value. When the two variables are independent of each other, the M value is 0. Updated input feature matrix X t The expression of ′ is: X t ′=X t ×M。 5. The energy system operation optimization method based on solar thermal energy storage according to claim 4 is characterized in that: The improvements to the LSTM layer include the introduction of a bias optimization parameter c t , based on the LSTM layer forward and backward prediction, the specific operations include: Using the updated input feature matrix X t 'Improve the LSTM layer and get: f t =σ(W fx x t ′+W fh h t-1 +b f ) i t =σ(W ix x t ′+W ih h t-1 +b i ) o t =Sigmoid(W ox x t ′+W oh h t-1 +b o ) Among them, x t ′ is the input of the forget gate, h t-1 is the intermediate output, S t-1 is the state memory unit, f t 、i t , h t , o t are the states of the forget gate, input gate, input node, intermediate output, state unit and output gate respectively, W fx , W ix , W gx , W ox They are respectively the forget gate, input gate, input node, output gate and forget gate input x t ′, the matrix weight, W fh , W ih , W gh , W oh They are respectively the forget gate, input gate, input node, output gate and intermediate output h t-1 The matrix weights to be multiplied, b f , b i , b g , b o They are the bias items of the forget gate, input gate, input node, and output gate, respectively, indicating the bitwise multiplication of vector elements, and φ indicates the change of tanh function; Introducing bias optimization parameter c t , and make forward and backward predictions: in, are the load prediction values of the forward and backward LSTM units at time t, respectively, t-1 , o t+1 are the load forecast values at time t-1 and t+1 respectively, For the forward and backward calculation of LSTM, a t 、b t are the forward and backward output weights respectively.
6. The energy system operation optimization method based on solar thermal energy storage according to claim 1 is characterized in that: Based on the prediction error, the uncertainty set divided into multiple intervals is constructed by using the probability distribution function, and the load uncertainty is described by the uncertainty set to obtain the uncertain load model. The specific operations include: Based on the prediction error, the probability distribution function is used to construct the uncertainty set divided into multiple intervals, the time period is and The total number of time periods divided between b intervals, the time of each interval Depending on the deviation and probability distribution of the corresponding interval, the expression of the uncertain load model is obtained as follows: in, For uncertain load power, is the load forecast value, are the upper and lower error limits corresponding to time period b, For variables, by introducing variables Limit the power of uncertain load to within the scope; The constraints are: in, is the maximum deviation rate allowed between the deviation value and the true value, N b 、N t Represent the total number of intervals and the total time respectively.
7. The energy system operation optimization method based on solar thermal energy storage according to claim 1 is characterized in that: The operation optimization model based on the solar thermal energy storage effect and the uncertain load obtained from the uncertain load model is constructed and expressed as: in: P tes (t)=H tes (t)η1 Where C′ is the total cost of day-ahead scheduling, C1′, C2′, and C3′ are daily operation costs, start-up and shutdown costs, and equipment maintenance costs, respectively. T is the scheduling period. h(C′)=0, g(C′)≤0 are the equality constraints and non-equality constraints satisfied by day-ahead scheduling. P w (t), P p (t), P pb (t), P tes (t) are the power of wind turbine, photovoltaic, solar thermal power generation system and solar thermal energy storage system at the tth moment, c w 、c p 、c pb are the unit operating costs of wind turbines, photovoltaic and solar thermal power generation systems, c tes is the unit operating cost of the CSP energy storage system, H tes (t) is the charge / discharge heat of the CSP system at time t, η1 is the thermal-electric conversion efficiency of the CSP system, C2 is the start-up and shutdown cost of wind power, photovoltaic power generation, and CSP power generation, SU W ,SU P ,SU PB are the startup costs of wind power, photovoltaic power generation, and solar thermal power generation, SD W , SD P , SD PB are the shutdown costs of wind power, photovoltaic power generation, and solar thermal power generation, are binary variables for wind turbine, photovoltaic, solar thermal power generation system, and solar thermal energy storage system started at time t, are binary variables indicating that the wind turbine, photovoltaic, solar thermal power generation system, and solar thermal energy storage system stop at time t. When the value is 1, the energy block starts at time t. When the value is 1, the energy block stops at time t, SU TES Start-up cost of heat-to-electric conversion for CSP storage, SD TES The startup cost of charging the CSP system, OM WT OM PV ,OM PB ,OM TES They are the maintenance cost coefficients of wind turbines, photovoltaic, solar thermal power generation systems, and solar thermal energy storage systems; The constraints for running the optimization model are: 0≤P g,t ≤P g,max in, are the rated power of wind power and photovoltaic units respectively, are the lower limits of the ramp rates of wind, photovoltaic and solar thermal power generation systems, is the upper limit of the ramp rate of wind, photovoltaic and solar thermal power generation systems. is the maximum output power of the CSP system, is the binary variable of the CSP system started at time t, is the binary variable for starting and stopping each unit, P g,t is the power of the tie line, P g,max is the maximum allowable power of the tie line, For uncertain load power, To charge and release heat for solar thermal energy storage, is a binary variable, The heat storage unit stores heat. is a binary variable, When the heat storage unit releases heat, are the maximum heat storage and heat release powers of the thermal energy unit, is the heat output of the CSP system at time t, H csp Absorb total heat for the concentrated solar collector system.
8. Energy system operation optimization system based on solar thermal energy storage, characterized by: The system comprises: The load forecasting module is used to collect weather data and load data, build a load forecasting model based on the collected historical weather data and historical load data, and obtain a load forecast value using the weather data, historical load data and the load forecasting model, and obtain a forecast error based on the load forecast value and the collected actual load data; The uncertain load model building module is used to construct an uncertain set divided into multiple intervals based on the prediction error using the probability distribution function, and to describe the load uncertainty through the uncertain set to obtain the uncertain load model; The operation optimization model module is used to construct an operation optimization model based on the effect of solar thermal energy storage and the uncertain load obtained from the uncertain load model, and use the operation optimization model to obtain the system operation optimization strategy.
9. The energy system operation optimization system based on solar thermal energy storage according to claim 8 is characterized in that: The operations performed by the load forecasting module include: Construct a convolutional long short-term memory neural network CNN-LSTM network. The CNN-LSTM network structure includes an input layer, a convolutional neural network layer, a long short-term memory neural network layer, and an output layer. The convolutional neural network layer is represented by a CNN layer, and the long short-term memory neural network layer is represented by an LSTM layer. Historical weather data and historical load data are input through the input layer. Introduce the mutual information method to update the data output by the CNN layer; Improvements to the LSTM layer, including the introduction of a bias optimization parameter c t , forward and backward prediction based on LSTM layer.
10. The energy system operation optimization system based on solar thermal energy storage according to claim 8, characterized in that: The mutual information method is introduced to update the data output by CNN, and the specific operations include: Determine the input feature data set and obtain the input feature matrix X corresponding to the load data set output by the CNN layer t , input feature matrix X t The expression is: For the input feature matrix X t After normalization, we can obtain the importance value fluctuation matrix M of the input features that change over time, which is expressed as follows: Among them, o' t is the input feature matrix X at time t C The output load data obtained, x t,z Represented as the input feature matrix X t The input data of the zth row and tth column in M(x t,z , o' t ) is the correlation between the input data and the corresponding column output load data, which is represented by the M value. The stronger the correlation between the two variables, the larger the M value. When the two variables are independent of each other, the M value is 0. Updated input feature matrix X t The expression of ′ is: X t ′=X t ×M。 11. The energy system operation optimization system based on solar thermal energy storage according to claim 8, characterized in that: The improvements to the LSTM layer include the introduction of a bias optimization parameter c t , based on the LSTM layer forward and backward prediction, the specific operations include: Using the updated input feature matrix X t 'Improve the LSTM layer and get: f t =σ(W fx x t ′+W fh h t-1 +b f ) i t =σ(W ix x t ′+W ih h t-1 +b i ) o t =Sigmoid(W ox x t ′+W oh h t-1 +b o ) Among them, x t ′ is the input of the forget gate, h t-1 is the intermediate output, S t-1 is the state memory unit, f t 、i t , h t , o t are the states of the forget gate, input gate, input node, intermediate output, state unit and output gate respectively, W fx , W ix , W gx , W ox They are respectively the forget gate, input gate, input node, output gate and forget gate input x t ′, the matrix weight, W fh , W ih , W gh , W oh They are respectively the forget gate, input gate, input node, output gate and intermediate output h t-1 The matrix weights to be multiplied, b f 、b i 、b g 、b o They are the bias items of the forget gate, input gate, input node, and output gate, respectively, indicating the bitwise multiplication of vector elements, and φ indicates the change of tanh function; Introducing bias optimization parameter c t , and make forward and backward predictions: in, are the load prediction values of the forward and backward LSTM units at time t, respectively, t-1 , o t+1 are the load forecast values at time t-1 and t+1 respectively, For the forward and backward calculation of LSTM, a t 、b t are the forward and backward output weights respectively.
12. The energy system operation optimization system based on solar thermal energy storage according to claim 8, characterized in that: The specific operations of the uncertain load model building module include: Based on the prediction error, the probability distribution function is used to construct the uncertainty set divided into multiple intervals, the time period is and The total number of time periods divided between b intervals, the time of each interval Depending on the deviation and probability distribution of the corresponding interval, the expression of the uncertain load model is obtained as follows: in, is the load forecast value, are the upper and lower error limits corresponding to time period b, For variables, by introducing variables Limit the power of uncertain load to within the scope; The constraints are: in, is the maximum deviation rate allowed between the deviation value and the true value, N b 、N t Represent the total number of intervals and the total time respectively.
13. The energy system operation optimization system based on solar thermal energy storage according to claim 8, characterized in that: The operation optimization model based on the solar thermal energy storage effect and the uncertain load obtained from the uncertain load model is constructed and expressed as: in: P tes (t)=H tes (t)η1 Where C′ is the total cost of day-ahead scheduling, C1′, C2′, and C3′ are daily operation costs, start-up and shutdown costs, and equipment maintenance costs, respectively. T is the scheduling period. h(C′)=0, g(C′)≤0 are the equality constraints and non-equality constraints satisfied by day-ahead scheduling. P w (t), P p (t), P pb (t), P tes (t) are the power of wind turbine, photovoltaic, solar thermal power generation system and solar thermal energy storage system at the tth moment, c w 、c p 、c pb are the unit operating costs of wind turbines, photovoltaic and solar thermal power generation systems, c tes is the unit operating cost of the CSP energy storage system, H tes (t) is the charge / discharge heat of the CSP system at time t, η1 is the thermal-electric conversion efficiency of the CSP system, C2 is the start-up and shutdown cost of wind power, photovoltaic power generation, and CSP power generation, SU W ,SU P ,SU PB are the startup costs of wind power, photovoltaic power generation, and solar thermal power generation, SD W , SD P , SD PB are the shutdown costs of wind power, photovoltaic power generation, and solar thermal power generation, are binary variables for wind turbine, photovoltaic, solar thermal power generation system, and solar thermal energy storage system started at time t, are binary variables indicating that the wind turbine, photovoltaic, solar thermal power generation system, and solar thermal energy storage system stop at time t. When the value is 1, the energy block starts at time t. When the value is 1, the energy block stops at time t, SU TES Start-up cost of heat-to-electric conversion for CSP storage, SD TES The startup cost of charging the CSP system, OM WT ,OM PV ,OM PB ,OM TES They are the maintenance cost coefficients of wind turbines, photovoltaic, solar thermal power generation systems, and solar thermal energy storage systems; The constraints for running the optimization model are: 0≤P g,t ≤P g,max in, are the rated power of wind power and photovoltaic units respectively; are the lower limits of the ramp rates of wind, photovoltaic and solar thermal power generation systems, is the upper limit of the ramp rate of wind, photovoltaic and solar thermal power generation systems. is the maximum output power of the CSP system, is the binary variable of the CSP system started at time t, P is the binary variable for starting and stopping each unit; g,t is the power of the tie line, P g,max is the maximum allowable power of the tie line, For uncertain load power, To charge and release heat for solar thermal energy storage, is a binary variable, The heat storage unit stores heat. is a binary variable, When the heat storage unit releases heat, are the maximum heat storage and heat release powers of the thermal energy unit, is the output heat of the CSP system at time t; H csp Absorb total heat for the concentrated solar collector system.
14. A computer storage medium storing a program for optimizing the operation of an energy system based on solar thermal energy storage, wherein the program for optimizing the operation of an energy system based on solar thermal energy storage, when executed by at least one processor, implements the steps of the method for optimizing the operation of an energy system based on solar thermal energy storage at any time as claimed in any one of claims 1 to 7.
