An integrated energy system operation optimization method based on intelligent group optimization engine

Abstract

The present invention discloses an integrated energy system operation optimization method based on an intelligent group optimization engine, which belongs to the technical field of energy system optimization, and includes: S1: based on the selected kernel function, calculating the similarity between different data in the historical operation data of the integrated energy system, and obtaining linearly separable data based on the similarity; S2: constructing a linear activation function in a perceptron neural network in the intelligent group optimization engine based on the prediction variables in the predicted demand; S3: constructing an optimization objective function; S4: obtaining a demand forecast value in the optimization objective function based on the linear activation function; S5: based on a particle swarm optimization algorithm in the intelligent group optimization engine, using the demand forecast value to obtain the optimal control strategy of the energy conversion equipment, and optimizing the system operation. The method solves the problem that the perceptron neural network is limited to processing linearly separable data and limited to outputting discrete data, and overcomes the problem in the prior art that it is difficult to ensure accuracy while having high real-time performance.

Description

Technical Field

[0001] The present invention relates to the technical field of energy system optimization, and in particular to a method for optimizing the operation of an integrated energy system based on an intelligent group optimization engine. Background Art

[0002] The integrated energy system realizes the conversion, storage and distribution of energy by integrating various energy resources. In daily life, in order to meet the needs of users, it is necessary to adjust the output power of energy conversion equipment, especially when encountering extreme weather, the internal cooling load or heat load will increase sharply, or when there is a sudden large-scale change in the demand for electricity, heat or cooling energy, the system needs to quickly adjust the control strategy of the energy conversion equipment, so that the real-time requirements of system control in these cases are very high, in order to maintain a high degree of stability, ensure that the energy supply is not interrupted, and avoid losses or safety accidents. Although the commonly used deep neural network has high accuracy, due to its multi-layer structure, as the number of layers increases, the parameters and calculation amount in the network increase sharply, so its calculation efficiency is low, and it is difficult to meet the requirements of real-time and high stability under extreme conditions. Although the perceptron neural network has high calculation efficiency and can meet the real-time requirements, it is mainly used to process linearly separable data and outputs discrete data, which is not suitable for integrated energy systems. Therefore, it is a technical problem that is difficult to solve at present to ensure accuracy while having high real-time performance. Summary of the invention

[0003] In view of the problem that the prior art is difficult to ensure accuracy while having high real-time performance, the present invention provides an integrated energy system operation optimization method based on an intelligent group optimization engine. The similarity between different data in the historical operation data is obtained through a kernel function, so that the linearly inseparable historical operation data is linearly separable in a high-dimensional space. The predictive variables in the forecast demand are used to construct a linear activation function in the perceptron neural network in the intelligent group optimization engine so that the perceptron neural network outputs discrete data, which solves the problem that the perceptron neural network is limited to processing linearly separable data and limited to outputting discrete data. Therefore, the present invention overcomes the problem that the prior art is difficult to ensure accuracy while having high real-time performance.

[0004] In order to solve the above technical problems, the present invention provides an integrated energy system operation optimization method based on an intelligent group optimization engine, comprising the following steps:

[0005] S1: selecting a kernel function based on the historical operation data characteristics and predicted demand of the integrated energy system, calculating the similarity between different data in the historical operation data of the integrated energy system based on the selected kernel function, and obtaining linearly separable data based on the similarity;

[0006] S2: Build a smart group optimization engine based on the perceptron neural network and particle swarm optimization algorithm, and build a linear activation function in the perceptron neural network in the smart group optimization engine based on the prediction variables in the predicted demand;

[0007] S3: Construct an optimization objective function using the purchased energy price, the predicted value in the forecast demand, and the user's overall load;

[0008] S4: inputting the real-time acquired operation data into the perceptron neural network in the intelligent group optimization engine, and obtaining the demand forecast value in the optimization objective function based on the linear activation function;

[0009] S5: Based on the particle swarm optimization algorithm in the intelligent group optimization engine, the demand forecast value in the optimization objective function is used to obtain the optimal control strategy of the energy conversion equipment, and then optimize the operation of the integrated energy system.

[0010] After adopting the above technical solution, the present invention has the following advantages:

[0011] Taking into account the characteristics of the perceptron neural network, a kernel function is used to obtain the similarity between different data in the historical operation data, so that the linearly inseparable historical operation data can be linearly separable in a high-dimensional space, and the nonlinear historical operation data is converted into linear data that can be used by the perceptron neural network. At the same time, the prediction variables in the prediction requirements are used to construct the linear activation function in the perceptron neural network in the intelligent group optimization engine, so that the perceptron neural network outputs continuous data that can be used for the optimization objective function, which solves the problem that the perceptron neural network is limited to processing linearly separable data and outputting discrete data. Combined with the characteristics of the perceptron neural network with high computational efficiency and the ability to meet real-time requirements, the present invention achieves the effect of taking into account both accuracy and high real-time performance.

[0012] By using the particle swarm optimization algorithm to continuously optimize the control strategy of energy conversion equipment, the integrated energy system can more accurately match the user's load demand, reduce the deviation and imbalance between supply and demand, thereby improving the stability and reliability of the integrated energy system and improving resource utilization;

[0013] The combined use of kernel functions, linear activation functions, and particle swarm optimization algorithms can more accurately predict and respond to changes in load demand while balancing accuracy and high real-time performance. While ensuring real-time performance, it significantly improves accuracy in many aspects, thereby significantly improving the stability and reliability of the integrated energy system.

[0014] Preferably, in S1, the kernel function selected based on the historical operation data characteristics and predicted demand of the integrated energy system includes:

[0015] Clean the historical operation data of the integrated energy system, remove outliers, remove missing values ​​and normalize them;

[0016] The data obtained at the same time point in the historical operation data is analyzed with the predicted demand to obtain local features of the data, the data of the same category in the historical operation data is analyzed with the predicted demand to obtain global features of the data, and a kernel function is selected based on the local features and global features of the data.

[0017] In this scheme, by analyzing the data obtained at the same time point and the predicted demand, the dependency between different categories of data and the impact of the dependency on the predicted demand can be discovered. By analyzing the data and the predicted demand of the same category, the law of change of the same category of data over time and the impact of the law on the predicted demand can be discovered. The kernel function is selected based on the dependency and law, so that the kernel function can better adapt to different data distributions and characteristics, thereby improving the accuracy of obtaining linearly separable data. At the same time, by selecting a suitable kernel function, the calculation efficiency is further improved, thereby improving real-time performance.

[0018] Preferably, in S1, the calculating the similarity between different data in the historical operation data of the integrated energy system based on the selected kernel function, and obtaining linearly separable data based on the similarity includes:

[0019] S1a: using the selected kernel function to measure the distance between the data of the same category at different times in the historical operation data, based on the distance, obtaining the similarity between different data in the historical operation data through an exponential function, and mapping the historical operation data to a high-dimensional feature space based on the similarity to obtain high-dimensional data;

[0020] S1b: Based on the high-dimensional data, a linear classifier is used for training. If the accuracy obtained by training with the linear classifier is greater than the preset accuracy, the high-dimensional data is linearly separable data. If it is less than the preset accuracy, feature data is selected from the historical operation data based on the prediction needs, and the feature data is used as the historical operation data to execute S1a, or the parameters of the linear classifier are adjusted to execute S1b.

[0021] In this scheme, the distance between the same category of data at different times in the historical operation data is measured by the kernel function, and the exponential function is used to convert this distance into similarity, which can more accurately reflect the internal connection between the historical operation data, so as to map the historical operation data to the high-dimensional space, so that the linearly inseparable data can be linearly separable, and the linear classifier is used for training to determine whether the high-dimensional data is linearly separable, and the prediction accuracy and reliability of the high-dimensional data in the current feature space are evaluated. When the historical operation data is inseparable, considering the influence of the historical operation data with poor dependence in the historical operation data, if it is less than the preset accuracy rate, feature data is selected from the historical operation data, which improves the purity and quality of the historical operation data, or considering the irrationality of adjusting the parameters of the linear classifier, if it is less than the preset accuracy rate, the parameters of the linear classifier are adjusted, so that the linear classifier can better adapt to the characteristics of the historical operation data, improve the classification ability and generalization ability of the linear classifier, and under the joint influence of the above characteristics, the linearly inseparable data can be accurately made linearly differentiable.

[0022] Preferably, S2 includes:

[0023] S21: Construct an intelligent swarm optimization engine based on the perceptron neural network and particle swarm optimization algorithm, and obtain the weights of different categories of data in linearly separable data based on the perceptron neural network in the intelligent swarm optimization engine;

[0024] S22: The weight and the linearly separable data are used as independent variables, and the predicted variables in the predicted demand are used as dependent variables, thereby constructing a linear activation function in the perceptron neural network in the intelligent group optimization engine.

[0025] In this solution, since different categories of data in linearly separable data have different effects on the predictive variables in the prediction requirements, and since the perceptron neural network can efficiently calculate the impact of the input data on the output results, the perceptron neural network can be used to obtain the weights of different categories of data in the linearly separable data, thereby improving the calculation efficiency and the accuracy of the output results. Considering the binary classification characteristics of the perceptron neural network, the linear activation function is constructed so that the perceptron neural network is no longer limited to outputting discrete data such as binary classification, but can output continuous data, which significantly improves the versatility and flexibility of the perceptron neural network.

[0026] Preferably, in S21, the step of obtaining weights of different categories of data in linearly separable data based on a perceptron neural network in an intelligent group optimization engine includes:

[0027] S21a: inputting the linearly separable data into the perceptron neural network in the intelligent group optimization engine, classifying the linearly separable data using the classification linear function in the perceptron neural network based on the weight values ​​randomly preset for different categories of data in the linearly separable data, and thereby obtaining a predicted classification result;

[0028] S21b: Compare the predicted classification results with the actual classification results. If the number of successful comparisons is greater than the preset number, the weight values ​​preset for different categories of data in the linearly separable data are the weights of different categories of data in the linearly separable data. If the number of successful comparisons is less than or equal to the preset number, the weight values ​​preset for different categories of data in the linearly separable data are corrected based on the loss function, and the weight values ​​preset for different categories of data in the linearly separable data obtained after the loss function correction are used as the weight values ​​preset for different categories of data in the linearly separable data, and S21a is continued. By comparing the predicted classification results with the actual classification results, the rationality of the obtained weights is verified, and the weights are corrected when they are unreasonable, thereby improving the accuracy of weight distribution.

[0029] Preferably, in S3, the predicted value in the predicted demand includes the overall cooling power available from the integrated energy system under the current energy conversion equipment control strategy, the overall heating power available from the integrated energy system under the current energy conversion equipment control strategy, and the overall electricity power available from the integrated energy system under the current energy conversion equipment control strategy.

[0030] Preferably, S3 includes:

[0031] The optimization objective function , where represents the current operating cost of the integrated energy system, represents the penalty function, Represents the total cooling load required by the user, represents the total heat load required by the user, Indicates the overall electrical load required by the user, is the total cooling power of the integrated energy system under the current energy conversion equipment control strategy, is the total heating power of the integrated energy system under the current park energy conversion equipment control strategy, It is the total power available for the comprehensive energy system under the current energy conversion equipment control strategy. The optimization objective function includes the total cooling load, heating load and electric load required by the user, as well as the cooling, heating and electric power that the system can provide under the current energy conversion equipment control strategy, ensuring that while meeting user needs, energy waste and loss are reduced as much as possible, resource utilization is improved, and the economic benefits of the system are improved.

[0032] Preferably, S5 includes:

[0033] S51: Obtaining an optimization objective function value based on a demand forecast value in the optimization objective function;

[0034] S52: Determine whether the optimization objective function value meets the preset requirements. If so, the energy conversion equipment control strategy corresponding to the optimization objective function value is the optimal control strategy of the energy conversion equipment. If not, the energy conversion equipment control strategy is improved based on the particle swarm optimization algorithm, and the data corresponding to the energy conversion equipment control strategy obtained through the improvement is used as the real-time operation data to execute S4. By continuously optimizing the control strategy of the energy conversion equipment, the system can more accurately match the user's load demand, reduce the deviation and imbalance between supply and demand, improve the system's operating efficiency, and then improve the system's stability and reliability, while improving resource utilization.

[0035] Preferably, in S52, if the conditions are not met, improving the control strategy of the energy conversion device based on the particle swarm optimization algorithm includes:

[0036] Use the first and second formulas in the particle swarm optimization algorithm to obtain the global optimal solution, and improve the control strategy of the energy conversion equipment based on the global optimal solution;

[0037] The first formula is: , where and is the acceleration constant, and is a random number between 0 and 1, is the inertia weight, is the velocity of the ith particle in the (k+1) generation, is the velocity of the ith particle in the kth generation, is the position of the ith particle in the kth generation, is the optimal position of the ith particle in the kth generation, is the optimal position of the particle group in the kth generation;

[0038] The second formula is: , where is the position of the ith particle in the (k+1) generation, is the position of the ith particle in the kth generation, is the velocity of the ith particle in the (k+1)th generation.

[0039] The beneficial effects of this program:

[0040] Taking into account the characteristics of the perceptron neural network, a kernel function is used to obtain the similarity between different data in the historical operation data, so that the linearly inseparable historical operation data can be linearly separable in a high-dimensional space, and the nonlinear historical operation data is converted into linear data that can be used by the perceptron neural network. At the same time, the prediction variables in the prediction requirements are used to construct the linear activation function in the perceptron neural network in the intelligent group optimization engine, so that the perceptron neural network outputs continuous data that can be used for the optimization objective function, which solves the problem that the perceptron neural network is limited to processing linearly separable data and outputting discrete data. Combined with the characteristics of the perceptron neural network with high computational efficiency and the ability to meet real-time requirements, the present invention achieves the effect of taking into account both accuracy and high real-time performance.

[0041] Specifically, the kernel function is used to measure the distance between the same category of data at different times in the historical operation data, and the exponential function is used to convert this distance into similarity, which can more accurately reflect the internal connection between the historical operation data, so as to map the historical operation data to a high-dimensional space, so that the linearly inseparable data can be linearly separable. At the same time, due to the high computational efficiency of the kernel function, the overall computational efficiency is further improved by using the kernel function.

[0042] Specifically, by using the perceptron neural network to obtain the weights of different categories of data in linearly separable data, the computational efficiency and accuracy of the output results are improved. Considering the binary classification characteristics of the perceptron neural network, the linear activation function is constructed so that the perceptron neural network is no longer limited to outputting discrete data such as binary classification, but can output continuous data, which significantly improves the versatility and flexibility of the perceptron neural network, and further improves the computational accuracy and efficiency;

[0043] By using the particle swarm optimization algorithm to continuously optimize the control strategy of energy conversion equipment, the integrated energy system can more accurately match the user's load demand, reduce the deviation and imbalance between supply and demand, thereby improving the stability and reliability of the integrated energy system and improving resource utilization;

[0044] The combined use of kernel function, linear activation function and particle swarm optimization algorithm can more accurately predict and respond to changes in load demand on the basis of taking into account both accuracy and high real-time performance. While ensuring real-time performance, it significantly improves accuracy in many aspects, thereby significantly improving the stability and reliability of the integrated energy system.

[0045] The problem of difficulty in ensuring accuracy while having high real-time performance is solved.

[0046] The present invention also provides a storage medium, in which computer executable instructions are stored. When the computer executable instructions are loaded and executed by a processor, the steps of the integrated energy system operation optimization method based on the intelligent group optimization engine are implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Other features, objects and advantages of the present invention will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings. The drawings are only for the purpose of illustrating preferred embodiments and are not to be considered as limiting the present invention. Also, the same reference symbols are used throughout the drawings to represent the same parts.

[0048] Figure 1 It is a flow chart of the integrated energy system operation optimization method based on the intelligent group optimization engine of the present invention;

[0049] Figure 2 It is a structural schematic diagram of the comprehensive energy system in the present invention;

[0050] Figure 3 It is the first convergence process curve of the iterative optimization of the particle swarm optimization algorithm in the present invention;

[0051] Figure 4 It is the second convergence process curve of the iterative optimization of the particle swarm optimization algorithm in the present invention;

[0052] Figure 5 It is the third convergence process curve of the iterative optimization of the particle swarm optimization algorithm in the present invention. DETAILED DESCRIPTION

[0053] In order to make the objectives, 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 implementation method described herein is only an optimal embodiment of the present invention, which is only used to explain the present invention and does not limit the scope of protection of the present invention. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0054] Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flow charts. Although the flow charts describe the operations (or steps) as sequential processes, many of the operations (or steps) therein can be implemented in parallel, concurrently, or simultaneously. In addition, the order of the operations can be rearranged. The process can be terminated when its operation is completed, but can also have additional steps not included in the drawings; the process can correspond to a method, function, procedure, subroutine, subprogram, etc.

[0055] Embodiment 1:

[0056] like Figure 1 As shown, the integrated energy system operation optimization method based on the intelligent group optimization engine includes the following steps:

[0057] S1: Select a kernel function based on the historical operation data characteristics and predicted demand of the integrated energy system, calculate the similarity between different data in the historical operation data of the integrated energy system based on the selected kernel function, and obtain linearly separable data based on the similarity.

[0058] In S1, the kernel function selected based on the historical operation data characteristics and predicted demand of the integrated energy system includes:

[0059] Clean the historical operation data of the integrated energy system, remove outliers, remove missing values, and normalize;

[0060] The data obtained at the same time point in the historical operation data is analyzed with the predicted demand to obtain local features of the data, the data of the same category in the historical operation data is analyzed with the predicted demand to obtain global features of the data, and a kernel function is selected based on the local features and global features of the data.

[0061] In S1, the calculating the similarity between different data in the historical operation data of the integrated energy system based on the selected kernel function, and obtaining linearly separable data based on the similarity includes:

[0062] S1a: using the selected kernel function to measure the distance between the data of the same category at different times in the historical operation data, based on the distance, obtaining the similarity between different data in the historical operation data through an exponential function, and mapping the historical operation data to a high-dimensional feature space based on the similarity to obtain high-dimensional data;

[0063] S1b: Based on the high-dimensional data, a linear classifier is used for training. If the accuracy obtained by training with the linear classifier is greater than the preset accuracy, the high-dimensional data is linearly separable data. If it is less than the preset accuracy, feature data is selected from the historical operation data based on the prediction needs, and the feature data is used as the historical operation data to execute S1a, or the parameters of the linear classifier are adjusted to execute S1b.

[0064] The integrated energy system is a trigeneration integrated energy system for the park, such as Figure 2As shown, AB is a gas boiler with a maximum heating power of 400kW; CERG is an electric refrigeration equipment with a maximum cooling power of 300kW; WARG is an absorption refrigeration equipment with a maximum cooling power of 300kW; CHP is a cogeneration equipment with a maximum power supply power of 120kW and a maximum heating power of 160kW; the energy conversion equipment control strategy of the park's comprehensive energy system is composed of the electric power supplied to the electric refrigeration equipment , Natural gas power supplied to gas boilers , supply heat power to absorption refrigeration equipment , Natural gas power for combined heat and power generation equipment And the park's external power Consists of.

[0065] Historical operation data includes energy conversion equipment control strategy data under all operating conditions (such as the power supplied to the electric refrigeration equipment , Natural gas power supplied to gas boilers , supply heat power to absorption refrigeration equipment , Natural gas power for combined heat and power generation equipment And the park's external power data) and energy conversion equipment output data (total available power supply in the park 、Total heating power of the park 、Total cooling power of the park ) etc. When selecting the kernel function, if the local features of the data are more obvious, you can choose the Gaussian kernel function; if the global features of the data are more obvious, you can choose the polynomial kernel function; if the local features and global features of the data are not obvious or both are obvious, you can choose the mixed kernel function. The expression of the Gaussian kernel function is , where All are input data of the Gaussian kernel function, which are historical running data here. express and The Euclidean distance between is the bandwidth parameter of the Gaussian kernel function. The expression of the polynomial kernel function is , where are all input data of the polynomial kernel function, which are historical running data here. express The inner product of is a constant term, is the degree of the polynomial, and the expression of the mixed kernel function is , where is the Gaussian kernel function, is the polynomial kernel function, is the weighting coefficient, ranging from 0 to 1.

[0066] The linear classifier finds a hyperplane, for example, a straight line in two-dimensional space, a plane in three-dimensional space, and so on to high-dimensional space, and uses the hyperplane to separate data points of different categories, thereby obtaining linearly separable data. The simple decision boundary of the linear classifier is expressed as , where is the weight vector, which is the weight of different categories of historical running data. To output the feature vector, here it is the historical running data, is a bias term. In this embodiment, the weight vector can also be used to obtain the predicted value in the predicted demand. If the accuracy rate obtained by training with a linear classifier is less than the preset accuracy rate, feature data is selected from the historical operation data based on the predicted demand, and the feature data is used as the historical operation data. The predicted demand is the overall cooling power available for the integrated energy system under the current energy conversion equipment control strategy, the overall heating power available for the integrated energy system under the current energy conversion equipment control strategy, and the overall power available for the integrated energy system under the current energy conversion equipment control strategy. The feature data are the electric power supplied to the electric refrigeration equipment, the natural gas power supplied to the gas boiler, the thermal power supplied to the absorption refrigeration equipment, the natural gas power supplied to the cogeneration equipment, the power data of the purchased electric power of the park, the total power available for the park, the total heating power available for the park, and the total cooling power available for the park.

[0067] By selecting kernel functions based on dependencies and rules, the kernel functions can better adapt to different data distributions and characteristics, improve the accuracy of obtaining linearly separable data, and further improve the efficiency of calculation by selecting appropriate kernel functions.

[0068] S2: Build an intelligent swarm optimization engine based on the perceptron neural network and particle swarm optimization algorithm, and build a linear activation function in the perceptron neural network in the intelligent swarm optimization engine based on the prediction variables in the predicted demand.

[0069] The S2 includes:

[0070] S21: Construct an intelligent swarm optimization engine based on the perceptron neural network and particle swarm optimization algorithm, and obtain the weights of different categories of data in linearly separable data based on the perceptron neural network in the intelligent swarm optimization engine;

[0071] S22: The weight and the linearly separable data are used as independent variables, and the predicted variables in the predicted demand are used as dependent variables, thereby constructing a linear activation function in the perceptron neural network in the intelligent group optimization engine.

[0072] In S21, the weights of different categories of data in the linearly separable data are obtained based on the perceptron neural network in the intelligent group optimization engine, including:

[0073] S21a: inputting the linearly separable data into the perceptron neural network in the intelligent group optimization engine, classifying the linearly separable data using the classification linear function in the perceptron neural network based on the weight values ​​randomly preset for different categories of data in the linearly separable data, and thereby obtaining a predicted classification result;

[0074] S21b: Compare the predicted classification result with the actual classification result. If the number of successful comparisons is greater than the preset number, the weight values ​​randomly preset for different categories of data in the linearly separable data are the weights of different categories of data in the linearly separable data. If the number of successful comparisons is less than or equal to the preset number, the weight values ​​preset for different categories of data in the linearly separable data are corrected based on the loss function, and the weight values ​​preset for different categories of data in the linearly separable data obtained after correction by the loss function are used as the weight values ​​randomly preset for different categories of data in the linearly separable data, and continue to execute S21a.

[0075] The perceptron neural network consists of an input layer and an output layer. In the input layer, data is received and passed to the output layer, and the output layer outputs the classification result based on the weighted sum of the input signal and the relationship with the threshold. The particle swarm optimization algorithm is an optimization algorithm based on swarm intelligence. Each particle represents a potential solution. The position of the particle represents the parameter value of the solution, and the speed represents the moving direction and step size of the solution in the search space. By iteratively updating the speed and position of the particles, the optimal solution or a solution close to the optimal solution can be found in the search space. The prediction variables in the demand include the overall cooling power available for the integrated energy system under the current energy conversion equipment control strategy, the overall heating power available for the integrated energy system under the current energy conversion equipment control strategy, and the overall power available for the integrated energy system under the current energy conversion equipment control strategy. The expression of the linear activation function is , where is the predictor variable in forecasting demand, is the power consumed in the conversion process, n is the number of factors affecting the predictive variables in the forecast demand, that is, the number of different categories of data in the historical operation data, Indicates the specific value of the first influencing factor, It indicates the influence of the first influencing factor on the predicted variable in the forecast demand, and other parameters can be obtained by analogy. In addition to constructing the linear activation function, the function in the perceptron neural network can also be improved under the condition that the weights obtained by the perceptron neural network meet the requirements, and the bias value in the function is changed to the power consumed in the conversion process.

[0076] By using the perceptron neural network to obtain the weights of different categories of data in linearly separable data, the computational efficiency and the accuracy of the output results are improved. Considering the binary classification characteristics of the perceptron neural network, the linear activation function is constructed so that the perceptron neural network is no longer limited to outputting discrete data such as binary classification, but can output continuous data, which significantly improves the versatility and flexibility of the perceptron neural network.

[0077] S3: Construct an optimization objective function using the purchased energy price, the predicted value in the forecasted demand, and the user’s overall load.

[0078] In S3, the predicted values ​​in the predicted demand include the overall cooling power available from the integrated energy system under the current energy conversion equipment control strategy, the overall heating power available from the integrated energy system under the current energy conversion equipment control strategy, and the overall electricity power available from the integrated energy system under the current energy conversion equipment control strategy.

[0079] The S3 includes:

[0080] The optimization objective function , where represents the current operating cost of the integrated energy system, represents the penalty function, Represents the total cooling load required by the user, represents the total heat load required by the user, Indicates the overall electrical load required by the user, is the total cooling power of the integrated energy system under the current energy conversion equipment control strategy, is the total heating power of the integrated energy system under the current park energy conversion equipment control strategy, It is the total power that can be supplied by the comprehensive energy system under the current energy conversion equipment control strategy.

[0081] , and so on and , where the total cooling load required by users, the total heating load required by users and the total electricity load required by users are the operating constraints, and the total cooling power available for the integrated energy system under the current energy conversion equipment control strategy, the total heating power available for the integrated energy system under the current park energy conversion equipment control strategy and the total electricity power available for the integrated energy system under the current energy conversion equipment control strategy are adjustable control strategies. The optimization objective function contains a relatively comprehensive range of data types, ensuring that while meeting user needs, energy waste and loss are reduced as much as possible, resource utilization is improved, and the economic benefits of the system are improved.

[0082] S4: Input the real-time acquired operation data into the perceptron neural network in the intelligent group optimization engine, and obtain the demand forecast value in the optimization objective function based on the linear activation function. The real-time acquired operation data and the historical operation data include the same data types, which will not be repeated here.

[0083] S5: Based on the particle swarm optimization algorithm in the intelligent group optimization engine, the demand forecast value in the optimization objective function is used to obtain the optimal control strategy of the energy conversion equipment, and then optimize the operation of the integrated energy system.

[0084] The S5 includes:

[0085] S51: Obtaining an optimization objective function value based on a demand forecast value in the optimization objective function;

[0086] S52: Determine whether the optimization objective function value meets the preset requirements. If so, the energy conversion equipment control strategy corresponding to the optimization objective function value is the optimal control strategy of the energy conversion equipment. If not, the energy conversion equipment control strategy is improved based on the particle swarm optimization algorithm. The data corresponding to the energy conversion equipment control strategy obtained through the improvement is used as the real-time operation data, and S4 is executed.

[0087] In S52, if the control strategy of the energy conversion device is not satisfied, then the control strategy is improved based on the particle swarm optimization algorithm, including:

[0088] Use the first and second formulas in the particle swarm optimization algorithm to obtain a global optimal solution, and improve the energy conversion equipment control strategy based on the global optimal solution;

[0089] The first formula is:

[0090] , where and is the acceleration constant, and is a random number between 0 and 1, is the inertia weight, is the velocity of the ith particle in the (k+1) generation, is the velocity of the ith particle in the kth generation, is the position of the ith particle in the kth generation, is the optimal position of the ith particle in the kth generation, is the optimal position of the particle group in the kth generation;

[0091] The second formula is:

[0092] , where is the position of the ith particle in the (k+1) generation, is the position of the ith particle in the kth generation, is the velocity of the ith particle in the (k+1)th generation.

[0093] is the inertia weight between 1.4 and 0.4, which is gradually reduced from 1.4 to 0.4 as the number of iterations increases. The population size is 1000, and the number of iterations is 80. After continuous iterative updates, the particle group gradually moves to the optimal solution in the solution space, and finally obtains the global optimal solution. Based on the global optimal solution, the control strategy of the energy conversion equipment is improved to minimize the optimization target, such as Figure 3 As shown in the figure, under the constraints of the current period of the park's purchased energy price and the overall electricity, heat and cooling loads required by the park users, the one-hour operating cost of the current park's comprehensive energy system gradually decreases from 124.6 yuan to 91.9 yuan with the iterative optimization of particle swarm optimization. The total energy supply of the optimized park and the overall electricity, heat and cooling loads required by the park users achieve a supply-demand balance, and minimize the current comprehensive energy system operating cost under the constraint of meeting the overall electricity, heat and cooling loads required by the park users. The effects of the comprehensive energy system energy conversion equipment control strategy before and after optimization are shown in Table 1:

[0094] Table 1. Comparison of the first effects before and after optimization of the control strategy of energy conversion equipment in the integrated energy system

[0095]

[0096] like Figure 4 As shown in the figure, under the constraints of the current period of the park's purchased energy price and the overall electricity, heat and cooling loads required by the park users, the one-hour operating cost of the current park's comprehensive energy system gradually decreases from 247.9 yuan to 194.2 yuan with the iterative optimization of particle swarm optimization. The total energy supply of the optimized park and the overall electricity, heat and cooling loads required by the park users achieve a supply-demand balance, and minimize the current comprehensive energy system operating cost under the constraint of meeting the overall electricity, heat and cooling loads required by the park users. The effects of the comprehensive energy system energy conversion equipment control strategy before and after optimization are shown in Table 2:

[0097] Table 2. Comparison of the second effect before and after optimization of the control strategy of energy conversion equipment in the integrated energy system

[0098]

[0099] like Figure 5As shown in the figure, under the constraints of the current period of the park's purchased energy price and the overall electricity, heat and cooling loads required by the park users, the one-hour operation cost of the current park's comprehensive energy system gradually decreases from 325.1 yuan to 296.9 yuan with the iterative optimization of particle swarm optimization. The total energy supply of the park after optimization and the overall electricity, heat and cooling loads required by the park users achieve a balance between supply and demand, and minimize the current comprehensive energy system operation cost under the constraint of meeting the overall electricity, heat and cooling loads required by the park users. The effects of the comprehensive energy system energy conversion equipment control strategy before and after optimization are shown in Table 3:

[0100] Table 3. Comparison of the third effect before and after optimization of the control strategy of energy conversion equipment in the integrated energy system

[0101]

[0102] By continuously optimizing the control strategies of energy conversion equipment, the system can more accurately match the user's load requirements, reduce the deviation and imbalance between supply and demand, improve the system's operating efficiency, and thus improve the system's stability and reliability, while also improving resource utilization, thereby achieving cost reduction and efficiency improvement in the park.

[0103] Embodiment 2:

[0104] This embodiment also provides a storage medium, in which computer executable instructions are stored. When the computer executable instructions are loaded and executed by a processor, the steps of the integrated energy system operation optimization method based on the intelligent group optimization engine are implemented.

[0105] The specific implementation modes described above are preferred implementation modes of the present invention, and are not intended to limit the specific implementation scope of the present invention. The scope of the present invention includes but is not limited to the specific implementation modes. All equivalent changes made in accordance with the shape and structure of the present invention are within the protection scope of the present invention.

Claims

1. A comprehensive energy system operation optimization method based on an intelligent group optimization engine, characterized in that: The following steps are involved: S1: Select a kernel function based on the historical operation data characteristics and predicted demand of the integrated energy system, calculate the similarity between different data in the historical operation data of the integrated energy system based on the selected kernel function, and obtain linearly separable data based on the similarity; wherein, the kernel function is selected by the local characteristics of the data and the global characteristics of the data in the historical operation data; the distance between the same category of data at different times in the historical operation data is measured using the selected kernel function, and based on the distance, the similarity between different data in the historical operation data is obtained by an exponential function; S2: Build a smart group optimization engine based on the perceptron neural network and particle swarm optimization algorithm, and build a linear activation function in the perceptron neural network in the smart group optimization engine based on the prediction variables in the predicted demand; The S2 includes: S21: Construct an intelligent swarm optimization engine based on the perceptron neural network and particle swarm optimization algorithm, and obtain the weights of different categories of data in linearly separable data based on the perceptron neural network in the intelligent swarm optimization engine; S22: taking the weight and the linearly separable data as independent variables, taking the predicted variable in the predicted demand as the dependent variable, and then constructing a linear activation function in the perceptron neural network in the intelligent group optimization engine; S3: Construct an optimization objective function using the purchased energy price, the predicted value in the forecast demand, and the user's overall load; In S3, the predicted value in the predicted demand includes the overall cooling power available from the integrated energy system under the current energy conversion equipment control strategy, the overall heating power available from the integrated energy system under the current energy conversion equipment control strategy, and the overall electricity power available from the integrated energy system under the current energy conversion equipment control strategy; The S3 includes: The optimization objective function OF = cost + come (L c , P c )+comp(L h +P h )+comp(L e , P e ), where cost represents the current operating cost of the integrated energy system, comp() represents the penalty function, and L c Indicates the total cooling load required by the user, L h Indicates the total heat load required by the user, L e Indicates the total electrical load required by the user, P c is the total cooling power of the integrated energy system under the current energy conversion equipment control strategy, P h is the total heating power of the integrated energy system under the current park energy conversion equipment control strategy, P e It is the total power available for the integrated energy system under the current energy conversion equipment control strategy; S4: inputting the real-time acquired operation data into the perceptron neural network in the intelligent group optimization engine, and obtaining the demand forecast value in the optimization objective function based on the linear activation function; S5: Based on the particle swarm optimization algorithm in the intelligent group optimization engine, the demand forecast value in the optimization objective function is used to obtain the optimal control strategy of the energy conversion equipment, and then optimize the operation of the integrated energy system.

2. The integrated energy system operation optimization method based on the intelligent group optimization engine according to claim 1 is characterized in that: In S1, the kernel function selected based on the historical operation data characteristics and predicted demand of the integrated energy system includes: Clean the historical operation data of the integrated energy system, remove outliers, remove missing values, and normalize; The data obtained at the same time point in the historical operation data is analyzed with the predicted demand to obtain local features of the data, the data of the same category in the historical operation data is analyzed with the predicted demand to obtain global features of the data, and a kernel function is selected based on the local features and global features of the data.

3. The integrated energy system operation optimization method based on the intelligent group optimization engine according to claim 1 is characterized in that: In S1, the calculating the similarity between different data in the historical operation data of the integrated energy system based on the selected kernel function, and obtaining linearly separable data based on the similarity includes: S1a: Mapping the historical operation data to a high-dimensional feature space based on the similarity to obtain high-dimensional data; S1b: Based on the high-dimensional data, a linear classifier is used for training. If the accuracy obtained by training with the linear classifier is greater than the preset accuracy, the high-dimensional data is linearly separable data. If it is less than the preset accuracy, feature data is selected from the historical operation data based on the prediction needs, and the feature data is used as the historical operation data to execute S1a, or the parameters of the linear classifier are adjusted to execute S1b.

4. The integrated energy system operation optimization method based on the intelligent group optimization engine according to claim 1 is characterized in that: In S21, the step of obtaining weights of different categories of data in the linearly separable data based on the perceptron neural network in the intelligent group optimization engine includes: S21a: inputting the linearly separable data into the perceptron neural network in the intelligent group optimization engine, and classifying the linearly separable data using the classification linear function in the perceptron neural network based on the weight values ​​randomly preset for different categories of data in the linearly separable data, thereby obtaining a predicted classification result; S21b: Compare the predicted classification result with the actual classification result. If the number of successful comparisons is greater than the preset number, the weight values ​​randomly preset for different categories of data in the linearly separable data are the weights of different categories of data in the linearly separable data. If the number of successful comparisons is less than or equal to the preset number, the weight values ​​preset for different categories of data in the linearly separable data are corrected based on the loss function, and the weight values ​​preset for different categories of data in the linearly separable data obtained after correction by the loss function are used as the weight values ​​randomly preset for different categories of data in the linearly separable data, and continue to execute S21a.

5. The integrated energy system operation optimization method based on the intelligent group optimization engine according to claim 1 is characterized in that: The S5 includes: S51: Obtaining an optimization objective function value based on a demand forecast value in the optimization objective function; S52: Determine whether the optimization objective function value meets the preset requirements. If so, the energy conversion equipment control strategy corresponding to the optimization objective function value is the optimal control strategy of the energy conversion equipment. If not, the energy conversion equipment control strategy is improved based on the particle swarm optimization algorithm. The data corresponding to the energy conversion equipment control strategy obtained through the improvement is used as the real-time operation data, and S4 is executed.

6. The integrated energy system operation optimization method based on the intelligent group optimization engine according to claim 5 is characterized in that: In S52, if the control strategy of the energy conversion device is not satisfied, the control strategy based on the particle swarm optimization algorithm is improved, including: Use the first and second formulas in the particle swarm optimization algorithm to obtain the global optimal solution, and improve the control strategy of the energy conversion equipment based on the global optimal solution; The first formula is: Where c1 and c2 are acceleration constants, r1 and r2 are random numbers between 0 and 1, and W is the inertia weight. is the velocity of the ith particle in the (k+1) generation, is the velocity of the ith particle in the kth generation, is the position of the ith particle in the kth generation, is the optimal position of the ith particle in the kth generation, gBest k is the optimal position of the particle group in the kth generation; The second formula is: In the formula, is the position of the ith particle in the (k+1)th generation, is the position of the ith particle in the kth generation, is the velocity of the ith particle in the (k+1)th generation.

7. A storage medium, characterized in that: The storage medium stores computer executable instructions, which, when loaded and executed by a processor, implement the steps of the integrated energy system operation optimization method based on an intelligent group optimization engine as described in any one of claims 1 to 6.

Images