CN118195104A - Method, device, medium and product for optimizing configuration of commercial complex building power supply system - Google Patents

Abstract

The invention discloses a method, a device, a medium and a product for optimizing configuration of a commercial complex building power supply system, and relates to the field of commercial complex buildings. The method comprises the steps of firstly defining thermal coefficients of different thermal areas, taking thermal coefficients, building heights and functional areas of built commercial complex buildings in the different thermal areas as input quantities, taking load configuration data under a unit area as output quantities, constructing a data set, then adopting a genetic algorithm to optimize an initial weight threshold of a BP neural network, utilizing the data set to train the BP neural network after assignment to obtain a load capacity prediction model, finally using the trained BP neural network to predict load configuration data of the newly built commercial complex buildings in the future, and determining optimal load configuration data of the built commercial complex buildings based on the load capacity prediction model and an improved particle swarm algorithm based on random distribution time lag. The invention can accurately predict and optimize the power load and the transformer capacity of the commercial complex building.

Description

Method, device, medium and product for optimizing configuration of commercial complex building power supply system

Technical Field

The invention relates to the field of commercial complex architecture, in particular to a method, a device, a medium and a product for optimizing configuration of a commercial complex architecture power supply system.

Background

At present, a traditional design method is adopted for a commercial complex building power supply and distribution system, and the influence of the complex environment on the operation characteristics of the transformer is not considered. Conventional designs may result in significant losses in the transformer during operation, severely impacting the economics of the system operation, and also impacting the actual operating efficiency of the transformer.

Disclosure of Invention

The invention aims to provide an optimal configuration method, device, medium and product of a commercial complex building power supply system, which can accurately predict and optimize the power load and transformer capacity of the commercial complex building.

In order to achieve the above object, the present invention provides the following.

An optimization configuration method for a commercial complex building power supply system comprises the following steps: the method comprises the steps of defining the value of a thermal coefficient of a summer hot and winter cold region to be 1 by taking the transformer capacity average value of commercial complex buildings in the summer hot and winter cold region as a reference value, and determining the ratio of the transformer capacity average value of the commercial complex buildings in each thermal region except the summer hot and winter cold region to the reference value as the value of the thermal coefficient of each thermal region except the summer hot and winter cold region. Constructing a data set by taking the thermal coefficient, the building height and the functional ecological area of the built commercial complex building in different thermal areas as input quantity and the load configuration data in unit area as output quantity; the load configuration data includes power load, air conditioning load, lighting load, reactive power loss, and transformer capacity. And optimizing an initial weight threshold of the BP neural network by adopting a genetic algorithm, obtaining an optimal initial weight threshold, and assigning the optimal initial weight threshold to the BP neural network. And acquiring a load capacity prediction model by utilizing the BP neural network after the data set training assignment. And inputting the thermal coefficient, the building height and the functional ecological area of the future newly-built commercial complex building into the load capacity prediction model to predict the load configuration data of the future newly-built commercial complex building. According to the thermal coefficient, building height and functional ecological area of the built commercial complex building, the load capacity prediction model is used as an objective function of an improved particle swarm algorithm based on random distribution time lag, and optimal load configuration data of the built commercial complex building is determined; the improved particle swarm algorithm based on random distribution time lag introduces random distribution time lag in a speed update model.

Optionally, the building method uses thermal coefficients, building heights and functional areas of the built commercial complex buildings in different thermal areas as input quantities, uses load configuration data in unit area as output quantities, and builds a data set, and the building method further comprises the following steps: unifying the units of the load configuration data, and representing the load configuration data in the form of unit area; outliers in the functional stateful area and outliers in the load configuration data are removed.

Optionally, the BP neural network includes an input layer, an implicit layer, and an output layer connected in sequence; the number of neurons of the input layer is 6; the number of neurons of the hidden layer is 10, and the activation function of the hidden layer is a tanh function; the number of neurons of the output layer is 6, and the activation function of the output layer is a linear activation function.

Optionally, the speed update model of the improved particle swarm algorithm based on random distribution time lag is as follows:

。

Wherein, 、/>The speeds of the ith particle in k+1 iterations and k iterations are respectively represented; w represents an inertia factor; /(I)Representing the cognitive acceleration coefficient,/>Representing social acceleration coefficient,/>And/>First and second acceleration coefficients representing a distributed time lag term,/>=/>And/>=/>；/>、/>、/>And/>Is a random number uniformly distributed in [0,1 ]; /(I)Representing the number of delay iterations; /(I)Representing the optimal position of the ith particle in k iterations,/>Indicating that the ith particle is at k-/>Optimal position of the secondary iteration; /(I)Represents the position of the particle globally optimal in k iterations,/>Represented in k-/>Iterating the position of the globally optimal particle; /(I)Representing the current position of the ith particle at k iterations; /(I)And/>Respectively representing the intensity factors of individual distribution delay terms and global distribution delay terms according to the evolution state xi; /(I)Representing an n-dimensional vector, each element in the n-dimensional vector being randomly selected from 0 or 1; n represents the upper bound of the distribution delay.

The position update model of the improved particle swarm algorithm based on the random distribution time lag is as follows: ; wherein/> Representing the current position of the ith particle at k+1 iterations.

The inertia weight updating model of the improved particle swarm algorithm based on the random distribution time lag is as follows: ; wherein/> And/>Respectively representing the maximum value and the minimum value of the inertia weight; item represents the current iteration number and maxiter represents the maximum iteration number.

The acceleration coefficient updating model of the improved particle swarm algorithm based on the random distribution time lag is as follows: And/> 。

Wherein,Representing the final value of the cognitive acceleration coefficient,/>Representing the initial value of the cognitive acceleration coefficient of the ith particle,/>Representing the final value of the social acceleration coefficient,/>The initial value of the social acceleration coefficient of the i-th particle is represented.

Optionally, in the velocity update modelAnd/>The determining method of (1) comprises the following steps: according to the formulaCalculating a distance between the particles; wherein/>Represents the average distance between the ith particle and other particles, s represents the particle population size, D represents the particle size,/>Representing the current position of the jth particle in k iterations; according to the distance between particles, the formula/>, is usedCalculating an evolution factor; wherein/>Representing evolution factors,/>Representing the distance between the current location and the global best particle,/>And/>Representing the minimum and maximum values, respectively, of the distance between particles in the population of particles; when the value of the evolution factor is less than 0.5,/>And/>All take values of 0; when the value of the evolution factor is greater than or equal to 0.5,/>And/>The average value is 0.01.

Optionally, the workflow of the improved particle swarm algorithm based on the random distribution time lag specifically comprises the following steps: initializing the size of the particle population, the initial position of the particles and the initial velocity of the particles; calculating the fitness of each particle; updating the optimal position of each particle and the position of the globally optimal particle according to the fitness; calculating a distance between the particles; calculating an evolution factor according to the distance between the particles; determining the intensity factor of a distributed time delay term according to the evolution state xi according to the evolution factor; updating the inertia weight; updating the cognitive acceleration coefficient and the social acceleration coefficient; updating the randomly generated distributed delay information according to the intensity factor, the updated inertia weight, the updated cognitive acceleration coefficient and the updated social acceleration coefficient; updating the speed and the position of the particles according to the updated random distributed delay information, the speed updating model and the position updating model; increasing the value of the iteration times by 1; if the iteration number after the value is increased is smaller than the maximum iteration number, executing the step of calculating the fitness of each particle; if the iteration times after the value increase is equal to the maximum iteration times, ending, and taking the position of the updated globally optimal particle as the optimal position of the particle.

A computer apparatus, comprising: the system comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the computer program to realize the steps of the method for optimizing and configuring the commercial complex building power supply system.

A computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of the above-described method for optimizing the configuration of a commercial complex building power supply system.

A computer program product comprising a computer program which, when executed by a processor, performs the steps of the commercial complex building power system optimization configuration method described above.

According to the specific embodiments provided by the invention, the following technical effects are disclosed.

The embodiment of the invention provides a method, a device, a medium and a product for optimizing configuration of a commercial complex building power supply system, which provides a new thermal coefficient concept, adds a thermal region into a load capacity prediction model in a definite numerical form, and enables the thermal region to serve as a new influence factor to optimize accurate training of the load capacity prediction model; optimizing an initial weight threshold of the BP neural network by adopting a genetic algorithm, so that a load capacity prediction model obtained by training the assigned BP neural network has higher prediction accuracy on load configuration data of a newly built commercial complex building in the future; aiming at the established commercial complex building, the improved particle swarm optimization based on random distribution time lag still has a better convergence effect on the multi-dimensional problem optimization of the commercial complex building, can avoid sinking into local optimum, and has higher accuracy by optimizing and configuring the electricity load capacity of the commercial complex building by calling the load capacity prediction model.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

Fig. 1 is a schematic flow chart of a method for optimizing configuration of a commercial complex building power supply system according to embodiment 1 of the present invention.

Fig. 2 is a schematic structural diagram of a BP neural network according to embodiment 1 of the present invention.

Fig. 3 is a flowchart of an improved neural network according to embodiment 1 of the present invention.

Fig. 4 is a schematic diagram of a training effect of a BP neural network according to embodiment 1 of the present invention.

Fig. 5 is a flowchart of the improved particle swarm algorithm based on the random distribution time lag according to embodiment 1 of the present invention.

Fig. 6 is a schematic diagram of an optimization convergence curve of Sphere function provided in embodiment 1 of the present invention.

Fig. 7 is a schematic diagram of an optimized convergence curve of Rosenbrock functions provided in embodiment 1 of the present invention.

Fig. 8 is a schematic diagram of an optimization convergence curve of the Ackley function provided in embodiment 1 of the present invention.

Fig. 9 is a schematic diagram of an optimized convergence curve of RASTRIGIN functions provided in embodiment 1 of the present invention.

Fig. 10 is a schematic diagram of an optimized convergence curve of Griewank functions provided in embodiment 1 of the present invention.

Fig. 11 is a graph comparing convergence curves of the improved particle swarm algorithm based on the random distribution time lag and the basic particle swarm algorithm in the severe cold region provided in embodiment 1 of the present invention.

Fig. 12 is a graph showing a comparison of convergence curves of the improved particle swarm algorithm and the basic particle swarm algorithm based on the random distribution time lag in the cold regions according to example 1 of the present invention.

Fig. 13 is a graph comparing convergence curves of the improved particle swarm algorithm based on the random distribution time lag and the basic particle swarm algorithm in the summer hot and winter cold regions provided in the embodiment 1 of the present invention.

Fig. 14 is a graph comparing convergence curves of the improved particle swarm algorithm based on the random distribution time lag and the basic particle swarm algorithm in the summer heat and winter warm region provided in embodiment 1 of the present invention.

Fig. 15 is a graph comparing convergence curves of the modified particle swarm algorithm and the basic particle swarm algorithm based on random distribution time lags in the mild regions provided in example 1 of the present invention.

Fig. 16 is a schematic diagram of a program window according to embodiment 1 of the present invention.

Fig. 17 is an internal structural view of the computer device.

Detailed Description

The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to the appended drawings and appended detailed description.

Example 1.

As shown in fig. 1, a method for optimizing configuration of a commercial complex building power supply system in this embodiment includes the following steps.

Step 1: the method comprises the steps of defining the value of a thermal coefficient of a summer hot and winter cold region to be1 by taking the transformer capacity average value of commercial complex buildings in the summer hot and winter cold region as a reference value, and determining the ratio of the transformer capacity average value of the commercial complex buildings in each thermal region except the summer hot and winter cold region to the reference value as the value of the thermal coefficient of each thermal region except the summer hot and winter cold region.

The invention provides a new concept of thermal coefficient aiming at the optimization problem of a commercial complex building power supply system, so as to explore the influence degree of climate on the capacity configuration of a transformer and an electric load.

Firstly, according to the thermal region division of the civil building thermal design specification (GB 50176-2016), the regions where the collected and processed commercial complex buildings are located are ordered according to the severe cold region, the summer hot and winter warm region and the mild region. After finishing, the data volume collected in the summer hot and winter cold region is found to be the largest, so that the average value of the transformer capacity of the commercial complex building in the summer hot and winter cold region is taken as a reference value, and the thermal coefficient is defined as 1. The ratio of the transformer capacity average value of the commercial complex building in the waste heat working area to the reference value is the value of the thermal coefficient of the thermal area, and can be expressed as the formula (1).

(1)。

Wherein,Representing the thermal coefficient of the i' th thermal region,/>Representing the reference value,/>The transformer capacity average of the commercial complex building in the ith thermal area is shown.

The calculation of the above formula shows that the thermal coefficient in the severe cold region is 0.95, the thermal coefficient in the cold region is 1.13, the thermal coefficient in the summer, the winter and the warm region is 1.06, and the thermal coefficient in the mild region is 0.89. It should be noted that, considering the electricity utilization characteristics of the areas where different commercial complex buildings are located and the problem of reserving the margin in the design stage, the value of the thermal coefficient may float up and down.

Step 2: constructing a data set by taking the thermal coefficient, the building height and the functional ecological area of the built commercial complex building in different thermal areas as input quantity and the load configuration data in unit area as output quantity; the load configuration data includes power load, air conditioning load, lighting load, reactive power loss, and transformer capacity.

The sample collection method in the data set comprises the following steps: by collecting load data of a commercial complex building which is built and operated for more than one year, the load influencing factors including a thermal coefficient k', a building height h and main functional status areas (office area, hotel area, commercial area and other areas) are determined. The load components mainly comprise power load, air conditioning load, lighting load, reactive power loss and transformer capacity. Because the commercial complexes studied in the invention have different building scales, the building height and the overall differentiation of each functional status area are large, and the data are required to be preprocessed in order to ensure the universality and generality of input parameters: firstly unifying units, and more clearly reflecting the arrangement condition of each power load and the capacity of the transformer in a unit area mode. Wherein the units of the lighting load, the power load and the air conditioning load areThe unit of transformer capacity per unit area is/>Reactive loss is expressed as/>. Subsequently, outlier data in the sample is removed. The outlier data mainly comprises data of unreasonable design of the power load and the capacity of the transformer or the range far exceeding the normal operation range. Such data is not representative and cannot meet modeling requirements. In addition, some projects are outdoor projects such as recreation ground, the outdoor area of the projects is far larger than the indoor area, and the projects are inconsistent with the research content of the invention and belong to outlier data.

The above pretreatment can be summarized as: unifying the units of the load configuration data, and representing the load configuration data in the form of unit area; outliers in the functional stateful area and outliers in the load configuration data are removed.

The method is characterized in that load influence factors are taken as independent variables, load components are taken as dependent variables, meanwhile, the complexity of a multi-input multi-output optimization and prediction model is considered, the thermal coefficient k' of a commercial complex building, the building height h and main functional business areas (office area, hotel area, business area and other areas) are taken as input quantities, and the power load, air conditioning load, lighting load, reactive power loss and transformer capacity under unit area are taken as output quantities.

Step 3: and optimizing an initial weight threshold of the BP neural network by adopting a genetic algorithm, obtaining an optimal initial weight threshold, and assigning the optimal initial weight threshold to the BP neural network.

A BP neural network model optimized based on a genetic algorithm is established, and meanwhile, the neural network error initialized by parameters is used as a fitness value of a target commercial complex building load, and the optimal individual genes, namely the optimal initial weight threshold of the BP neural network, are searched through selection, crossing and mutation operations of the genetic algorithm.

The BP neural network optimization based on the genetic algorithm mainly comprises three parts of determining a neural network structure, optimizing an initial weight threshold value by the genetic algorithm and predicting the neural network. The BP neural network mainly determines a neural network structure according to input quantity, output quantity, fitting functions and the like, wherein 6 neurons in the input layer, 10 neurons in the hidden layer, the activation function of the hidden layer is a tanh function, 6 neurons in the output layer are output, and the activation function is a linear activation function. The structure of the BP neural network is shown in FIG. 2, W in FIG. 2 represents the weight, and b represents the bias of the neuron.

Each individual commercial complex building is a specific neural network structure, that is, each individual contains an initial weight threshold for the BP neural network, and each individual calculates the fitness value via a fitness function. Firstly, an optimal initial weight threshold value is selected through selection, crossing and mutation operation by utilizing a genetic algorithm, and then is endowed to the BP neural network to start training, and at the moment, the initial weight threshold value of the BP neural network is set as the optimal initial weight threshold value obtained through optimization by utilizing the genetic algorithm.

And secondly, optimizing the mean square error by using a random gradient descent method to ensure the accuracy of the mean square error. And meanwhile, the system can automatically train according to the training set, the verification set and the test set divided by the data set until the model generalization performance is relatively best, and stop training, obtain the model and store the model. The trained neural network model can be used for optimizing and predicting the electric load of the commercial complex building, and the flow diagram of the improved neural network is shown in fig. 3. The flow of the improved neural network shown in fig. 3 is: inputting genetic algorithm parameters, generating an initial population, calculating an objective function value of the initial population, crossing and mutating, calculating an objective function value of a new population, updating the population and generating a child population, judging whether termination conditions are met, if so, training the BP neural network by using the optimal initial weight and bias, and outputting a trained BP neural network; if not, returning to execute the crossing step. The trained BP neural network is the subsequent load capacity prediction model.

Step 4: and acquiring a load capacity prediction model by utilizing the BP neural network after the data set training assignment.

The thermal coefficient reflects the climate environment of different areas in China, and the difference of the climate environment leads to different electricity utilization habits of different areas, so that obvious electricity utilization peaks can appear at a certain moment in the operation process of the transformer. In order to study the relation between the thermal coefficient k' of different areas and the power load and transformer capacity of the commercial complex building, five types of thermal areas of the original data are respectively input into the assigned BP neural network for training and prediction, and the training result of the assigned BP neural network is shown in figure 4.

As shown by the training result, the optimal objective function value of the trained BP neural network is 0.287, the training fitting error is 0.155, and the test training errors of the power load and the reactive power loss under the unit area are above 95%, which indicates that the trained BP neural network has higher training precision and accuracy.

Step 5: and inputting the thermal coefficient, the building height and the functional ecological area of the future newly-built commercial complex building into the load capacity prediction model to predict the load configuration data of the future newly-built commercial complex building.

After the BP neural network is trained, prediction analysis is carried out on unit area power load, air conditioning load, illumination load, reactive power loss and transformer capacity configuration values in different thermal areas, and the prediction results are shown in tables 1 to 5.

TABLE 1 Power load prediction error per unit area

TABLE 2 prediction error of air conditioner load per unit area

TABLE 3 prediction error of illumination load per unit area

TABLE 4 reactive loss prediction error per unit area

Table 5 prediction error of transformer capacity per unit area

The prediction results show that the prediction of the power load, the air conditioning load, the lighting load, the reactive power loss and the transformer capacity in unit area has higher training precision under different thermal coefficients, and the training precision is below 5%.

Step 6: according to the thermal coefficient, building height and functional ecological area of the built commercial complex building, the load capacity prediction model is used as an objective function of an improved particle swarm algorithm based on random distribution time lag, and optimal load configuration data of the built commercial complex building is determined; the improved particle swarm algorithm based on random distribution time lag introduces random distribution time lag in a speed update model.

In order to further ascertain the relationship between the thermal coefficient k', the building height h, the functional ecological area and the actual power load of the commercial complex building and the transformer capacity, the design of the existing load capacity is optimized, and an Improved particle swarm algorithm (Improved PARTICLE SWARM Optimization with Randomly distributed TIME DELAY, RDTD-PSO) based on random distribution time lag is established. The improved particle swarm algorithm abstracts each commercial complex building into 'particles' without mass and volume, each particle individual has own initial random position and speed before algorithm iteration starts, and each particle in the population can carry out information interaction, so that the individual of the whole population can be close to the particle individual with the optimal position on the premise of retaining the diversity characteristic of the individual. However, due to the complexity of the electrical load data of the commercial complex building, the traditional basic particle swarm algorithm is easy to converge too fast, so that the defects of local minimum value, population diversity reduction and the like are easily caused. The main novelty of the improved particle swarm algorithm based on randomly distributed time lags is the introduction of randomly distributed time lags into the velocity update model. More specifically, a certain number of historical individual best particles and global best particles are randomly selected according to the evolution state. Compared with the traditional delayed particle swarm algorithm, the delay term of the improved particle swarm algorithm is selected by randomly multiplying 0 or 1 by two random numbers, so that random distribution time lag is introduced in the population speed updating process, the random delay can better utilize the accumulated population to search the evolution history, the overall accuracy of the algorithm is further improved, the whole population has stronger capability of avoiding the problem of trapping in local optimal search, and the population keeps proper balance between convergence and diversity.

Compared with the traditional basic particle swarm algorithm, the speed and position updating, inertia weight and acceleration coefficient updating process of the RDTD-PSO algorithm is shown in the following formula.

(2)。

(3)。

(4)。

(5)。

The formula (2) is a speed update model, the formula (3) is a position update model, the formula (4) is an inertia weight update model, and the formula (5) is an acceleration coefficient update model.

Wherein,And/>Respectively representing the maximum value and the minimum value of the inertia weight; item represents the current iteration number and maxiter represents the maximum iteration number. In general, the larger the inertial weight is, the better the global exploration, and the smaller the inertial weight is, the better the local development is. Acceleration coefficient/>And/>Updated according to equation (5)/>And/>A first acceleration coefficient and a second acceleration coefficient which are distributed time lag terms and are respectively equal to/>、/>I.e./>=/>And/>=/>。/>Representing the initial value of the cognitive acceleration coefficient of the ith particle,/>The initial value of the social acceleration coefficient of the i-th particle is represented. /(I)Represents the final value of the cognitive acceleration coefficient,The final value of the social acceleration coefficient is represented. It is noted that/>=2.5，/>=0.5，/>=0.5，/>=2.5 Is determined from a number of experimental experiences. N is the upper bound of the distributed time delay; /(I)An n-dimensional vector is declared, wherein each element is randomly selected from 0 or 1; /(I)、/>、/>And/>Is a random number uniformly distributed in [0,1 ]; /(I)Intensity factor representing individual distribution delay term according to evolution state ζ,/>Representing the intensity factor of the global distributed delay term according to the evolution state ζ; /(I)、/>The speeds of the ith particle in k+1 iterations and k iterations are respectively represented; /(I)Representing the optimal position of the ith particle in k iterations,/>Indicating that the ith particle is at k-/>Optimal position of the secondary iteration; /(I)Represents the position of the particle globally optimal in k iterations,/>Represented in k-/>Iterating the position of the globally optimal particle; representing the current position of the ith particle at k iterations; /(I) Representing the current position of the ith particle at k+1 iterations. The term following the second row of plus signs in equation (2) is an individual distribution delay term, and the term following the third row of plus signs in equation (2) is a global distribution delay term.

In the RTDT-PSO algorithm proposed by the present invention, the speed and position are updated according to the evolution state, which depends on the evolution factor. By evolution factors/>To reveal the search features of the RTDT-PSO algorithm.

First, since the evolution factor is calculated based on the distance between each individual particle, the average distance between the ith particle and the other particlesThe expression is as follows.

(6)。

Where S represents the particle population size and D represents the particle size. Subsequent evolution factorsExpressed by formula (7).

(7)。

Wherein the method comprises the steps ofRepresenting the distance between the current location and the global best particle, the subscript g represents the global best particle in the particles,/>And/>Representing the minimum and maximum values, respectively, of the distance between particles in the population of particles. /(I)AndIs composed of evolution factor/>Determined/>From the distance between the particles. Evolution factor/>And/>、The relationship between them is shown in Table 6-Abbreviated as/>，/>Abbreviated as/>。

Attention is paid here to the number of delay iterationsRelationship with k iterations: when/>At < k >, the velocity update model is executed according to (1), otherwise set/>=0. I.e. when/>When not less than k,/>The value is 0, and the speed update model is as follows.

。

On the other hand, the selection of the inertial weights and acceleration coefficients is critical to the implementation of improved particle swarm algorithms, and the balance of global and local search performance is obtained by adjusting the inertial weights. In the invention, the selection of the inertia weight is linearly reduced from a larger value to a smaller value, so that the improved particle swarm algorithm has stronger global searching capability at the beginning of operation and stronger local searching capability near the end of operation.

As a result of the addition of the distributed time-lapse term containing past personal and global best particle history information. Therefore, the PTDT-PSO algorithm can evade local optimization, and the search space is explored and utilized more thoroughly than the classical particle swarm algorithm. The flow of the improved particle swarm algorithm based on a randomly distributed time lag is shown in FIG. 5.

The workflow of the improved particle swarm algorithm based on randomly distributed time lags shown in FIG. 5 specifically comprises the following steps.

6.1: Initializing improved particle swarm algorithm parameters. Namely: the size of the particle population, the initial position of the particles and the initial velocity of the particles are initialized.

6.2: Evaluating population fitness of all particles, updating/>And saved as history information. Namely: the fitness of each particle is calculated.

6.3: The average distance for each particle is calculated. Namely: updating the optimal position of each particle and the position of the globally optimal particle according to the fitness; the distance between the particles is calculated.

6.4: Calculating individual evolution factors of each particle, namely: the evolution factor is calculated from the distance between the particles.

6.5: The current evolution state is determined. Namely: determining the state of evolution based on the evolution factorIs used for the intensity factor of the distributed time delay term.

6.6: The inertial weights are updated.

6.7: And updating the acceleration coefficient. Namely: and updating the cognitive acceleration coefficient and the social acceleration coefficient.

6.8: Updating the randomly occurring distributed delay information. Namely: and updating the randomly generated distributed delay information according to the intensity factor, the updated inertia weight, the updated cognitive acceleration coefficient and the updated social acceleration coefficient.

6.9: Update speed and location. Namely: and updating the speed and the position of the particles according to the updated random distributed delay information, the speed updating model and the position updating model.

6.10: K″=k+1. Namely: the value of the iteration number is increased by 1.

6.11: K″ is up to a maximum number of iterations.

6.12: If the number of iterations after the increase of the value is smaller than the maximum number of iterations, sub-step 6.2 is performed.

6.13: If the iteration times after the value increase is equal to the maximum iteration times, ending, and taking the position of the updated globally optimal particle as the optimal position of the particle.

The invention takes five standard test functions as objective functions, tests the improved particle swarm algorithm based on random distribution time lag, measures the improved performance of the algorithm by recording the mean value and relative error, and the test result is shown in table 7.

Table 7 comparison of test function results

As can be seen from the comparison results of the test functions, the optimizing precision and accuracy of the RTDT-PSO algorithm are superior to those of the PSO algorithm, and the global optimal solution can be effectively searched in the optimizing process of the five test functions, so that the defect that the PSO algorithm is easy to be trapped into local optimal is avoided. The optimized convergence curves of the five test functions are shown in fig. 6 to 10, the PSO algorithm is the basic particle swarm algorithm in fig. 6 to 10, and the RTDT-PSO algorithm is the modified particle swarm algorithm in fig. 6 to 10.

As can be seen from the convergence curves shown in fig. 6 to 9, the Sphere function, rosenbrock function, ackley function, RASTRIGIN function all have a faster decreasing amplitude and approach to the optimal solution at the early stage of the improved particle swarm algorithm iteration. Meanwhile, the method also has larger descending amplitude in the later iteration stage, so that the improved particle swarm algorithm has higher searching efficiency in the global, and the defect that the basic particle swarm algorithm is easy to fall into local optimum is avoided.

In addition, although the Griewank function shown in fig. 10 has a similar drop amplitude in the early stages of the iteration, the drop amplitude in the later stage modified particle swarm algorithm is significantly greater than that of the basic particle swarm algorithm. This means that the global searching efficiency of the improved particle swarm algorithm in the later iteration period is higher than that of the basic particle swarm algorithm, so that the improved particle swarm algorithm has higher optimizing precision.

By comparing simulation results of the five test functions, the basic particle swarm algorithm is more prone to be trapped into a local extremum, a global space cannot be effectively explored in the problem of multi-dimensional commercial complex power supply load optimization, and convergence performance is poor. The convergence speed of the improved particle swarm algorithm in the later stage of the algorithm is larger than that of the standard algorithm, and the global optimal solution can be searched through rapid convergence, so that the convergence effect of the algorithm is improved, and the situation of sinking into local optimal is effectively avoided. Therefore, the improved strategy based on the distribution time lag has a great improvement on the algorithm performance of particle swarm.

The training and prediction flow of the BP neural network optimized by the genetic algorithm can be known, and the BP neural network after training can be used as an objective function of the improved particle swarm algorithm proposed by the chapter, so that the optimization work of the power load, the transformer capacity and the reactive power loss design value of the commercial complex building is completed. The optimal load capacity allocation and improved particle swarm algorithm convergence curves of the target commercial complex building under different thermal regions (severe cold region, summer hot winter warm region, mild region) are sequentially shown in fig. 11 to 15.

As can be seen from the convergence curves shown in fig. 11 to 15, in the transformer capacity optimization process under the thermal region corresponding to k' =0.95, the improved particle swarm algorithm always maintains a larger descending amplitude in the iterative updating process, the searching efficiency of the global space is far higher than that of the basic particle swarm algorithm, and finally the algorithm searches for an optimal value after 60 iterations. The searching efficiency of the global space is high. The descending amplitude of the thermal regions of k '=1 and k' =1.06 at the initial stage of iteration is far larger than that of a standard algorithm, the convergence effect of the improved particle swarm algorithm is better, and the global searching efficiency is higher at the later stage of iteration. In the optimizing process of k' =1.13 and the thermotechnical area, the basic particle swarm algorithm has a faster descending amplitude at the initial stage of iteration, has stronger capacity of approaching to the optimal solution, but has lower convergence performance at the later stage of iteration, and has lower searching efficiency on the global space, so that the basic particle swarm algorithm is finally trapped into a local minimum value.

As can be seen from fig. 11 to fig. 15, compared with the basic particle swarm algorithm, the improved particle swarm algorithm has better searching performance in the global domain, and the improved algorithm effectively increases the convergence rate of the algorithm, effectively avoids the problem that the population is easy to fall into local optimum, ensures the diversity of the population, and has remarkable performance improvement for solving the multi-objective electricity load optimization problem of the commercial complex building.

Meanwhile, the improved particle swarm algorithm gives the design configuration values of the optimal transformer capacity under different thermal regions, and the building heights and main functional performance areas of the corresponding commercial complex buildings are shown in the table 8. In view of the accuracy of the BP neural network model training and prediction based on the genetic optimization algorithm, the output result has higher accuracy in the electric design stage of the commercial complex building and has stronger reference in the electric design stage.

Table 8 optimal transformer capacity configuration and building scale

The purpose of step 5 prediction is to create a commercial complex building service for the future, while the object of step 6 optimization is the established commercial complex building.

The advantages of the above method can be expressed as follows.

1. The method provides a new thermal coefficient k' concept, and a thermal region is added into a prediction optimization model in a definite numerical form through a statistical analysis method, so that the thermal region is used as a new influence factor to help to know the change rule of the electric load of the commercial complex building more clearly and optimize the accurate training of the prediction model.

2. The method establishes a BP neural network model based on genetic algorithm optimization, and establishes a load capacity prediction model which takes a thermal coefficient k', a building height h and each functional state area of a commercial complex building as input quantity and takes electricity load (lighting load, power load and air-conditioning load), transformer capacity and reactive loss as output quantity, wherein the model has higher prediction accuracy and the precision is within 5 percent.

Meanwhile, the trained BP neural network can be used as an objective function, and a convenient program window can be established to be applied to power supply and distribution design of a commercial complex building based on the objective function, and the program window is shown in fig. 16.

In the design process of the newly built commercial complex building project, engineers can directly input parameters such as a thermal coefficient k ' of a target building, a building height h, various functional status areas and the like at the following interfaces, and can directly output power load, air-conditioning load, lighting load, reactive power loss and transformer capacity design values under a unit area by clicking a ' prediction ' button. Because of the accuracy of neural network model training, each output value is theoretically closest to the actual running condition of the commercial complex building, and has a stronger reference basis.

3. Compared with the traditional basic particle swarm algorithm, the improved particle swarm algorithm based on the random distribution delay has a better convergence effect even in the multi-dimensional problem optimization of the commercial complex building, and meanwhile, can avoid sinking into local optimum.

Meanwhile, by calling the trained BP neural network, the power consumption load capacity of the commercial complex building is optimally configured, and the commercial complex building height with the optimal power consumption load, transformer capacity and minimum reactive power loss and each main function state area under the target thermal region can be output. Because BP neural network training and prediction have higher precision, commercial complex building data and each power consumption load capacity configuration output by the improved particle swarm algorithm have higher accuracy, and have stronger reference value in the electric design stage.

Example 2

A computer apparatus, comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor executing the computer program to perform the steps of the commercial complex building power supply system optimal configuration method in embodiment 1.

Example 3

A computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the commercial complex building power supply system optimization configuration method in embodiment 1.

Example 4

A computer program product comprising a computer program which when executed by a processor performs the steps of the commercial complex building power system optimization configuration method of embodiment 1.

Example 5

A computer device, the internal structure of which may be as shown in fig. 17. The computer device includes a processor, a memory, an Input/Output interface (I/O) and a communication interface. The processor, the memory and the input/output interface are connected through a system bus, and the communication interface is connected to the system bus through the input/output interface. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer device is used to store the pending transactions. The input/output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by the processor to implement the commercial complex building power supply system optimization configuration method in embodiment 1.

The object information (including, but not limited to, object device information, object personal information, etc.) and the data (including, but not limited to, data for analysis, stored data, presented data, etc.) according to the present invention are information and data authorized by the object or sufficiently authorized by each party.

Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, database, or other medium used in embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, high density embedded nonvolatile Memory, resistive random access Memory (ReRAM), magneto-resistive random access Memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric Memory (Ferroelectric Random Access Memory, FRAM), phase change Memory (PHASE CHANGE Memory, PCM), graphene Memory, and the like. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, and the like. By way of illustration, and not limitation, RAM can be in the form of a variety of forms, such as Static Random access memory (Static Random access memory AccessMemory, SRAM) or dynamic Random access memory (Dynamic Random Access Memory, DRAM), and the like. The databases referred to in the embodiments provided herein may include at least one of a relational database and a non-relational database. The non-relational database may include, but is not limited to, a blockchain-based distributed database, and the like. The processor referred to in the embodiments provided in the present invention may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, or the like, but is not limited thereto.

The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.

The principles and embodiments of the present invention have been described in detail with reference to specific examples, which are provided to facilitate understanding of the method and core ideas of the present invention; also, it is within the scope of the present invention to be modified by those of ordinary skill in the art in light of the present teachings. In view of the foregoing, this description should not be construed as limiting the invention.

Claims (10)

1. An optimization configuration method for a commercial complex building power supply system is characterized by comprising the following steps:

Defining the value of the thermal coefficient of the summer hot and winter cold region as 1 by taking the transformer capacity average value of the commercial complex buildings in the summer hot and winter cold region as a reference value, and determining the ratio of the transformer capacity average value of the commercial complex buildings in each thermal region except the summer hot and winter cold region to the reference value as the value of the thermal coefficient of each thermal region except the summer hot and winter cold region;

Constructing a data set by taking the thermal coefficient, the building height and the functional ecological area of the built commercial complex building in different thermal areas as input quantity and the load configuration data in unit area as output quantity; the load configuration data includes power load, air conditioning load, lighting load, reactive power loss and transformer capacity;

Optimizing an initial weight threshold of the BP neural network by adopting a genetic algorithm, obtaining an optimal initial weight threshold, and assigning the optimal initial weight threshold to the BP neural network;

Acquiring a load capacity prediction model by utilizing the BP neural network after the data set training assignment;

Inputting the thermal coefficient, the building height and the functional ecological area of the future newly-built commercial complex building into the load capacity prediction model to predict load configuration data of the future newly-built commercial complex building;

According to the thermal coefficient, building height and functional ecological area of the built commercial complex building, the load capacity prediction model is used as an objective function of an improved particle swarm algorithm based on random distribution time lag, and optimal load configuration data of the built commercial complex building is determined; the improved particle swarm algorithm based on random distribution time lag introduces random distribution time lag in a speed update model.

2. The method for optimizing configuration of a power supply system for a commercial complex building according to claim 1, wherein the method for optimizing configuration of a power supply system for a commercial complex building according to claim 1 is characterized in that a thermal coefficient, a building height and a functional area of the commercial complex building built in different thermal areas are used as input amounts, load configuration data per unit area is used as output amounts, and a data set is constructed by:

unifying the units of the load configuration data, and representing the load configuration data in the form of unit area;

Outliers in the functional stateful area and outliers in the load configuration data are removed.

3. The optimal configuration method for the commercial complex building power supply system according to claim 1, wherein the BP neural network comprises an input layer, an hidden layer and an output layer which are sequentially connected;

the number of neurons of the input layer is 6;

the number of neurons of the hidden layer is 10, and the activation function of the hidden layer is a tanh function;

The number of neurons of the output layer is 6, and the activation function of the output layer is a linear activation function.

4. The method for optimizing and configuring a commercial complex building power supply system according to claim 1, wherein the speed update model of the improved particle swarm algorithm based on the random distribution time lag is as follows:

；

Wherein, 、/>The speeds of the ith particle in k+1 iterations and k iterations are respectively represented; w represents an inertia factor; /(I)Representing the cognitive acceleration coefficient,/>Representing social acceleration coefficient,/>And/>First and second acceleration coefficients representing a distributed time lag term,/>=/>And/>=/>；/>、/>、/>And/>Is a random number uniformly distributed in [0,1 ]; /(I)Representing the number of delay iterations; /(I)Representing the optimal position of the ith particle in k iterations,/>Indicating that the ith particle is at k-/>Optimal position of the secondary iteration; /(I)Represents the position of the particle globally optimal in k iterations,/>Represented in k-/>Iterating the position of the globally optimal particle; /(I)Representing the current position of the ith particle at k iterations; /(I)AndRespectively representing the intensity factors of individual distribution delay terms and global distribution delay terms according to the evolution state xi; /(I)Representing an n-dimensional vector, each element in the n-dimensional vector being randomly selected from 0 or 1; n represents the upper bound of the distributed delay;

The position update model of the improved particle swarm algorithm based on the random distribution time lag is as follows:

；

Wherein, Representing the current position of the ith particle at k+1 iterations;

The inertia weight updating model of the improved particle swarm algorithm based on the random distribution time lag is as follows:

；

Wherein, And/>Respectively representing the maximum value and the minimum value of the inertia weight; item represents the current iteration number, maxiter represents the maximum iteration number;

The acceleration coefficient updating model of the improved particle swarm algorithm based on the random distribution time lag is as follows:

；

；

Wherein, Representing the final value of the cognitive acceleration coefficient,/>Representing the initial value of the cognitive acceleration coefficient of the ith particle,/>Representing the final value of the social acceleration coefficient,/>The initial value of the social acceleration coefficient of the i-th particle is represented.

5. The method for optimizing configuration of a commercial complex building power supply system according to claim 4, wherein the speed update model isAnd/>The determining method of (1) comprises the following steps:

According to the formula Calculating a distance between the particles; wherein/>Represents the average distance between the ith particle and other particles, s represents the particle population size, D represents the particle size,/>Representing the current position of the jth particle in k iterations;

Based on the distance between particles, using the formula Calculating an evolution factor; wherein/>Representing evolution factors,/>Representing the distance between the current location and the global best particle,/>And/>Representing the minimum and maximum values, respectively, of the distance between particles in the population of particles;

When the value of the evolution factor is less than 0.5, And/>All take values of 0;

when the value of the evolution factor is greater than or equal to 0.5, And/>The average value is 0.01.

6. The method for optimizing configuration of a commercial complex building power supply system according to claim 4, wherein,

When (when)At < k, the velocity update model is

；

When (when)When not less than k,/>The value is 0, and the speed update model is

。

7. The method for optimizing configuration of a commercial complex building power supply system according to claim 5, wherein the workflow of the improved particle swarm algorithm based on randomly distributed time lags specifically comprises:

initializing the size of the particle population, the initial position of the particles and the initial velocity of the particles;

Calculating the fitness of each particle;

Updating the optimal position of each particle and the position of the globally optimal particle according to the fitness;

Calculating a distance between the particles;

Calculating an evolution factor according to the distance between the particles;

Determining the intensity factor of a distributed time delay term according to the evolution state xi according to the evolution factor;

Updating the inertia weight;

updating the cognitive acceleration coefficient and the social acceleration coefficient;

updating the randomly generated distributed delay information according to the intensity factor, the updated inertia weight, the updated cognitive acceleration coefficient and the updated social acceleration coefficient;

Updating the speed and the position of the particles according to the updated random distributed delay information, the speed updating model and the position updating model;

increasing the value of the iteration times by 1;

if the iteration number after the value is increased is smaller than the maximum iteration number, executing the step of calculating the fitness of each particle;

If the iteration times after the value increase is equal to the maximum iteration times, ending, and taking the position of the updated globally optimal particle as the optimal position of the particle.

8. A computer apparatus, comprising: a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to perform the steps of the method for optimizing the configuration of a commercial complex building power supply system according to any one of claims 1-7.

9. A computer readable storage medium having stored thereon a computer program, characterized in that the computer program when executed by a processor implements the steps of the method for optimizing the configuration of a commercial complex building power supply system according to any one of claims 1-7.

10. A computer program product comprising a computer program which, when executed by a processor, implements the steps of the method for optimizing the configuration of a commercial complex building power supply system according to any one of claims 1-7.