Grid investment capability prediction based on path analysis and BP neural network

. With the more complex investment environment of China's power grid, the accurate prediction of the investment ability of power grid enterprises has become an important prerequisite for managers to make precise investment decisions. This paper first selects the factors affecting the investment capacity of the power grid from the internal and external environment, and establishes the index system of the factors affecting the investment capacity. Secondly, the path analysis is used to deeply explore the interaction relationship and influence degree of each index and investment capacity. Finally, the maximum investment capacity of the power network can be predicted based on the BP neural network prediction model. The results show that the BP neural network prediction model can achieve higher prediction accuracy when predicting the power grid investment capability.


Introduction
The environment for grid investment has become increasingly complex in the context of the dual carbon target and new power system transformation.Accurate prediction of the grid's investment capacity is a prerequisite for grid companies to make investment decisions.Therefore, in order to make an accurate investment capacity assessment of the grid in the complex context, a set of simple and comprehensive index system and a scientific prediction model are needed to forecast the grid investment capacity, so that the grid can make a reasonable investment strategy in the new situation.At present, some literature has made preliminary exploration on the influencing factors of investment capacity from qualitative and quantitative perspectives.Qualitatively, literature [1] analyzed the investment capacity of enterprises from two perspectives of investment efficiency and investment scale of grid enterprises, and summarized the main influencing factors.The literature [2] investigates the influence of management power on corporate investment, and the results show that power intensity is significantly and positively related to the level of corporate investment.Quantitatively, the literature [3] used the TOPSIS method to analyze the excellent level of investment capacity of grid companies in each year.The literature [4], on the other hand, investigates the dynamic feedback relationship between the amount of grid investment and investment efficiency based on system dynamics.The literature [5] verified the influence mechanism between grid investment and macro-environmental factors based on co-integration theory and vector error correction model.In order to accurately grasp the investment capacity of grid enterprises, some studies forecast the grid investment capacity.The literature [6] proposed a multi-region grid investment capacity measurement method based on assetliability ratio, which provides a basis for the investment plan of grid enterprises.The literature [7] constructs an investment capacity measurement model adapted to the power system reform through comprehensive plan balance optimization, and reasonably allocates the investment of each project.Compared with the above traditional forecasting model, the intelligent forecasting algorithm can avoid the errors caused by subjectivity.The literature [8] constructed a provincial grid investment scale prediction model based on MLR and REF neural network, and the model prediction error is small.The literature [9], on the other hand, used an improved support vector machine model to forecast and analyze the investment capacity of a power grid company.The literature [10] performs a variational modal decomposition and forecasts it using multiple support vector machine models.A small part of the literature predicts the grid investment capacity based on two and more models.The literature [11] proposed a LSTNET-ship2 network model for grid investment capacity prediction by applying a combination of convolutional networks, CRU networks, and autoregressive models.The literature [12] proposed a combined forecasting model combining gray forecasting model, BP neural network and multiple regression model, thus combining the advantages of various methods to forecast the grid investment scale and further improving the forecasting accuracy.From the above literature analysis, it can be obtained that there may be serious coupling relationship among the factors of grid investment capacity.Therefore, when selecting the influencing factors, the original information is retained as much as possible, so that the simplified influencing factors truly reflect the level of grid investment capacity.And a suitable intelligent prediction model is selected to improve the accuracy of grid investment capacity prediction.Therefore, this paper firstly selects the factors affecting the grid investment capacity from the internal and external environment, and analyzes the mutual influence relationship and the degree of influence between each index and the investment capacity.Finally, the maximum investment capacity of the grid is predicted based on the BP neural network prediction model.The prediction results show that the proposed prediction method can predict the grid investment capacity more accurately.

Construction of index system
The main purpose of the analysis of investment ability influencing factors is to screen and identify the main influencing factors and lay the foundation for the investment ability measurement model.This paper identifies the main influencing factors of investment capacity by combining qualitative and quantitative methods.

Analysis of the influencing factors
This paper comprehensively constructs the influencing factor index system of power grid investment capacity from both internal factors and external factors.Internal factors are divided into business status and management status.External factors are divided into market environment, economic environment and policy environment.The index system of the factors influencing the investment capacity is shown in Table 1.

Diameter analysis
Make the model correction using SPSS Amos, and finally get the investment capacity model path map as shown in Figure 1.The total effect of the factors on the investment capacity should be the sum of the direct diameter coefficient and the indirect diameter coefficient, and only consider the direct diameter coefficient is not comprehensive.Therefore, in order to make the results more reasonable and reliable, the investment ability modeling should be composed of internal financial indicators and external factor correction coefficient.

BP neural network prediction model
The relevant factors affecting the grid investment capacity are used as input elements to forecast the grid investment capacity.the learning of BP neural network consists of two main parts.
(1) Forward propagation of the input signal According to the BP neural network structure, the various unit input and output functions in the implied layer are as follows: 1 In formula: ij w j b is the weight between the i-th neurons in the input layer and the j-th neurons in the hidden layer; ) =1,2, , ; =1,2, ,r In formula: jk w is the weights between the j-th neuron in the hidden layer and the k-th neuron in the output layer; (2) Reverse propagation of error information The error of the BP network is calculated as follows: The reverse optimization process is to correct the network parameters, and the output layer correction algorithm is as follows: In formula: q k t and q k o represent the target and output values of the k-th neurons in the q-th sample output layer.

Case study
This paper takes the investment data of a municipal power grid as an example.First, key indicators are screened based on path analysis, and the interaction relationship between indicators is analyzed.Secondly, BP neural network model is used to predict the investment ability of power grid.Finally, and the prediction results are compared with the actual value of investment ability.(2) Analysis of investment ability prediction results

Sample selection and division
The investment capacity prediction results are shown in Table 3, and the actual investment pair is shown in Figure 4. Before 2016, Chongqing Electric Power Company's investment capacity showed a rapid growth trend and peaked in 2016; from 2017-2020, the investment capacity gradually decreased; in 2021 and 2022, the investment capacity gradually increased and then leveled off; from 2011-2024, the investment capacity can meet the investment demand.

Conclusion
This paper first selects the factors affecting the investment capacity of the power grid from the internal and external environment, and establishes the index system of the factors affecting the investment capacity.On this basis, the path analysis is used to analyze the interaction relationship between various indicators and investment capacity.Finally, the power grid maximum investment capacity is predicted based on the BP neural network prediction model.The results show that: (1) The actual investment of the power grid in each region is within the investment capacity, and the prediction method is practical and feasible.
(2) In recent years, power grid investment remains high but the overall downward trend.

Figure 1
Figure 1 Path analysis of key indicators of investment

SHS
Web of Conferences 151, 01043 (2022) https://doi.org/10.1051/shsconf/202215101043EMSD 2022 j b represents the output threshold of the neurons in the hidden layer; ( )   is a transfer function representing the hidden layer.By processing the multi-layer neurons, the output results of the output layer can be calculated: 1 ( Data on factors affecting investment capacity from 2011-2020 were used as a training set sample, and data in 2021 were used as a test set sample.

4. 2
Investment capacity forecast (1) Analysis of the model prediction effect After multiple training sessions, training set predicted values and training set scatter fitting results, as shown in Figures2 and 3, respectively.The predicted value of investment ability almost coincides with the expected value.The strong correlation between the predicted value and the true value, and the correlation coefficient is reaching 1.0000.

Figure 2
Figure 2 The predicted and actual values of the training set are compared

Figure 4
Figure 4 Comparison of actual investment amount and maximum investment capacity value

Table 1
Index system of influencing factors of investment capacity Using SPSS software, calculate the correlation coefficient between each influencing factor and the historical investment ability.If the Pearson coefficient is greater than 0.8, it indicates that the factor affects the investment ability greatly.Factors with a Pearson coefficient greater than 0.8 are shown in Table 2.