I suggest that the first step in determining the fixed portion of a mixed cost (a cost that is partially fixed and partially variable) is to graph the data. Label the vertical or y-axis of the graph as Total Manufacturing Overhead (or Total Electricity Cost if you are analyzing the individual components of overhead). Label the horizontal or x-axis of the graph as total machine hours (or some other indicator of volume). Then put a point on the graph for each of the past 12 months. If January had $100,000 of overhead and 6,000 MHs, enter a point on the graph where those two amounts intersect. After plotting all 12 months, you might see a pattern and/or you might see something unusual. Investigate the unusual before proceeding. Perhaps there was an accounting error or a very unusual situation that is not likely to recur. Once you are confident that the data is reasonable you can proceed. (Please note that the overhead costs need to be on an accrual basis. In other words, you need to have the costs that occurred when the machine hours occurred. Your data will be misleading if you relate this month's machine hours to the electricity bill that was paid this month, if that bill is for the previous month's electricity usage. Check the meter reading dates on the bill to be sure the dates of the electricity usage agree with the dates of the machine hours.)
Once you eliminated any "outliers" from the graph (or scattergraph) and you are confident that the dollar amounts are related to the activity on the x-axis, you can proceed. One technique is the High-Low method. This method uses only two of the points or months on your graph: the point where the MHs were the highest and the point where the MHs were the lowest...so long as those two points are not outliers. Let's assume that the highest number of MHs occurred in September. At that point the MHs were 10,000 and the total overhead cost was $140,000. Let's assume that the lowest activity occurred in January when the MHs were 6,000 and the total overhead cost was $100,000. If the range of these MHs (6,000 to 10,000) are not outside the range of normal activity, we know by definition that the fixed manufacturing costs will not change in this range as the MHs change. That means that any change in the total manufacturing costs in this range must be the change in the variable costs. Here's the formula: Variable Cost Rate = Change in total costs divided by the change in the MHs. In our example, Variable Cost Rate = $40,000 ($140,000 minus $100,000) divided by 4,000 MHs (10,000 MHs minus 6,000 MHs) equals $10 per MH. The total variable costs at the lowest activity = 6,000 MH times $10 variable rate = $60,000. Since the total overhead costs were $100,000 the fixed costs must have been $40,000. At the highest level of activity, the total variable costs = 10,000 MHs X $10 = $100,000. Since the total costs at this level were $140,000 the fixed costs must be $40,000. So we estimate (based on just two points) that the manufacturing overhead is $40,000 per month + $10 per MH.
A more sophisticated method for separating the fixed costs from the mixed costs would be to use all of the points (rather than only the highest and lowest). The technique to accomplish this is regression analysis. The use of regression analysis not only gives you the best equation for all the points, it also generates statistics on how much of the change in manufacturing overhead is caused by the MHs. Undoubtedly you will find that MHs are causing only part of the change in the overhead costs. This feedback is important if it prompts you to identify the other variables that are causing the manufacturing overhead costs to change.
In summary, the High-Low method is an overly simplistic tool. Be careful and spend time understanding the relationships between activities and costs. Manufacturing processes and costs are probably driven by many activities (not just MHs) and not all products may require the same activities.